Chemically gated artificial nanochannels for programmable subcellular signal modulated transport regulation

Abstract

Biomimetic artificial nanochannels have been developed rapidly because of their promising potentials in biomedical applications. Here we report the design of chemically gated artificial nanochannels to perform transmembrane channel-mimetic permeability transition via the multi-functional DNA components modified at the inner surface, whereby the structure and charge of the DNA components can be tuned by multiple key chemical signals. We realize the targeted capture of a single mitochondrion in single living cells, which allows in situ response of multiple mitochondrial signals (Ca²⁺ / ROS / H⁺) and the subsequent delicate control of permeability transition. Further study of rotenone (ROT) induced ROS / Ca²⁺ release and mitochondrial membrane potential loss demonstrate that the nanochannels can response to complex chemical signals at a localized subcellular region in spite of the complicated intracellular environment. Finally, we report the advanced applications of nanochannels for evaluating and regulating the interaction network between mitochondria and other organelles.

Introduction

Transmembrane channels reside in lipid membranes on the interface between two aqueous compartments, where they are regulated to control the translocation of diverse solutes across membranes¹. In living systems, transmembrane channel permeability is programmably modulated by a combination of chemical signals, which enables stimuli-responsive transition of the transport performance to meet diverse physiological needs and adapt to environmental changes^2,3. For example, the plasma membrane-resident ion channels govern the transport of various ions into and out of living cells^4,5,6,7. Another example is the mitochondrial permeability transition pore (mPTP) located at mitochondrial membranes, whose opening can be regulated by various stimuli, inducing increased permeability of mitochondria and triggering different cellular responses⁸. MPTP is a nonselective channel that spans the inner and outer mitochondrial membranes (OMMs), facilitating signal transmission or the exchange of materials between the cytoplasm and the mitochondrial matrix; thus, mPTP is involved in multiple biological functions, including modulation of oxidative stress responses, regulation of Ca²⁺ homeostasis, and stimuli-induced protein translocation⁹. Under low-stress conditions, mPTP is usually in a closed state, allowing ions and small molecules to pass through and maintaining normal mitochondrial functions. In contrast, elevated intracellular Ca²⁺ levels, oxidative stress, and fluctuations in blood glucose concentration can trigger the opening of mPTP, resulting in increased membrane permeability. The long-term opening of mPTP enables the influx of water, solute, and Ca²⁺ flow into the mitochondrial matrix in large quantities, causing the swelling of mitochondria and the programmed cell death¹⁰. These functions enable mPTP to be regulated by multiple chemical signals. To create biomimetic ion channels, a range of functional materials have been employed, including structurally organized polymers, DNA-based nanopores, G-quadruplex structures, metal-organic cuboctahedra, and solid-state nanochannels^{11,12,13,14,15,16}. These artificial nanochannels, inspired by biological systems, are engineered to replicate selective transport across membranes.

In recent years, biological transmembrane channels have inspired the construction of biomimetic artificial nanochannels to achieve precise replication of key transport properties observed in natural systems¹⁷. For example, Liu et al. constructed a biomimetic voltage-sensitive chloride nanochannel by integrating solid-state synthetic nanopores with chloride-responsive components, which achieved the reversible switching between the “on” and “off” states upon the presence and absence of Cl⁻¹⁸. Cantley et al. reported that nanopores incorporated into graphene membranes exhibited K⁺ selectivity and could be modulated by an applied gate voltage¹⁹. Zhou et al. developed a synthetic light-responsive ion channel membrane by employing conjugated microporous polymers showing photo-switchable behavior and enabling light-controlled ion transport across the membrane²⁰. Ou et al. reported a photo-responsive styrylpyrene-modified MOF allowing repeated cycles of small molecule loading and release²¹. Keum et al. established a modular conductive MOF-gated field effect transistor biosensor array to discriminate different small molecule neurotransmitters in biological fluids²². These biomimetic artificial nanochannels have garnered significant interest owing to their tunable architecture, versatile surface functionalization, adjustable ionic transport behavior, and enhanced structural stability^23,24,25. Therefore, biomimetic artificial nanochannels have provided compelling alternatives to conventional protein nanopores, showing great potential in signal regulations and subcellular biosensing^26,27,28. However, it remains challenging to actuate nanochannels to achieve real-time response to the stimulation of complex chemical signals in a complicated intracellular environment. To overcome these limitations, DNA-based gating systems have emerged as potential strategies for constructing biomimetic transmembrane channels due to their excellent reprogrammability, stimulus logic integration, and biocompatibility. For example, Dey et al. reported a DNA-functionalized membrane channel that allowed reversibly regulated protein transport, enabling stimuli-responsive translocation of proteins across lipid bilayers²⁹.

Here, by drawing inspiration from transmembrane biological channels, mPTP is used as a specific reference that can help to reflect the programmable modulation behavior under multiple chemical inputs, our strategy is to design chemically gated artificial nanochannels based on functionalized glass nanopipettes. Glass nanopipettes, also known as solid-state glass nanopores, are an attractive class of building blocks of artificial nanochannels with significant benefits including high temporal and spatial resolution, simple modification, and ease of introduction into living cells³⁰. For example, Xiong et al. realized a fluidic memristor exhibiting neuromorphic functions through a polyelectrolyte-confined fluidic architecture constructed from glass nanopipettes³¹. Recently, our group established rectifying artificial nanochannels exhibiting multiple interconvertible permeability states by using inner surface-functionalized glass nanopipettes³². By incorporating DNAzyme-substrate molecular switching systems, precise regulation of channel permeability was achieved. Nevertheless, its application was limited due to the exogenous factor-induced control mechanism and the lack of exploration to perform targeted response or regulation at a localized subcellular region.

We now present the direct and targeted modulation of chemically gated artificial nanochannels across the membrane of living cells. Using mitochondria as the targets, the outer surface of the nanopipettes is modified by mitochondrial targeting peptide (MLS) to facilitate the selective capture of single mitochondria, which enables in situ modulation of the channel permeability in response to multiple chemical signals (ROS/H⁺/Ca²⁺) by employing multi-functional DNA components for the modification of the inner surface. And we show that ROT-induced ROS/Ca²⁺ release and mitochondrial membrane potential loss can be delicate transport regulation by using the artificial nanochannels. Finally, we demonstrate that the programmable subcellular signal-modulated transport regulation of artificial nanochannels can be used for evaluating and regulating the interaction network between mitochondria and other organelles. Overall, we provide an updated view of the design strategy of artificial nanochannels, enabling quick and accurate responses based on multiple external chemical inputs.

Results

Design of chemically gated artificial nanochannels

Transmembrane channels enable stimuli-responsive regulation of the translocation of diverse solutes across cell membranes or organelle membranes. Precise regulation of transmembrane channels plays a vital physiological role in controlling molecular transport and metabolic processes^1,33. A typical example is mPTP, also known as mitochondrial mega-channel or multi-conductance channel³⁴. Depending on the degree of conductance, mPTP can vary from a narrow channel state to a large channel state that allows high flux transportation of molecules³⁵. As shown in Fig. 1a, mPTP conformation change coordinates to its conductance state. Voltage-dependent anion channel (VDAC), the key component of mPTP at the OMM, transitions from low- to high-conductance state in response to the loss of mitochondrial membrane potential (ΔΨ)³⁶. Adenine nucleotide translocator (ANT), the key component of mPTP at the inner mitochondrial membrane (IMM), can be opened under the coordinated stimulation of Ca²⁺ ions and ROS (reactive oxygen species)³⁷.

Fig. 1: Chemically gated artificial nanochannels for programmable modulated transport regulation.

a Mitochondrial permeability transition pore (mPTP) opens upon the coordinated stimulation in nature. IMM inner mitochondrial membrane, OMM outer mitochondrial membrane, IMS intermembrane space, VDAC voltage-dependent anion channel, ANT adenine nucleotide translocator. b Design of biomimetic chemically gated artificial nanochannels based on functionalized glass nanopipettes. Seq-1 from 5′ to 3′: a linker part (Linker) for inner surface attachment, Ca²⁺ specific DNAzyme (CD), and a Hemin-bonded G-quadruplex DNAzyme part (HG4). HG4 can catalyze the ROS-mediated oxidation of ABTS to produce metastable radicals (ABTS^•+ at pH = 5.5 and ABTS^•− at pH = 7.4). Seq-2 from 5′ to 3′: a substrate part (Substrate) and a short random DNA sequence (RD). c Programmable modulation of the artificial nanochannels under the synergetic effect of multiple key chemical signals. From left to right: the current–voltage (I–V) curve measured in a single nanopipette without any treatment (Null), different ICR ratio evolutionary paths upon the treatment of input signals (different concentrations of Ca²⁺/ROS/H⁺), the current–voltage (I–V) curve measured in a single nanopipette after stimulation (Output). The adding order of chemical stimuli: Path-1 (P1), Ca²⁺, ROS, H⁺; Path-2 (P2), ROS, Ca²⁺, H⁺; Path-3 (P3), ROS, H⁺, Ca²⁺; Path-4 (P4), H⁺, ROS, Ca²⁺; Path-5 (P5), Ca²⁺, H⁺, ROS; Path-6 (P6), H⁺; Ca²⁺; ROS. Source data are provided as a Source data file.

In order to mimic the programmable modulation behavior of mPTP under multiple chemical inputs, our strategy is to design chemically gated artificial nanochannels constructed from functionalized glass nanopipettes (Fig. 1b)^38,39,40. The internal surface of the nanopipettes is functionalized by a delicately designed multi-functional DNA structure composed of two DNA sequences with different regulatory modules (Supplementary Table 1). Seq-1 consists of three functional parts: a linker part (Linker) for inner surface attachment, a Ca²⁺-dependent DNAzyme part (CD)^41,42, and a Hemin-bonded G-quadruplex DNAzyme part (HG4)⁴³. Seq-2 contains two functional parts: a substrate part (Substrate) corresponding to the CD part in Seq-1 and a short random DNA sequence (RD). Here, CD and Substrate are complementary sequences, thus Seq-1 and Seq-2 are hybridized to form Seq-1/Seq-2 dimers (dsDNA) at the internal surface of nanopipettes, and then the functionalized glass nanopipettes are filled with electrolyte solution containing 2,2′-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid (ABTS))⁴⁴. On one hand, in the presence of Ca²⁺, the substrate can be cut from the ribonucleotide cleavage site, leading to the release of the segment adjacent to the 3′-end of Seq-2. The introduction of RD at the 3′-end of Seq-2 can introduce more negative charges, thus helping to boost the ionic current rectification (ICR) change upon the cleavage of the substrate. On the other hand, HG4 can catalyze the ROS-mediated oxidation of ABTS into ABTS^•+ at acid pH or ABTS^•− at neutral/alkalinity pH. Based on this, the inner surface charge can be changed due to the combination effect of Ca²⁺ triggered Substrate cleavage and HG4 catalyzed generation of ABTS^•+/ABTS^•−, and this change can be revealed by observing ICR⁴⁵. The functionalization of the inner and outer surfaces of the nanopipette, and the formation of Seq-1/Seq-2 dimers in the nanopipette were shown in Supplementary Fig. 1.

By employing functionalized nanopipettes as a proof of concept, we seek to accomplish transmembrane channel-mimetic permeability control of the artificial nanochannels under the synergetic effect of multiple key chemical signals (Ca²⁺/ROS /H⁺). As shown in Fig. 1c, through the acquisition of current–voltage (I–V) curves from a nanopipette (Supplementary Table 2), the ICR ratio (R = |I_−1.0 V|/I_1.0 V) evolutionary paths and permeability characteristics of the nanochannels at different stages are depicted and explained according to the action order of different stimuli. For example, Path-1 (P1) shows the dynamic path of ICR changing in a single nanochannel. The order of adding chemical stimuli is 1. Ca²⁺, 2. ROS (H₂O₂), 3. H⁺. The original state of an artificial nanochannel without any treatment (pH = 7.4) is defined as Null. At the Null state, Seq-1/Seq-2 dimers were present at the inner surface of the nanopipette, and a negative rectification (R > 1.0) was observed, as shown by the first blue dot in Fig. 1c, P1. Next, the addition of Ca²⁺ led to the departure of the fragment near the 3′-end of Seq-2 (Supplementary Fig. 2), resulting in the decrease of the ICR ratio (the second blue dot in Fig. 1c, P1). After ROS treatment, the charge density at the inner surface of the nanochannel was enhanced from the accumulation of ABTS^•−, leading to a significant rise in the ICR ratio (the third blue dot in Fig. 1c, P1)⁴⁶. Finally, upon the addition of H⁺ (pH 5.5), the inner surface charge could be tuned from negative to positive due to the amphoteric character of the HG4 part, leading to an inversion of the ICR curve, with positive rectification (R < 1.0, the red dot in Fig. 1c, P1). Similarly, by changing the order of chemical stimuli, we could obtain different ICR ratio evolutionary paths (P2–P6). Starting with the same Null value (the original state), six different paths were observed to reach the same destination (output). Based on continuous monitoring of the I–V curves of a single artificial nanochannel, we are able to look into the invisible black box to identify each intermediate step.

Fabrication and characterization of chemically gated artificial nanochannels

Before DNA functionalization, bare glass nanopipettes were fabricated through laser pulling of quartz capillaries for subsequent construction of chemically gated artificial nanochannels. Here, the pulling parameters were tuned to obtain nanopipettes with different tip diameters (Supplementary Table 3). As shown in Supplementary Fig. 3, nanopipettes with different tip sizes (50, 150, and 500 nm) were prepared, and their corresponding I–V curves are shown in Supplementary Fig. 4. When selecting nanopipette tip diameters, two factors were taken into consideration. First, the tip size should not be too large, so as not to cause cell damage. Second, to ensure adequate immobilization of DNA sequences onto the inner surface of the nanopipettes, the tip size should not be too small. Taking these factors together, nanopipettes with an opening tip size of 150 nm were selected for the following DNA modification. As shown in Fig. 2a, the inner surface of the bare nanopipettes was first functionalized with piranha solution to introduce hydroxyl groups (control), followed by modification with 3-aminopropyl-triethoxysilane (APTES) to generate amino groups. Subsequently, glutaraldehyde (GA) was applied to introduce electrically neutral aldehyde groups. Next, glutaraldehyde (GA) was used to induce electrically neutral aldehyde groups, which could be used for the linkage of DNA sequences (tagged with amino groups) via Schiff-base reaction. SEM images of the side view and top view of a nanopipette showed a typical conical shape and an opening pore of 150 nm (Fig. 2b). From the I–V curves recorded in a single nanopipette during successive modification steps (Fig. 2c), APTES functionalization resulted in a positively rectified ionic current. Following GA conjugation, the I–V curve became linear with no rectification, suggesting the formation of an electrically neutral surface. Upon DNA immobilization, a pronounced negative ICR emerged, confirming the successful attachment of DNA sequences. Additionally, variations in surface charge density were assessed via Zeta potential measurements using silica nanoparticles as a model to mimic the nanopipette’s inner wall (Fig. 2d). The results of Zeta potential measurements were in accordance with ICR trends, which demonstrated the successful fabrication of the DNA-functionalized nanopipettes.

Fig. 2: Characterization and rectification behavior of the chemically gated artificial nanochannels.

a Schematic illustration showing the procedure of modification at the inner surface of the nanopipettes. Step 1: piranha treatment to obtain hydroxyl groups. Step 2: 3-aminopropyl-triethoxysilane (APTES) treatment to induce amino groups. Step 3: glutaraldehyde (GA) treatment to form electrically neutral aldehyde groups. Step 4: modification of amino group-labeled Seq-1/Seq-2 dimers. b SEM images of nanopipettes after modification. Left: top view. Right: side view. c Current–voltage (I–V) curves measured in a single nanopipette at different steps of the modification procedure. d Zeta potentials measured at different steps of the modification procedure. Here, silica nanoparticles were used to simulate the inner wall of the nanopipettes to perform the same modification procedure in (a). e Schematic illustration of the single-input/single-output (SISO) model. f SISO operation in the chemically gated artificial nanochannels. Input from left to right: Ca²⁺, ROS, H⁺. Output from left to right: I, II, III. g Rectification ratio measured in nanopipettes modified by Seq-1/Seq-2 upon the addition of different metal ions (Cu²⁺, Ba²⁺, Zn²⁺, Mg²⁺, Na⁺, Ag⁺, Fe³⁺ and all the metal ions were 2 µM). h Current–voltage (I–V) curves measured in the chemically gated artificial nanochannels before and after single-input stimuli corresponding to (f). Inset: ICR ratios corresponding to the I–V curves. Error bars in (c, d, g, h) represent the standard deviation of three independent experimental repeats, and the measure of the center represents their corresponding mean value. Significance of the ICR ratio was assessed by t-test (n = 3 independent experiments and the data were presented as the mean values ± SDs, *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). All statistics were calculated using two-tailed paired t-tests. Source data are provided as a Source data file.

Before operating the programmable modulation behavior in the artificial nanochannels, the length of Seq-1 was optimized by changing the number of HG4 (Supplementary Table 4). From I–V curves (Supplementary Fig. 5) of nanopipettes modified by CD-HG4 sequences with different numbers of HG4 (CDNs-HG4-1 with one HG4, CDNs-HG4-2 with two HG4, CDNs-HG4-3 with three HG4 parts), it could be seen that the CDNs-HG4-1 modified nanopipette showed a negative rectification even at the acidic pH, which was unfavorable for the discrimination of ICR changes. For the CDNs-HG4-2 modified nanopipette, there was no significant change between the current measured at pH 7.4 and that at pH 5.5 under −1.0 V. CDNs-HG4-3 modified nanopipette could clearly differentiate the ICR behaviors at different pH values. Thus, CDNs-HG4-3 with three HG4 parts was chosen for the following experiments. The concentration of DNA sequences was optimized to be 1 μM (Supplementary Fig. 6). Next, we examined the stability of DNA-modified nanopipettes by measuring the I–V curves at multiple time points under different pH values. As shown in Supplementary Fig. 7, the nanopipettes remained stable after 2 h. Control experiments using bare nanopipettes without any modification were also tested. As shown in Supplementary Fig. 8, there was no significant change between the I–V curves measured at different pH values, which indicated that the reversal of ICR in DNA-modified nanopipettes could be attributed to the amphoteric property of HG4. Moreover, the nanopipettes demonstrated reversibility over five consecutive cycles (Supplementary Fig. 9), confirming their ability to enable precise and reversible modulation of the inner surface charge.

To apply the obtained chemically gated artificial nanochannels, we used them to build input-and-output configurations for coupling the in situ response of chemical signals with permeability transition regulation. Here, we first checked the single-input/single-output (SISO) model in the chemically gated artificial nanochannels (Fig. 2e). Three different chemical inputs (Ca²⁺, ROS, H⁺) were tested. As shown in Fig. 2f, the input of Ca²⁺ (pH = 7.4) could activate the DNAzyme part (CD) to cut its substrate, leading to the departure of the fragment near the 3′-end of Seq-2. While control experiments using other metal ions (Cu²⁺, Ba²⁺, Zn²⁺, Mg²⁺, Na⁺, Ag⁺, Fe³⁺, 2 µM each) as input showed negligible changes of ICR ratio (Fig. 2g), which confirmed the specificity of Seq-1 to Ca²⁺. The above results induced a significant decrease in the ICR ratio upon the reduced negative charge of the DNA framework (Fig. 2h, I). The change of inner surface charge density in the nanopipette before and after the treatment of Ca²⁺ was verified by Zeta potential measurement using silica nanoparticles (Supplementary Fig. 10). Next, ROS (5.0 mM H₂O₂) was used as the single input. In the design of Seq-1, hemin-bonded G4-DNAzyme (HG4) was capable of catalyzing the H₂O₂-mediated oxidation of ABTS, exhibiting functionality similar to HRP but with enhanced stability (Supplementary Fig. 11)^47,48,49. Upon the oxidation of ABTS, ABTS^•− was generated, which could bind with DNA sequences with high affinity. In neutral or alkaline solutions, ROS treatment led to a marked enhancement in negative charge density on the nanopipette’s inner surface, caused by the accumulation of ABTS^•−, resulting in a significant increase of the ICR ratio (Fig. 2h, II). When using H⁺ as the input signal, the surface charge shifted from negative to positive (Fig. 2h, III), a behavior likely due to the amphoteric nature of G4. The variation in inner surface charge density under different pH conditions was confirmed through Zeta potential measurements employing silica nanoparticles (Supplementary Fig. 10). Collectively, these data demonstrate that the designed chemically gated artificial nanochannels can operate by sensing biologically diverse and important cellular chemical signals as input to yield corresponding output.

Programmable integration of signal modulation in chemically gated artificial nanochannels

To evaluate the modularity of the designed chemically gated artificial nanochannels, we tested the potential of more complex input-output configurations. As shown in Fig. 3a, a daisy chain configuration was built to perform the multiple-input/multiple-output (MIMO) model. In this model, one input induces a corresponding ICR response as the output, which then interacts with the next input to produce a new ICR output. In the MIMO model, the order of the input signals for an operational procedure does not affect the final ICR ratio, but it does affect the order in which the output objects appear. As illustrated in Fig. 3b, different orders of input signals yielded different evolutionary paths of the ICR ratio. Specifically, for Operational procedure-1 (I), the input of α (control) produced an original ICR response α* (null), which was then interacted with the next input β (Ca²⁺) to produce an ICR response (output β*). By analogy, the following input γ (ROS) produced the output γ*, and the final input δ (H⁺) produced the output δ*. The ICR ratio evolutionary paths measured under different operational procedures (I-VI) were depicted according to the action order of different input signals, with different permeability states (H₁-H₄) of the chemically gated artificial nanochannels. The ICR ratio evolutionary paths were shown on the far right in Fig. 3b. Path-1 (I): α* (ICR = 5.22), β* (ICR = 2.07), γ* (ICR = 2.50), δ* (ICR = 0.15). Path-2 (II): α* (ICR = 4.94), γ* (ICR = 6.37), β* (ICR = 2.67), δ* (ICR = 0.16). Path-3 (III): α* (ICR = 5.86), γ* (ICR = 6.80), δ* (ICR = 0.53), β* (ICR = 0.14). Path-4 (IV): α* (ICR = 6.20), δ* (ICR = 0.61), γ* (ICR = 0.27), β* (ICR = 0.13). Path-5 (V): α* (ICR = 5.56), β* (ICR = 2.84), δ* (ICR = 0.54), γ* (ICR = 0.13). Path-6 (VI): α* (ICR = 5.20), δ* (ICR = 0.76), β* (ICR = 0.27), γ* (ICR = 0.19). The ICR ratio evolutionary paths were shown in Supplementary Table 5. By depicting the I–V curves at different permeability states (Supplementary Figs. 12–18), the invisible black box of intermediate steps could be revealed.

Fig. 3: Complex input-output configurations in chemically gated artificial nanochannels.

a Schematic illustration of the multiple-input/multiple-output (MIMO) model. b Operation details of the multiple-input/multiple-output (MIMO) model. Operational procedure-1 (I): first input α, first output α*; second input β, second output β*; third input γ, third output γ*; final input δ, final outputδ*. Operational procedure-2 (II): first input α, first output α*; second input γ, second output γ*; third input β, third output β*; final input δ, final output δ*. Operational procedure-3 (III): first input α, first output α*; second input γ, second output γ*; third input δ, third output δ*; final input β, final output β*. Operational procedure-4 (IV): first input α, first output α*; second input δ, second output δ*; third input γ, third output γ*; final input β, final output β*. Operational procedure-5 (V): first input α, first output α*; second input β, second output β*; third input δ, third output δ*; final input γ, final output γ*. Operational procedure-6 (VI): first input α, first output α*; second input δ, second output δ*; third input β, third output β*; final input γ, final output γ*. The ICR ratio evolutionary paths of different operational procedures were shown on the far right. Path-1: α*, β*, γ*, δ*. Path-2 (II): α*, γ*, β*, δ*. Path-3 (III): α*, γ*, δ*, β*. Path-4 (IV): α*, δ*, γ*, β*. Path-5 (V): α*, β*, δ*, γ*. Path-6 (VI): α*, δ*, β*, γ*. Input α: null (without the presence of Ca²⁺/ROS/H⁺). Input β: Ca²⁺. Input γ: ROS. Input δ: H⁺. H: hidden state. H₁: nanopipette at the first hidden state (null). H₂: nanopipette at the second hidden state after the treatment of the second input. H₃: nanopipette at the third hidden state after the treatment of the third input. H₄: nanopipette at the final hidden state after the treatment of the final input. Outputs were evaluated by corresponding ICR ratios. Error bars in (b) represent the standard deviation of three independent experimental repeats (n = 3), and the measure of the center represents their corresponding mean value. Significance of the mean ICR ratio was assessed by t-test (see Supplementary Fig. 18). Source data are provided as a Source data file.

Our observations suggest that the designed chemically gated artificial nanochannels can respond to multiple chemical stimuli and display information about the operational procedure. Thus, our system can be used to operate programmable and complex input-output configurations.

In situ response of multiple chemical signals at the mitochondrial surface in a single living cell

Subsequently, we investigated whether the chemically gated artificial nanochannels could be used to perform subcellular targeting and chemical signal-induced modulation in single living cells. Based on the previous design, the surface of the nanopipettes was treated with APTES and GA, which could be used for the linkage of mitochondrial targeting peptide (MLS)⁵⁰. The sequence of MLS was shown in Supplementary Table 6. The MLS-modified nanochannels were then used for the capture of a selected mitochondrion in a single living cell. Here, 4T1 cells (mouse breast cancer cells) were selected for the following experiments. The Calcein-AM/PI double stain kit was used to test the viability of cells after inserting nanopipettes. From the microscopic cell images (Supplementary Fig. 19), the apoptosis fluorescence was negligible, demonstrating that the damage of the nanopipette to cells was not obvious after 1 h of penetration. Mito-Tracker Red was used for the staining of mitochondria in living cells, and individual mitochondria could be indicated by separately dispersed fluorescence spots inside the cell. To rule out the possible overlap between a selected mitochondrion and the nanopipette tip, the tip was pulled out of the cell if an individual mitochondrion was observed to be located at the orifice. As shown in Fig. 4a, a nanopipette was inserted into the cell membrane and moved closely to a single mitochondrion (a selected fluorescence spot) for 15 min. The slight movement of the nanopipette tip led to the synchronous movement of the mitochondrion, indicating successful capture of the mitochondrion by the nanopipette tip (Fig. 4a zoom). The continuous fluorescence observation confirmed that the mitochondrion was stably captured at the nanopipette tip after 2 h, enabling the lasting monitoring of the action order of mitochondrial chemical signals by the artificial nanochannels. A control experiment using a nanopipette without MLS modification showed that no mitochondrion was captured at the nanopipette tip (Supplementary Fig. 20), which ruled out the possibility of non-specific adsorption of mitochondrion at the nanopipette tip.

Fig. 4: In situ response of multiple mitochondrial signals (Ca²⁺/ROS/H⁺) during ROT-induced apoptosis.

a Microscopic images showing the targeting and capturing process of a selected mitochondrion in a single living 4T1 cell. BF bight-field. Mito-tracker: red channel. Scale bar: 20 µm. Scale bar in zoom: 500 nm. b The ICR ratio evolutionary path upon the treatment of ROT. c Schematic illustration of ROT-induced apoptosis. d Confocal microscopic images of 4T1 cells treated by ROT (10 µM) for different times and then stained with HyPer, GCaMP6s and JC-1. BF bright-field. HyPer: green channel (ex: 488 nm, em: 520 nm). GCaMP6s: green channel (ex: 488 nm, em: 520 nm). JC-1: red channel (ex: 550 nm, em: 570 nm). Scale bar: 20 µm. e–g FL intensity comparison histogram of 4T1 cells treated by ROT for different times in (d). Error bars in (b, e, f, g) represent the standard deviation of three independent experimental repeats (n = 3 independent samples), and the measure of the center represents their corresponding mean value. Control groups in (b, e, f, g) are corresponding data points measured at 0 min (samples without any treatment). Significance of the ICR ratio was assessed by t-test (n = 3 independent experiments and the data were presented as the mean values ± SDs, *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). All statistics were calculated using two-tailed paired t-tests. Source data are provided as a Source data file.

ROT-induced mitochondrial damage could trigger the release of chemical molecules (ROS/Ca²⁺/H⁺) from the mitochondrial matrix to the surface of OMM. The released chemical molecules could be detected in order through the chemically gated artificial nanochannels. Upon the treatment of ROT (10 μM) to stimulate the 4T1 cells, the ICR ratio of the mitochondrion captured nanopipette was gradually increased from 2.00 to 5.69 in 60 min (Fig. 4b, Stage 1), which proved the ROT-triggered ROS production by the mitochondrion. This phenomenon probably could be attributed to the ROT-activated ROS production within individual mitochondrion could trigger the long-term opening of mPTP, and further stimulate ROS generation in mitochondrion, which is known as ROS-induced ROS release (RIRR)⁵¹. Based on this, the ROS level released by mitochondria gradually increased with time. According to previous reports, ROS burst may induce severe and irreversible endoplasmic reticulum (ER) stress, thereby promoting cell death and amplifying ER stress responses^52,53, which can cause disturbance of Ca²⁺ homeostasis of mitochondrion, resulting in the overloading of mitochondrial Ca^2+54,55,56. At this stage, the fast efflux of ROS from mitochondria and the mitochondrial membrane potential loss resulted in a rapid drop in ROS and the leakage of H⁺, which could be revealed by the sharp decline of the ICR ratio from 5.69 to 1.74 in 30 min (Fig. 4b, Stage 2). Next, the ICR ratio increased again from 1.74 to 6.67 (Fig. 4b, Stage 3). This was caused by the exacerbated mitochondrial oxidative stress (EMOS) upon the continuous Ca²⁺ overloading⁵⁷. The whole process of ROT-induced apoptosis was illustrated in Fig. 4c. In addition, the simultaneous presence of multiple chemical molecules, such as two inputs (ROS/Ca²⁺, Ca²⁺/H⁺ and ROS/H⁺ Supplementary Figs. 21–23) or three inputs (ROS/Ca²⁺/H⁺, Supplementary Fig. 24) was used to build simple models to simulate the environment of mitochondrial matrix, and only beginning and ending states could be observed due to the simultaneous interaction between multiple inputs and DNA components. These results suggested that the MLS-modified nanopipette was located on the OMM rather than in the mitochondrial matrix, where detectable changes could be observed from the ICR ratio. In contrast, there was no significant change in ICR ratios of the control group without ROT treatment (Supplementary Fig. 25). For comparison, the burst of ROS in the cytoplasm during ROT stimulation was monitored using the nanopipette (without the modification of MLS at the outer surface). As shown in Supplementary Fig. 26, the ICR ratio was gradually increased from 1.36 to 2.67 in 120 min. And the response from the cytoplasm was reasonable because ROS burst from mitochondrion would quickly diffuse to the whole cell, leading to a limited amount of ROS detected by nanopipette. For the detection of ROS in the cytoplasm of cells without the treatment of ROT, no significant change in ICR ratios was observed (Supplementary Fig. 27).

Considering ROT could trigger a complex cascade, making interpretation of nanopipette readouts difficult, here we chose selective inducers, such as ionomycin for Ca²⁺ overloading and Bz-423 for ROS bursting, to assess the specific sensing of chemical molecules released by mitochondria. Ionomycin is a Ca²⁺ ionophore produced by Streptomyces conglobatus to increase intracellular Ca²⁺ levels. As shown in Supplementary Fig. 28a, the ICR ratio decreased from 2.78 to 1.41 in 20 min upon the treatment of ionomycin, indicating a rapidly increasing level of Ca²⁺⁵⁸. The mitochondrial Ca²⁺ overload could increase mitochondrial permeability, thus further aggravating Ca²⁺ overload and ROS busrt⁵⁹, as the ICR increased to 7.70 at 120 min. Bz-423 is a pro-apoptotic 1,4-benzodiazepine that targets the oligomycin-sensitivity conferring protein in mitochondria, thereby triggering the production of ROS^60,61,62. As shown in Supplementary Fig. 28b, the ICR ratio kept increasing in 20 min upon the addition of ionomycin, indicating a ROS burst. The following irreparable ER stress-triggered overloading of mitochondrial Ca²⁺ could be observed by the decrease in ICR from 5.39 to 3.99 at 30 min. Next, the EMOS upon the Ca²⁺ overloading led to an increase in the ICR ratio. The role of VDAC in permitting solute exchange was also considered. As shown in Supplementary Fig. 29, VBIT-4 was used to pharmacologically inhibit VDAC. The ICR evolutionary path of mitochondrion in 4T1 cells treated with VBIT-4 and ROT was similar to that in the cytoplasm treated with ROT, which indicated suppressed mitochondrial permeability upon VDAC inhibition, since no detectable changes were observed from the ICR ratio evolutionary path. Together, the above results demonstrated the successful establishment of a universal tool based on chemical-gated artificial nanochannels for the measurement of complex chemical signals at a localized subcellular region in spite of the complicated intracellular environment, providing insights into the action order of different chemical signals upon specific cellular events. The level of intracellular ROS, Ca²⁺, and ΔΨ in 4T1 cells after the treatment of ROT was examined by using LV-CMV-Mito-HyPer7-WPRE (HyPer), LV-0106LV-EF1a-GCaMp6s-WPRE (GCaMp6s), and JC-1, respectively. Each cell staining experiment was performed by using a single specific dye, and the excitation and emission wavelengths of HyPer, GCaMP6, and JC-1 were shown in Supplementary Table 7. Control experiments conducted on untreated bare 4T1 cells showed negligible spontaneous FL (Supplementary Fig. 30). As shown in Fig. 4d, in the Hyper channel, the green fluorescence intensity kept increasing in 60 min, indicating the activation of the RIRR (ROS-induced ROS release) process in mitochondria upon the treatment of ROT. Next, the leakage of ROS from mitochondria into cytoplasm resulted in a rapid drop at 90 min (Fig. 4e). It was worth noting that the green fluorescence of GCaMP6s reached the maximum at 90 min (Fig. 4f), which indicated that the ROS burst in the first 60 min triggered ER stress, causing disturbance of Ca²⁺ homeostasis in mitochondria and ER, thus inducing the subsequent transfer of Ca²⁺ from ER to mitochondria. Then the overload of Ca²⁺ in mitochondria could lead to excessive ROS production (Fig. 4e, 120 min), which in turn, EMOS and further promoted cell apoptosis. Besides, the red FL in the JC-1 channel dramatically dropped at 60 min (Fig. 4g), indicating the loss of mitochondrial membrane potential, which may explain the rapid drop of ICR ratio at 90 min due to the combination effect of the fast efflux of ROS from mitochondria and the leakage of H⁺ from the mitochondrial intermembrane space (Supplementary Fig. 31).

Collectively, these findings provide a framework of design principles for constructing the chemically gated artificial nanochannels, which enable programmable quick and accurate response to subcellular signals in living cells.

Evaluation and regulation of the interaction network between different organelles

Since we have achieved the targeted capture, regulation, and in situ response of a single mitochondrion in single living cells using the chemically gated artificial nanochannels, next we investigated whether the artificial nanochannels could be applied for the evaluation and regulation of the interaction networks between mitochondrion and other organelles.

Mitochondria (Mito) play a dual role, serving not only as essential organelles for cellular respiration but also as reservoirs for key metabolites, including lipids, protons, calcium, iron, and ATP. Thus, Mito are involved in multiple apoptosis and inflammation pathways through a complex interplay with different organelles^63,64. The connections between Mito and other organelles are highly dynamic, and Mito-centered networks with other organelles (ER, Lyso) are essential for maintaining the stability of the intracellular environment and are involved in the initiation and progression of numerous diseases. Nevertheless, technical limitations have limited the understanding of molecular components and substructural dynamics of organelle interaction networks, as the molecular temporal changes remain difficult to observe. Here, the proposed chemically gated artificial nanochannels achieve precise capture of individual mitochondria and monitoring of molecular temporal changes in local subcellular regions under various physiological and pathological states through the integration of multi-functional DNA modules. Rotenone (ROT) induces mitochondrial apoptosis through selective inhibition of mitochondrial respiratory chain complex I, thus promoting the production of ROS in mitochondria and inducing the apoptosis of cells⁶⁵. To avoid the possible influence of calcium ions on mitochondrial-ER/Lyso connections, we use DNAzymes specifically activated by other metal ions (Zn²⁺/Mg²⁺) to build connections between different organelles.

As shown in Fig. 5a, here we used metal ion-specific DNAzymes to perform dynamic regulation of interactions between Mito and ER/Lyso. Upon modification of Mito targeting peptides, DNAzyme strands (Zn²⁺-DNAzyme and Mg²⁺-DNAzyme) could be localized to mitochondria, and their corresponding substrate strands were localized to ER (Sub-1) or Lyso (Sub-2, Supplementary Table 8). The hybridization between DNAzymes and substrates enabled the specific linkage between Mito and ER/Lyso. In the presence of specific metal ions (Zn²⁺/Mg²), the DNAzymes could be activated to induce substrate cleavage to break the interaction network between Mito and ER/Lyso. Based on this, the dynamic regulation of assembly and disassembly between Mito and ER/Lyso was achieved in living cells. Specifically, Zn²⁺ and Mg²-specific DNAzymes were modified with mitochondrial targeting peptide (MLS) through click reaction between N₃-group and DBCO at 5′-end, and then tagged with AF 488 at 3′-end for locating Mito. The corresponding substrates, Sub-1 (substrate of Zn²⁺-DNAzyme), were modified with an endoplasmic reticulum targeting peptide (ELS) at the 5′-end, tagged with Atto 425 at the 3′-end for locating the ER. Sub-2 (substrate of Mg²⁺-DNAzyme) was modified with lysosome targeting peptide (LLS) at the 5′-end, and tagged with TAMRA at the 3′-end for locating Lyso. Here, Lipofectamine™ 2000 transfection reagent (Lipo-2000) was used to transfect nucleic acids into cells. Zn²⁺/Mg²⁺-DNAzyme strands were combined with Lipo-2000 solution to form liposomal vesicles incorporating DNA (Lipo-1/2) and then sent into the cells for 2 h incubation. After that, the obtained Sub-1/2-integrated liposome vesicles (Lipo-1′/2′) were incubated for another 2 h to acquire the Mito-ER/Lyso assembly cells (the detailed experimental information is shown in “Methods”). As shown in Supplementary Fig. 32, commercial tracker dyes, including Mito-Tracker (Mito-Tr, red), ER-Tracker (ER-Tr, red), and Lyso-Tracker (Lyso-Tr, green) were used to stain 4T1 cells treated with Zn²⁺-DNAzyme strand (green), sub-1 (blue), and sub-2 (red), respectively. The fluorescence of DNA strands and organelle trackers was highly overlapped. The Pearson’s correlation coefficient (PCC) between Mito-Tr and Zn²⁺-DNAzyme strand was calculated to be 0.88. The PCC between Er-Tr and sub-1 was 0.79, and the PCC between Lyso-Tr and sub-2 was 0.78. These results demonstrated the targeting ability of DNA strands to selected organelles. On the contrary, Mito-Tr, ER-Tr, and Lyso-Tr were used to stain 4T1 cells treated with Zn²⁺-DNAzyme strand (green, without the modification of MLS), sub-1 (blue, without the modification of ELS), and sub-2 (red, without the modification of LLS), respectively. The fluorescence of DNA strands and organelle trackers did not overlap. The PCC between Mito-Tr and Zn²⁺-DNAzyme strand was calculated to be 0.67. The PCC between Er-Tr and sub-1 was 0.37, and the PCC between Lyso-Tr and sub-2 was 0.17 (Supplementary Fig. 33). The above results further demonstrated the targeting ability of DNA strands to selected organelles and the successful assembly between organelles. As shown in Fig. 5b, i), upon the addition of Zn²⁺-DNAzyme and Sub-1, the assembly between Mito (green channel) and ER (blue channel) was observed from confocal microscopic images (0 min), which could also be seen from TEM images (Fig. 5b, ii). After that, the cells were treated with Zn²⁺ ions, and the FL intensity of Atto 425 was continuously decreasing within 30 min at the Mito area, while the FL intensity of AF 488 was almost unchanged. Similarly, after the addition of Mg²⁺-DNAzyme and Sub-2, the assembly between Mito (green channel) and Lyso (red channel) was observed from confocal microscopic images (Fig. 5c, 0 min), which could also be seen from TEM images (Fig. 5c, ii). As shown in Supplementary Fig. 34, the TEM images and FL co-ocalization analysis of control (without any tether), Mito-ER tether, and Mito-Lyso tether were repeated three times to confirm the tether between Mito-ER/Lyso. After that, the cells were treated with Mg²⁺ ions, and the FL intensity of TAMRA was continuously decreasing within 30 min at the Mito area, while the FL intensity of AF 488 was almost unchanged (Supplementary Fig. 35). Here, the changes of FL intensity in confocal images were quantified from the region of Mito, in order to observe the FL changes in situ in this region before and after adding metal ions. The decreased FL intensity upon Zn²⁺/Mg²⁺ activation indicated the intercellular disaggregation of Mito-ER /Lyso in this region. Throughout the entire cell region, the FL intensity did not change; thus, FL dyes were redistributed rather than quenched.

Fig. 5: In situ response of ROT-induced mitochondrial signals upon the interaction between mitochondria and other organelles.

a Schematic illustration showing the interaction network between different organelles based on metal ion-activated DNAzyme sequences. Zn²⁺-DNAzyme from 5′ to 3′: mitochondrion (Mito) targeting peptide, Zn²⁺ specific DNAzyme, and AF 488. Sub-1: the substrate of Zn²⁺ specific DNAzyme tagged with endoplasmic reticulum (ER) targeting peptide and Atto 425. Mg²⁺-DNAzyme from 5′ to 3′: Mito targeting peptide, Mg²⁺ specific DNAzyme, and AF 488. Sub-2: the substrate of Mg²⁺ specific DNAzyme tagged with lysosome (Lyso) targeting peptide and TAMRA. Lipo-1: liposomes containing Zn²⁺-DNAzyme. Lipo-1′: liposomes containing Sub-1. Lipo-2: liposomes containing Mg²⁺-DNAzyme. Lipo-2′: liposomes containing Sub-2. b Confocal microscopic images (i) and TEM images (ii) showing the assembly and disassembly between Mitochondria and ER in 4T1 cells. c Confocal microscopic images (i) and TEM images (ii) showing the assembly and disassembly between Mitochondria and Lysosomes in 4T1 cells. AF 488: green channel (ex = 488 nm). Atto 425: blue channel (ex = 405 nm). TAMRA: red channel (ex = 555 nm). Merge: the mixture of green and blue/red channels. Scale bar: 20 µm. d FL intensity comparison histogram of 4T1 cells treated by Zn²⁺ for different times in (b). e FL intensity comparison histogram of 4T1 cells treated by Mg²⁺ for different times in (c). f The ICR ratio evolutionary paths upon the treatment of ROT detected at the surface of a single mitochondrion upon Mito-ER assembly. g The ICR ratio evolutionary paths upon the treatment of ROT detected at the surface of a single mitochondrion upon Mito-Lyso assembly. Error bars in (d, e, f, g) represent the standard deviation of three independent experimental repeats (n = 3 independent samples), and the measure of the center represents their corresponding mean value. Control groups in (b, e, f, g) are corresponding data points measured at 0 min (samples without any treatment). Significance of the ICR ratio was assessed by t-test (n = 3 independent experiments and the data were presented as the mean values ± SDs, *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). All statistics were calculated using two-tailed paired t-tests. Source data are provided as a Source data file.

These results demonstrated that the linkage between Mito-ER and Mito-Lyso could be disconnected by adding corresponding metal ions (Zn²⁺/Mg²⁺). Control experiments without the treatment of Zn²⁺/Mg²⁺ were conducted (Supplementary Figs. 36 and 37), and the FL intensities in Mito-ER assembled cells and Mito-Lyso assembled cells were almost unchanged within 30 min. The corresponding histogram of FL intensities of the Mito area with or without metal ion treatment is shown in Fig. 5d, e. Control experiments of living cells without any treatment of DNAzymes/substrates were also performed to eliminate the interference of cell spontaneous FL (Supplementary Figs. 38 and 39). Together, the above results proved the successful linkage of Mito-ER and Mito-Lyso.

After achieving the linkage of Mito-ER and Mito-Lyso, the MLS-modified nanochannels were used for the capture of a mitochondrion in a single living cell and the subsequent dynamic monitoring of the chemical signals upon treatment of ROT. First, control experiments were performed to assess whether the tether of Mito-ER and Mito-Lyso affects key mitochondrial readouts. As shown in Supplementary Fig. 40, in the Hyper and GCaMP6 channels, the green FL intensity of Mito-ER and Mito-Lyso groups was insignificant. In the JC-1 channel, the red FL of Mito-ER and Mito-Lyso groups was similar to each other. The above results indicated that the tether between organelles did not affect the key mitochondrial readouts. Then, the interaction between Mito and ER was investigated. As shown in Fig. 5f, the ICR ratios kept increasing at first due to the RIRR effect in both situations (without/with the linkage of ER). Compared with the situation without the linkage of ER, the first turning point in the Mito-ER assembly cells could be seen earlier at 40 min because of the shortened distance between Mito and ER, which indicated that the artificial nanochannel could quickly detect the signal of Ca²⁺ with decreased ICR signal. Next, the ICR ratio increased again due to the EMOS effect at 50 min upon the continuous Ca²⁺-overloading. The interaction between Mito and Lyso was also tested (Fig. 5g). It was worth noticing that positive rectification was observed at the starting point (0 min) in Mito-Lyso assembly cells, which indicated decreased pH at the Mito area upon the approach of Lyso. As shown in Supplementary Fig. 41, the ICR ratio of Mito-ER (Mito-ER tethered, dark blue curve) was lower than Mito (Mito-ER untethered, light blue curve) on the whole, which could be attributed to the closer physical distance between Mito and ER to reduce the diffusion and dilution of Ca²⁺ into the cytoplasm. Thus, the efficiency of Ca²⁺ uptake by mitochondria was obviously improved. In addition, compared with Mito, the ICR response time of Mito-ER was shortened to 1 h. While in the situation of Mito-Lyso, positive rectification was observed at the starting point (0 min), indicating decreased pH at the Mito-Lyso contact region. According to a previous study, the addition of ROT can stimulate ROS generation in Mito and lead to the possible transfer of Ca²⁺ between Mito and Lyso, which could explain the slight decrease of ICR ratio from 0.99 to 0.78 in 10 min. Next, Mito was degraded through autophagy (mitophagy), a process in which damaged mitochondria (induced by ROT) are encapsulated by autophagosomes and subsequently fused with lysosomes to form autolysosomes, leading to the breakdown of their contents^66,67,68,69. Thus, the H⁺ from Lyso was gradually diluted, leading to an increase in pH value. Correspondingly, the ICR ratio gradually increased from 0.78 to 0.96 from 10 min to 40 min, and then kept increasing to reach 1.17 at 50 min, which could be attributed to the formation of an autophagosome. As shown in Supplementary Fig. 42, the confocal microscopic images of Mito-ER and Mito-Lyso, which used mismatching DNA structures, showed low PCC values, indicating the lack of organelle interaction without the strands’ hybridization. The above findings defined a comprehensive set of design guidelines for constructing the chemically gated artificial nanochannels, which enable programmable response to subcellular signals under different cellular processes, such as dynamic interaction between mitochondria and other organelles.

Together, the above results established a framework of design principles for constructing the chemically gated artificial nanochannels, which enable programmable response to subcellular signals under different cellular processes such as dynamic interaction between mitochondrion and other organelles (Table 1).

Table 1 Properties revealed from the interaction between different organelles

Discussion

In summary, we have developed biomimetic artificial nanochannels with chemically gated functions. The design of chemically gated artificial nanochannels is based on multi-functional glass nanopipettes to perform subcellular targeting and chemical signal-induced modulation in single living cells. The inner wall of the nanopipettes is functionalized by a delicately designed multi-functional DNA structure composed of two hybridized DNA sequences with chemical stimuli-responsive regulatory modules, and the nanopipettes are loaded with an electrolyte solution containing ABTS. Based on this, the surface charge of the inner wall can be changed upon the combination effect of multiple key chemical signals (Ca²⁺/ROS/H⁺), and this change can be revealed by monitoring the changes of ICR. Using functionalized nanopipettes as a proof of concept, we show that transmembrane channel-mimetic permeability transition functions have been successfully achieved. By modifying a mitochondrial targeting moiety at the outer surface of the nanopipettes, a single mitochondrion can be captured at the nanopipette tip to perform in situ modulation of the channel permeability in response to multiple mitochondrial chemical signals released from a single mitochondrion. Taken together, the results of this study present the design and performance of artificial nanochannels that can respond to complex chemical signals at a localized subcellular region in spite of the complicated intracellular environment. The proposed sensing strategy shows the advantages of simplicity, low cost, high stability, and little damage to cells. We here demonstrated that the chemically gated artificial nanochannels allow in situ response of multiple endogenous signals (Ca²⁺/ROS/H⁺) in single living cells, which can help to explore the action order of chemical signaling affected by specific cellular events. In the future, effort will be dedicated to potential applications in monitoring chemical signaling events in complex sample matrices such as tissues and living organisms, which may help to provide new insight into clinical practice. This study highlights the ability of chemically gated artificial nanochannels to implement lasting monitoring of the action order of local chemical signals in living cells without any external labeling. Finally, advanced applications of nanochannels for evaluating and regulating the interaction network between mitochondria and other organelles have been achieved. We believe that the results of this study will empower the future of the engineering of artificial nanochannels with customized gated functions, which are broadly applicable to implement sophisticated transport regulation. Our results will also open up a promising approach to build customized chemically gated artificial nanochannels for a wide range of applications. For instance, the inner surface of rectifying synthetic nanochannels could be modified to respond selectively to ATP, exhibiting specific sensitivity toward particular metal ions, thus enabling the ATP/metal ion coupled control of nanochannel properties. Another example is to construct a mitochondrial battery based on electron/proton transfer events during biological respiration, which may help to achieve efficient use of bioelectricity.

Methods

Materials

4T1 cells (Cat No. FH0366) were provided by Shanghai Fuheng Biotechnology Co., LTD (Shanghai, China). Rotenone (10 mM in DMSO), Lipo-2000, and Calcein-AM/PI were obtained from ThermoFisher Scientific Inc. (Waltham, MA, USA). DCFH-DA, Fluo-4 AM, and JC-1 probes were from Macklin (Shanghai, China). KCL, ZnCl₂, MgCl₂, CuSO₄, NaCl, 30% H₂O₂, 98% H₂SO₄ were obtained from Adamas-beta® (Shanghai, China). (3-aminopropyl)triethoxysilane (APTES, 98%), 50% Glutaric dialdehyde (GA) solution were from Energy Chemical (Shanghai, China). DNA markers were purchased from ZOMANBIO. All other chemical reagents were purchased from Sigma-Aldrich, Inc. (USA) of analytical grade. All solutions were prepared with ultrapure water generated by a Millipore Milli-Q water purification system (Billerica, MA, USA). DNA strands and the organelle-targeting sequences were synthesized and purified by Sangon Biotech (Shanghai), unless stated otherwise. All DNA samples were measured for concentration and stored in Milli-Q ultrapure water prior to use in subsequent experiments. The information on DNA sequences is given in Supplementary Tables 1, 4 and 8. And the detailed organelles locating sequence are given in Supplementary Table 6. MLS: mitochondrion targeting sequence. ELS: endoplasm reticulum targeting sequence. LLS: lysosome targeting sequence.

Apparatus

All fluorescence images were captured using a 60× water immersion objective on a laser scanning confocal microscope (Nikon A1R, Japan). Sample masses were determined using an analytical balance (ME 104, METTLER TOLEDO). Centrifugation of reagents was carried out with a centrifuge (Centrifuge 5430, Eppendorf). Scanning electron microscopy (SEM) imaging to examine nanopipette morphology was conducted using a field emission scanning electron microscope (Ultra55, Carl Zeiss Ltd., Germany). For electrochemical measurements, the nanopipette was mounted on the microscope stage using a holder (Axon Instruments, Union City, CA) and connected to an Axopatch 200B low-noise amplifier along with an Axon Digidata 1550A data acquisition system (Molecular Devices, Sunnyvale, CA).

Preparation of liposome-mediated transfection of DNA

First, organelles locating peptides (MLS, ELS, and LLS) modified DNA chains were dissolved in DMSO solution. Subsequently, 4 μL of 100 μM Zn²⁺-DNAzyme strands were combined with 50 μL of serum-free medium, while 2.0 μL of Lipo-2000 was dispersed in another 50 μL of serum-free medium and allowed to stand for 5 min at room temperature. The two mixtures were then blended and further incubated for 30 min under ambient conditions to form liposome-mediated transfection complexes of Zn²⁺-DNAzyme (designated as Lipo-1). Similarly, 4 μL of 100 μM Sub-1 strands were mixed with 50 μL of serum-free medium, and 2.0 μL of Lipo-2000 was separately diluted in 50 μL of serum-free medium, followed by a 5-min incubation at room temperature. After combining the two solutions and incubating them for an additional 30 min at room temperature, the resulting liposomal transfection complexes of Sub-1 (termed Lipo-1′) were obtained. Similarly, liposome-mediated transfection of Mg²⁺-DNAzyme chains (Lipo-2) and liposome-mediated transfection of Sub-2 (Lipo-2′) were also obtained.

Cell culture

4T1 cells were maintained in RPMI-1640 medium enriched with 10 % fetal bovine serum and 100 units/mL penicillin/streptomycin. The cell cultures were incubated in a humidified environment at 37 °C containing 5 % CO₂. For the fusion of DNA-loaded liposomes with cells, cancer cells were exposed to 100 μL of DNAzyme-encapsulated liposomal vesicles in 1640 medium for 2 h at 37 °C, followed by two washing steps. Then, the corresponding substrate-integrated liposome vesicles were added for another 2 h of incubation to achieve assembly of organelles.

Gel electrophoresis

To investigate the formation of synthetic transmembrane DNA receptors, agarose gel electrophoresis was performed in 1× TBE buffer at a constant voltage of 75 V for 45 min. Prior to electrophoresis, all samples were combined with 6× loading buffer (GenStar Biotech). For visualization, gels without intrinsic fluorescence were stained with Super GelRed (US Everbright Inc.). Images of the gels were captured using a Bio-Rad molecular imager under blue light excitation.

Preparation of nanopipettes

Borosilicate glass capillaries (BF100-58-10, Sutter Instrument Co., Novato, CA, USA), with an outer diameter of 1.0 mm and an inner diameter of 0.50 mm, were cleaned extensively using piranha solution for 30 min to eliminate surface-bound organic contaminants prior to pulling. Nanopipettes were fabricated using a Sutter P-2000 laser puller, with fabrication parameters detailed in Supplementary Table 3. To maintain consistent aperture geometry, the variation in pulling duration was kept within 0.1 s across preparations.

Functionalization of the inner and outer surface of nanopipettes

First, a 5% solution of aminopropyl triethoxysilane (APTES) in ethanol was introduced into the nanopipette tip to modify the inner surface, while the outer surface was functionalized by immersing the tip in the same APTES solution. Subsequently, the nanopipettes were rinsed thoroughly with ethanol and dried in a vacuum oven at 110 °C to form a uniform silanized layer. Next, both inner and outer surfaces were treated with a 2.5 % glutaraldehyde (GA) solution in ethanol for 4 h. Following a washing step, the DNAzyme solution was loaded into the nanopipettes, and the tip was submerged in a solution containing MLS (1 μM) for 12 h to facilitate conjugation.

ICR analysis

Two Ag/AgCl electrodes were employed to acquire electrochemical signals: one served as the working electrode, inserted into the nanochannel filled with electrolyte solution (0.02× PBS, pH 7.4), while the other acted as the reference electrode immersed in 1640 medium. Prior to measurements, the liquid-filled nanopipette tip was centrifuged to remove any entrapped air bubbles. The nanopipette was then mounted on a holder and connected to the headstage of an Axopatch 200B amplifier. Electrochemical recordings were conducted using a modified version of a previously established protocol, controlled by pClamp 10.7 software (Axon Instrument, Forest City, USA) running on a personal computer. Ionic current signals were recorded in gap-free mode at a sampling rate of 100 kHz, applying a 5 kHz low-pass Bessel filter. Data were analyzed using Clampfit 10.6, visualized in Origin 2021, and subsequently processed in Adobe Illustrator 2021 for final presentation.

Confocal laser scanning microscopy imaging

Confocal fluorescence imaging was carried out using a Nikon A1 laser scanning microscope (Japan) equipped with a 60× oil immersion objective. To minimize photobleaching, all procedures, including sample handling, staining, and image acquisition, were conducted under dark conditions. Fluorescence signals were collected within the following emission ranges: 400–450 nm for Atto 425 (excited at 405 nm), 500–550 nm for AF 488 (excited at 488 nm), and 570–620 nm for TAMRA (excited at 561 nm).