Forecasting the Reaction of DNA Modifying Enzymes on DNA Nanostructures by Coarse Grained Model for Stimuli-Responsive Drug Delivery

The reactivity of DNA modifying enzymes on their natural nucleic acid substrates has been fully understood. However, their reactivity on self-assembled nanostructures of nucleic acid is complicated and unpredictable. Here, we employed the molecular dynamic simulation to forecast the reactivity of tumor biomarker enzymes on DNA nanotubes by coarse grained model. It is found that the enzyme accessibility and the potential energy of the reaction products co-determine the structural change of DNA nanotubes. The reactivity can be regulated by the position of enzyme recognition site. According to the simulation results, stimuli-responsive drug nanocarrier with superior sensitivity and selectivity was developed. Drug payloads released in cancer cells is 3.7~5.5-fold higher than that in normal cells. The DNA nanocarrier equipped with cancer-specic aptamer AS1411 is used to deliver doxorubicin (DOX) to tumor-bearing mice not only effectively inhibiting tumor growth but also protecting major organs from drug-caused damage. This work provides new insights into the enzymatic reactivity of DNA nanostructures enriching the library of DNA-based reactions and heralding broad applications in nanomedicine.


Introduction
DNA nanotechnology offers controllable self-assembly on the nanoscale, allowing for the design of static structures, dynamic machines and computational architectures. Such nanotechnology provides unprecedented opportunities for nanomedicine applications including biosensing, cell modulation, bioimaging and drug delivery 1,2 . Owing to the excellent modularity and biocompatibility of DNA nanostructures, stimuli-responsive drug nanocarriers were established. A variety of drug molecules including chemotherapeutic agents 3 , small interfering RNA (siRNA) 4 , immunostimulatory sequences 5 , can be released in response to external and internal stimuli. The interaction of DNA nanostructure and enzymes particularly those DNA modifying enzymes or nucleases have drawn increasing attention because enzymes or nucleases are capable of enriching the applications in nanomedicine.
Nuclease activity is desired to be inhibited because nucleases present in the physiological environment can cause the undesired degradation of DNA nanostructures in real-life applications 3,6 . On the contrary, nuclease activity is desired to be maintained on DNA nanostructures for developing molecular probes 7 , responsive systems 8 , and complicated networks 9,10 . However, unlike the reaction between enzymes and natural substrates, the reactivity of enzymes on DNA nanostructure is complicated and unpredictable because the environment of enzyme recognition sites on nanostructures and the internal properties of nanostructures can be different from that of natural substrates.
Coarse-grained models provide a level of resolution between fully atomistic treatments and secondarystructure descriptions like the nearest-neighbor model 11 . Coarse-grained models represent individual nucleotides using a small number of interaction sites, which interact through effective potentials. If well parameterized, they can capture the known thermodynamic, structural and mechanical properties of DNA in a simple and naturally dynamical representation 12 . oxDNA a simulation code which implements the coarse-grained DNA model has been utilized to optimize DNA nanodevices including DNA tweezer 13 , origami 14 , and strand displacement reaction 15 . Therefore, this platform holds great potential to investigate the interaction of DNA modifying enzymes and DNA nanostructures.
Here, we forecast the reaction of DNA nanotubes and the enzyme, glycosylase (uracil degradation glycosylase) and endonuclease (Human apurinic/apyrimidinic (AP) endonuclease, APE1) which involves in cellular base excision repair (BER) pathway by coarse-grained model. These enzymes are signi cantly active in the cytoplasm of cancer cell line rather than that of the normal cell line, have emerged as promising biomarkers for cancer diagnostics and treatment. The simulation discloses that the orientation of enzyme recognition site and the potential energy of the product of enzymatic reaction exhibit synergistic effects on the responsiveness of DNA nanotubes. The nanotube was used to carry the anticancer drug doxorubicin (DOX). The simulation-guided location choice of enzyme recognition site permits highly sensitive and speci c drug release in cancer cells. DNA nanotubes equipped with tumorspeci c aptamers effectively inhibited tumors growth in tumor-bearing mice, and effectively protected major organs from damage. We anticipate that this work provides new insights into the design of functional DNA nanostructures and breaks through the bottleneck of stimuli-responsive nanocarriers in sensitivity and speci city.

Result And Discussion
Simulation of uracil-contained DNA nanotubes by coarse grained model BER pathway is signi cantly active in cancer cells 17 . The critical enzymes UDG and APE1 in this pathway are overexpressed in the cytoplasm of various cancer cells 18-20 . These two enzymes process uracilcontaining DNA. UDG as a glycosylase catalyzes the hydrolysis of the N-glycosidic bond from deoxyuridine to release uracil thereby generating AP site 21 . APE1 as an endonuclease cleaves DNA phosphodiester backbone at AP sites via hydrolysis leaving a one nucleotide gap with 3'-hydroxyl and 5' deoxyribose phosphate termini (Scheme 1) 22 . We implemented uracil-containing DNA nanotubes in response to these enzymes. Figure 1A shows the principle structure of DNA nanotube. The DNA tile consisting of ve single-stranded DNA is the building block of DNA nanotube. Sticky end hybridization allows for the nanotube formation. To make the DNA nanotubes responsive to UDG and APE1, thymine (T) needs to be substituted by uracil (U). Enzyme accessibility and the stability of enzymatic reaction product determine the degradation degree of DNA nanotubes in the presence of enzymes. Accordingly, the T to U substitution position is crucial for the reaction, which can be evaluated by coarse grained simulation. With the advance of in silico design tools such as caDNAno, Tiamat, CanDo, and DAEDALUS, the design of such DNA nanostructures is now a relatively fast and well-developed process 23 . To start simulation, we use caDNAno to design DNA nanotubes as the input structure for coarse grained model simulation on oxDNA platform 24 . According to previous study, such DNA nanotube has the diameter with an average of 13.5 nm, the circumference contains 14 DNA tiles (Figure 1, Figure S1) 24,25 . DNA nanostructures is usually stabilized by an appropriate Mg concentration (e.g. 12.5 mM). Under this ionic strength (12.5 mM Mg equals to 0.09397 M monovalent cation), 87% hydrogen bond strength was used as oxDNA simulation input because of the small root mean squared uctuations (RMSF) and its narrow distribution ( Figure S2). RMSF can observe how various traps modify the exibility of individual nucleotides, it can use single value decomposition alignment to calculate the mean position of every nucleotide, then again analyzes the trajectory to get RMSF of each particle from its mean position 26 . The RMSF can reveal the dynamics of each area of the nanotube 16,27 . A small and narrowly distributed RMSF re ects a stable structure.
First, we simulated the conformation of DNA nanotube to study the enzyme accessibility which depends on base orientation. We focused on fourteen T to U substitution sites on S1 and S5 strands which are hybridized with the bases of S3 strand (highlighted in Figure 1A). Under the simulation condition, the detailed environment of the fourteen positions are shown in Figure 1B and Figure S3. Brie y, S1-P3, S1-P1, and S5-P6 bases are outward, the others are inward. These results imply the U to T substitution at S1-P3, S1-P1, and S5-P6 may exhibit high enzyme accessibility.
Second, we investigated the stability of enzymatic reaction products by simulation. According to the enzyme properties, one-base gap can be generated after enzymatic cleavage. Figure 1C shows the enzymatic reaction product if the uracil is located at S1-P1 position. According to the simulation, fourteen reaction products of nanotubes whose tiles have one-base gap exhibit different potential energy ( Figure  1D). The control structure (no enzyme-caused break) with highest stability yields the lowest energy. S1-P5 shows the highest potential energy among all the candidates indicating that it has lowest stability. We analyzed the average structure and the distribution of RMSF of each nucleotide for the nanotubes. As shown in Figure 1E, the control structure (no enzyme-caused break) exhibits a small RMSF and a narrow distribution. At a glance of the positions on S1 strand, RMSF of S1-P5 is signi cantly higher than that of the other structures implying that S1-P5 can yield a high degradation degree. S5-P1 in S5 strand shows the similar feature as S1-P5 ( Figure S4). Hydrogen bonding is an important parameter that de nes the geometric structure of DNA nanotechnology 16 . Since the stable nal structure is designed as the theoretical minimum global free energy, the hydrogen bond is maximized. The change in the hydrogen bond occupancy rate shows that the current structure deviates from the designed structure and points to other important topological strain regions. Therefore, we calculated the occupancy of hydrogen bonds with a strength of less than 10% hydrogen bond strength to the total hydrogen bonds. The occupancy rate of hydrogen bonds indicates the degree of dissociation of the entire DNA nanotubes structure ( Figure  1F). Consistent with the results of potential energy and RMSF, S1-P5 yields the lowest occupancy of hydrogen bonds indicating a high degree of degradation.
By comprehensively analyzing the simulation results, we categorized the fourteen candidates into two classes: 'high quality', and 'low quality', which re ect the potential degradation degree of the nanotubes after enzymatic reaction. 'High quality' class has the following characteristics: outward base orientation or low stability of the cleaved tile. In contrast, the candidates in 'low quality' class have inward base orientation and high stability of the cleaved tile. Accordingly, for S1 strand, S1-P5, S1-P4, and S1-P3 belongs to 'high quality' class, S1-P1 and S1-P2 belongs to 'low quality' class.

Characterization and Responsiveness of DNA nanotubes
All candidates on S1 strand were tested to validate the simulation results. The nanotubes were assembled by the reported protocol 25,28 . Gel electrophoresis reveals that the assembled nanotubes have high purity, and T to U substitution has no in uence on the nanotube assembly ( Figure S5). Further characterization was performed by using total internal re ection uorescence (TIRF) microscopy which provides high S/N for near eld excitation 29 . As shown in Figure S6A, short strip was found, and the length of nanotube was estimated as ~400 nm which is consistent with the measurement by dynamic light scattering (DLS) ( Figure S6B). The reaction of uracil-containing nanotube and enzymes were rst characterized by gel electrophoresis (Figure 2A). Only when both enzymes are present can the nanotubes be degraded. The degradation degree of S1-P3, S1-P4, S1-P5 is signi cantly higher than the that of S1-P1 and S1-P2. This result is perfectly matched with the simulation results. S1-P3 and S1-P5 were chosen as the nanocarriers for drug delivery. Detailed characterization was carried out for S1-P3 and S1-P5. Fluorophore and quencher were labeled on their S3 and S4 strands for the measurement of reaction kinetics ( Figure S7). As shown in Figure 2B, the uorescence of S1-P3 and S1-P5 rapidly increases in the presence of UDG and APE1. In contrast, the control nanotube (no uracil) shows no uorescence enhancement. These two nanotubes can respond the low concentration of enzyme down to 200 pM of APE1 and 1 nM of UDG. The nanotubes show the same level of enzymatic reactivity as uracil-containing double-stranded substrate, which subverts traditional cognition that the reactivity of DNA nanostructures is weaker than that of natural substrates ( Figure S8). By the assistance of simulation, the reaction sensitivity of nucleic acid nanostructure can be modulated as that of natural substrate. Speci city is another important parameter for these nanotubes. Owing to the nature of nucleic acids, DNA nanostructures are vulnerable to non-speci c nucleases 30 . In contrast, our results show that the nanotubes are resistant to a variety of nucleases ( Figure S9). Tubular structure may make these nanotubes resistant to endonucleases, and the terminal structure of DNA tiles reduces the accessibility of exonucleases. The resistance of nanotubes to nonspeci c nuclease is superior than double-stranded DNA 31 . This result guarantees the speci city of the nanotube. Furthermore, TIRF imaging was performed to directly visualize the UDG/APE1-caused degradation of nanotubes ( Figure 2C). After enzymatic reaction, the uracil-containing nanotubes disappear, and the control nanotube keeps intact. Collectively, the simulation-guided nanotubes exhibit high sensitivity and speci city towards UDG and APE1, which makes them as excellent stimuli-responsive nanocarriers.
Next, we explored enzyme-stimulated drug release of the nanotubes. DOX is commonly used to as model drug to test the drug delivery system owing to its excellent antitumor effect and uorescence property 32 .
DOX is suitable for DNA-based nanocarriers because it intercalates into the minor groove of doublestranded DNA 33 . The low pH in tumor microenvironment (TME) is shared by healthy joints 34 , intracellular milieu 35 , and the stomach. pH-responsive drug release lacks tumor speci city. As shown in Figure 2D, the nanotube is not sensitive to pH, the disassembly of DNA nanotubes is slow at pH 7.4 (physiological) and 5.6 (endo-lysosome), and the drug release is less than 10% within 24h. In contrast, in the presence of BER enzymes, the DNA nanotubes are rapidly disintegrated, and more than 50% of the drug is released within 4h, the release can reach 77% within 24h. The loading e ciency of DNA nanotubes of 78%, which can be comparable with currently used nanomaterials 36,37 . These results imply that the nanotubes are robust and not sensitive to pH and can effectively release drugs under the environment containing highly expressed BER enzymes.

Cellular uptake of DNA nanotubes and drug delivery
The drug nanocarriers S1-P3 and S1-P5 were used to demonstrate the stimuli-responsive release in different cell lines. Highly inclined and laminated optical sheet (HILO) uorescence microscopy which provides high S/N for cellular uorescence excitation was used to cellular the cellular behaviors of the DNA nanotubes. The size and shape of DNA nanostructures has in uence on their cellular uptake e ciency 38 . We rst explored the cellular internalization of the nanotubes with different length which were assembled with different annealing speed, 18h for long nanotube and 1.5h for short nanotube ( Figure S10). As a result, the short nanotube exhibits signi cantly higher e ciency of internalization than the long one ( Figure 3A). The uorescence of long nanotube mainly accumulates on the cell membrane.
The long one-dimensional size results in the adsorption of lipid and DNA nanotube and the poor cellular uptake e ciency. Thus, short uracil-containing nanotube was used for drug delivery in cell.
The overexpression of BER enzymes UDG and APE1 have been veri ed in the cytoplasm of a variety of cancer cell lines 18 . We used uorophore and quencher labeled uracil-containing nanotubes to demonstrate the differential expression level of these enzymes in cancer and normal cell lines ( Figure   3B). The nanotubes exhibit signi cantly higher Cy5 uorescence in cancer lines HeLa, MCF7, and A549 than in normal cell line HEK-293 ( Figure 3B, S11, and S12). For S1-P5 nanotube, the mean uorescence intensity (MFI) is 7, 19, and, 4-fold higher in HeLa, MCF-7, and A549 cell lines than in HEK-293 cell line ( Figure 3B, S11, and S12). The control nanotube (no uracil) shows weak uorescence in all cell lines indicating the reaction speci city. APE1 inhibitor 7-nitroindole-2-carboxylicacid (NCA) was used to further con rm the speci city 39 . As expected, negligible uorescence was found in HeLa cell for S1-P5 nanotube in the presence of NCA ( Figure S13). All these results suggest that the uracil-containing nanotubes are capable of reacting with cellular UDG and APE1 accordingly.
DOX could ght against various cancers by embedding in genomic DNA, covalently binding to proteins involved in DNA replication and transcription, or inhibiting topoisomerase II, inducing apoptosis of tumor cells and the main target of DOX is nucleus DNA 40 . According to previous study, DNA nanostructures cannot enter nucleus 7 . Moreover, our results also show that the nanotubes mainly distribute in cytoplasm rather than nucleus ( Figure 3B, S11, and S12). Therefore, only when the nanotube drug carrier disintegrates in cytoplasm can the drug enter the nucleus. As expected, strong DOX uorescence can be observed in the nucleus of all the cancer cell lines when uracil-containing nanotubes are used as DOX carriers ( Figure 3C, S14, and S15). According to the MIF in nucleus, nanotube drug carriers release DOX 3.7~5.5-fold higher in cancer cell lines than in normal cell line ( Figure 3C, S14, and S15). In contrast, free DOX without using any carriers can undifferentiated enter the nucleus of cancer and normal cells ( Figure  3C, S14, and S15). We further evaluated the cell toxicity of this system. As shown in Figure S16, all the drug-free nanotubes (uracil and non-uracil) do not show any observable toxicity on cancer and normal cell lines. Uracil-containing nanotubes exhibit high toxicity on cancer cell but not on normal cell when DOX was loaded. The DOX loaded control nanotube (no uracil) have relatively weak toxicity to all cells.
Free DOX exhibit high toxicity to all types of cell. In this way, the drug effect can be precisely administrated by the BER enzyme-responsive DNA nanotubes allowing for selectively inhibiting cancer cells and protecting normal cells.
In vivo therapeutic e cacy Inspired by the excellent performance at the cellular level, we further evaluated the in vivo therapeutic e cacy. All animal experiments were approved by the ethics committee of our institute and were performed in accordance with the guidelines of the institution animal care and use. DNA nanotubes and DOX were applied to the tumor-bearing mice (HeLa cell) which were continuously treated for 12 days by tail vein injection of saline, free DOX, control-DOX, S1-P5-DOX (DOX equivalent 2 mg/kg), the administration was shown in Figure 4A. DNA nanotubes were equipped with AS1411 aptamer to enhance tumor targeting ( Figure S17) 41 . First, the in vivo enzymatic reaction was investigated by using the uorophore and quencher labeled nanotubes (no DOX). From the uorescence imaging, we found that the uracil-containing nanotube (S1-P5) shows stronger uorescence signal at the tumor site than the nonuracil nanotube within 4h ( Figure 4B). This implies that the aptamer equipped nanotubes can accumulate at the tumor, and BER enzymes are su ciently active to stimulate DNA nanotubes in the implanted tumor cells.
DOX-loaded nanotubes were used to show the anticancer effect. During 12 days of the therapeutic period, there is no obvious change of the body weights for all groups implying no signi cant toxicity for the DNA nanotubes ( Figure S18). The tumor volumes of all groups were recorded in different days. As shown in Figure 4C and 4D, the tumor volumes in the groups of PBS, free DOX, and control-DOX gradually increased by 15.5, 6.0, and 11.7 fold, respectively, while S1-P5-DOX could obviously inhibit the growth of tumor. Furthermore, the survival rate of HeLa-tumor-bearing mice was evaluated. Mice treated with S1-P5-DOX achieved a 100% survival rate after 50 days ( Figure 4E). Hematoxylineosin (H&E) and Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining revealed that the therapeutic delivered by S1-P5-DOX induced a higher level of nucleus necrosis and apoptosis of tumor tissue than the other groups ( Figure 4F). H&E staining analysis of major organs reveals the high biocompatibility of all the groups except for the free DOX group (Figure 4G and S19). Owing to the side effect of DOX, focal necrotic cell death (blue arrow) heart and liver can be observed ( Figure 4G) 42 . The above results indicate that the BER stimuli-responsive DNA nanotubes effectively reduce the side effects of chemotherapy drugs on the major organs, and massively accumulated at the tumor site then speci cally release the drug on demand to achieve the promising therapeutic effect.

Discussion
Targeted drug delivery remains effective approach for cancer therapeutics 43 . Stimuli-responsive drug carriers that undergo degradation or conformational change in response to external or endogenous stimuli have been engineered to control drug release in the tissue of interest effectively enhancing targetability and mitigating side effects 2,44,45 . The abnormal pH in microenvironment of tumor, cancer molecular biomarkers (e.g. nucleic acids, proteins), and even external energy sources (e.g. light, sound) have been employed as stimuli to design and engineer drug carriers 46-48 .
The sensitivity and speci city of drug carriers are the limiting factors for the e cacy of stimuli responsive drug delivery. Base excision repair (BER) pathway is a cellular mechanism that repairs damaged DNA induced by eliminating damaged (oxidized or alkylated) or inappropriate bases that are generated endogenously or induced by genotoxicants throughout the cell cycle 49 . BER recruits multiple enzymes including DNA glycosylase, endonuclease, polymerase, and ligase 22 . BER has been proven to be more active in tumor cells and the enzymes were signi cantly overexpressed than normal cells 17 . It is advantageous in the speci city and sensitivity to exploit the enzymes in BER pathway as the stimuli to develop responsive drug delivery system. Despite the rapid progress made in the eld of DNA based nanocarriers, it remains challenging to accurately predict the enzyme responsiveness at the design stage by using current design procedures because the interaction of complicated DNA nanostructure-enzyme is distinct from that of natural nucleic acid substrate-enzyme.
In this work, molecular dynamic (MD) simulation based on coarse grained model is employed to predict the reaction of BER enzymes which are DNA modifying enzymes and DNA nanotubes. With the guidance of MD simulation, uracil-containing DNA nanotubes can response to the BER enzymes in cancer cells with high sensitivity and speci city. The nanotubes hold potential to program the drug release kinetics by altering T to U substitution position. Owing to superior release selectivity, DNA nanotubes can release drugs in a variety of tumor cells (HeLa cells, MCF7 cells, and A549 cells) but not in normal cells (HEK 293 cells). DNA nanotubes equipped with AS1411 aptamers effectively inhibit tumor growth in HeLa-bearing mice, and protect major organs from damage. The simulation procedure provides new insights into the interaction of DNA nanostructures and DNA modifying enzymes which enriches the library of DNA-based reactions. This work is anticipated to nd broad applications in DNA chemistry and biomaterials.

Declarations
Author Contribution X.S. conceived the ideas and designed the study. Y.D., C.Z., and X.S. designed, performed and analysed the experiments. Y.T., L.Z. completed the molecular simulation. Y.D. wrote the manuscript, with critical revision by X.S. X.S. provided technical support and advice. All the authors read and approved the nal version of the manuscript.

Competing interests
There is no con ict to declare.

Data availability
The data that supports the plots within this paper and other nding of this study are available from the corresponding author upon reasonable request.

Supplementary information
Supplementary Figure S1-  52. oxDNA simulation. Structural analysis of the different types of the DNA nanotubes were achieved by using oxDNA which is a coarse-grained molecular dynamics (MD) simulation software. We used caDNAno to draw the DNA nanotubes, and export the le formats to "xx.top" and "xx.oxdna" through TacoxDNA. The position and orientation of each base and the overall DNA structure are intuitively presented in a visual image. The image is visualized by the oxDNA generated le in the oxView (https://sulcgroup.github.io/oxdna-viewer/), then we used "RELAX" option to oxDNA interaction, we can modify parameters in the input le, such as the number of steps, temperature and other conditions, and adjust hydrogen bond strength in the le of "oxDNA2_sequence_dependent_parameters" for simulation. The detailed simulation parameters and conditions are noted in Table S2. We used Python les to run the obtained trajectory les and extracted and analyzed the data to obtain the results of RMSF and hydrogen bond occupancy. Then we plotted the data and compared the data of different DNA nanotubes.
53. Preparation and characterization of DNA nanotubes. All sequences were dissolved and diluted in DNase/RNase-free deionized water. To synthesize tubular DNA, 2.5 µL of 20 µM of DNA sequences (S1-P1, S1-P2, S1-P3, S1-P4, and S1-P5, for sequences see Table S1) were mixed in TM buffer (40mM Tris,1 mM EDTA-Na 2 , 12.5 mM Mg 2+ , pH=8). The DNA solutions were heated to 90°C for 5 min and slowly cooled down to room temperature (1.5h or 18h). The products were characterized by 8% native polyacrylamide gel electrophoresis which was operated at 4°C for 1 h at a constant voltage of 120 V. The gel was subsequently stained with SYBR Gold dye. Cy5-labelled DNA nanotubes were imaged using TIRF microscope with a TIRF objective (100× magni cation, 1.49 NA, Nikon). The lenses, re ection mirrors, and dichroic mirrors were from Semrock (USA). For TIRF illumination, a solid laser of 520 nm was coupled into single-mode ber cable (Solamere Technologies). Samples containing nanotubes were imaged at 100 nM tile concentration in TM buffer.
54. Reaction kinetics of the nanotubes to nucleases. 1 µM Cy5 labeled strand (S3-Cy5), 1 µM BHQ3 labeled strand (S4-BHQ3) and 1 µM other strands (S1-P1, S1-P2, and S1-P5) were mixed and annealed in TM buffer to prepare dual-labeled nanotubes. In a typical uorescence dequenching assay, 250 nM nanotubes, and enzymes were incubated in TM buffer. Once the enzymes were added, The HEK293T cells were cultured in a DMEM medium with 10% fetal bovine serum and 1% penicillin/streptomycin under standard conditions (5% CO 2 , 37°C). The medium was replaced every 24 h, and the cells were digested with trypsin and resuspended in fresh complete medium before plating.
0. In vitro cellular uptake. The cellular uptake behavior of nanotubes and DOX delivery in cultured cell lines were investigated by using the HILO uorescence microscopy. According to the cell thickness and the S/N of single-molecule uorescence, the micrometer-driven optical rail for Z adjustment was adjusted to achieve HILO illumination. The cells were seeded into confocal dishes with a density of 4×10 5 cells/well and incubated with RPMI 1640 Medium in a humidi ed atmosphere containing 5% obtained trajectory les and extracted and analyzed the data to obtain the results of RMSF and hydrogen bond occupancy. Then we plotted the data and compared the data of different DNA nanotubes.
Preparation and characterization of DNA nanotubes. All sequences were dissolved and diluted in DNase/RNase-free deionized water. To synthesize tubular DNA, 2.5 μL of 20 μM of DNA sequences (S1-P1, S1-P2, S1-P3, S1-P4, and S1-P5, for sequences see Table S1) were mixed in TM buffer (40mM Tris,1 mM EDTA-Na 2 , 12.5 mM Mg 2+ , pH=8). The DNA solutions were heated to 90 °C for 5 min and slowly cooled down to room temperature (1.5h or 18h). The products were characterized by 8% native polyacrylamide gel electrophoresis which was operated at 4 °C for 1 h at a constant voltage of 120 V. The gel was subsequently stained with SYBR Gold dye. Cy5-labelled DNA nanotubes were imaged using TIRF microscope with a TIRF objective (100× magni cation, 1.49 NA, Nikon). The lenses, re ection mirrors, and dichroic mirrors were from Semrock (USA). For TIRF illumination, a solid laser of 520 nm was coupled into single-mode ber cable (Solamere Technologies). Samples containing nanotubes were imaged at 100 nM tile concentration in TM buffer.
Cell Culture. The HeLa cells, A549 cells and MCF 7 cells were cultured in a RPMI 1640 medium with 10% fetal bovine serum and 1% penicillin/streptomycin under standard conditions (5% CO 2 , 37 °C). The HEK293T cells were cultured in a DMEM medium with 10% fetal bovine serum and 1% penicillin/streptomycin under standard conditions (5% CO 2 , 37 °C). The medium was replaced every 24 h, and the cells were digested with trypsin and resuspended in fresh complete medium before plating.
In vitro cellular uptake. The cellular uptake behavior of nanotubes and DOX delivery in cultured cell lines were investigated by using the HILO uorescence microscopy. According to the cell thickness and the S/N of single-molecule uorescence, the micrometer-driven optical rail for Z adjustment was adjusted to achieve HILO illumination. The cells were seeded into confocal dishes with a density of 4×10 5