Introduction

Signal transduction is an important feature of living cells, which refers to the biological process of transferring chemical or physical signals throughout a living organism1,2,3. At the cellular level, signals are transmitted across or through a cell from outside to inside, which allows cells to respond to external environmental changes and communicate with other cells4,5,6,7,8,9. In natural systems, signal transduction mainly relies on transmembrane proteins known as receptors, which can be activated by external messengers10,11,12. Upon the interaction between receptors and ligands, a conformational change is triggered in transmembrane receptors, followed by the activation of a series of downstream cascade reaction pathways13,14,15,16,17. Inspired by natural receptors enable cells to detect, process, and respond to their environment cues, there is substantial interest in developing engineered synthetic receptors that can implement sophisticated customized sense-and-respond functions, as the structure of synthetic receptors can be tailored or modified more easily to adapt to different situations and purposes18,19,20. Ideally, synthetic receptors can be chemically encoded to respond to different input signals in a manner that triggers a change in the receptor state, and the change of the receptors can induce cascade reactions to actuate downstream changes in cell state. In addition, it is important that synthetic receptors can be easily embedded into living cell membranes without affecting cell viability.

In recent years, much progress has been made in constructing various synthetic transmembrane receptors to perform recognition-based transmembrane sensing and actuation21,22,23,24. Inspired by natural counterparts, researchers have focused on creating various functional molecules acting as receptors, including organic polymers, host-guest supramolecules, molecular motors/switches, protein-aptamer pairs, and DNA-based scaffolds25,26,27,28,29,30,31,32,33,34. Among them, DNA scaffolds are especially attractive. The improvement of high-throughput SELEX technology accelerates the screening of high-affinity DNA aptamers, which boosts the design of DNA nanostructure-based receptors with enhanced functional accuracy, such as DNA nanocages, DNA nanotubes, DNA polyhedrons, DNA origami, etc.35,36,37,38,39,40,41. These innovations enable DNA scaffolds to exhibit higher sensitivity and specificity in sensors and drug delivery systems42,43,44. In addition, the combination of DNA scaffolds with nanomedicine and synthetic biology has expanded their application prospects in real-time imaging, targeted therapy, and biosensing45,46. Therefore, synthetic receptors can be used for mimicking cellular signal transduction behaviors, such as ligand-induced receptor dimerization and activation of cascade reactions8,47. These synthetic receptors successfully mimic the cellular signal transduction behaviors, like ligand-induced receptor dimerization and activation of cascade reactions. Although most of these receptors have been proven to work well in synthetic liposomes to mimic living cells, it remains necessary to develop smarter, more biocompatible, and more customizable sense-and-respond receptors to enable user-defined responses and sophisticated customized functions that can be easily specified and precisely controlled in natural living cells, which imposes important challenges including the fast and stable anchoring of the synthetic receptors on living cell membrane, the engineering of the appropriate conformational change upon the recognition between the receptors and input information from changes in the cell’s environment, and specifically, the coupling between the downstream cascade events and emerging cell function regulations dictated by the cascade.

In this work, we report herein a flexible cell surface engineering strategy based on synthetic transmembrane DNA receptors to perform a dynamic cascade signal transduction process across the membrane of living cells. The synthetic DNA receptors are composed of three functional parts, including an extracellular signal reception part, a lipophilic transmembrane anchoring part, and an intracellular signal output part. Upon the input of external signals, the DNA receptors can form dimers on the cell surface, which induces downstream events via specific interaction between the dimers and cellular functional molecules. First, we validate the performance of our design by presenting the organization of a couple of membrane-spanning DNA receptors, where the hybridization of the two strands activates their dimerization, resulting in the generation of DNAzyme structures inside the cell. Upon the addition of corresponding metal ions, the activated DNAzyme structures can catalyze the cleavage of the substrates. Next, we develop the aptamer-locked design rules to tune the extracellular signal reception part of the synthetic DNA receptors, as different logic functions are implemented (AND, OR)48. Furthermore, we demonstrate the use of DNAzyme-catalytic substrate cleavage as a functional unit to regulate cell functions, including intercellular communication, gene transcription, protein expression, and cell apoptosis. Notably, based on the AND-controlled approach, we realize chemical messenger communication between donor T-cells and recipient cancer cells, indicating potential therapeutic applications. Taken together, this work demonstrates that synthetic transmembrane DNA receptors can be used to implement sophisticated and delicate cell regulation via precisely controlled inter- and intracellular signal transduction.

Results

Development of synthetic transmembrane DNA receptors for signal transduction

To evaluate the feasibility of the synthetic transmembrane DNA receptors for signal transduction, we have designed two transmembrane DNA strands (Seq1 and Seq2) to establish a signal output system. As shown in Fig. 1a, both Seq1 and Seq2 consisted of three functional parts: an extracellular part (L1 and L2) for signal recognition, a hydrophobic C12 spacer for membrane anchoring (C1 and C2), and an intracellular part (Dz1 and Dz2) for downstream signal output part (Supplementary Table 1). Here the L1 and L2 were designed to be complementary to facilitate the formation of Seq1/Seq2 dimers, while the Dz1 and Dz2 together comprised the complete sequence of a Mn2+-dependent DNAzyme. Before cell experiments, DNA strands were mixed with lipid solution to obtain DNA-integrated liposome vesicles (lipo-Seq1 and lipo-Seq2, Supplementary Fig. 1). The prepared lipo-Seq1 and lipo-Seq2 were characterized by using DLS and zeta potential measurements (Supplementary Fig. 2). The integrated DNA in the liposome can be directly transported onto the cell membrane via a vesicle-cell fusion process (Fig. 1b)34. The DNA-integrated liposomes showed a significantly increased fusion efficiency than that of bare DNA strands (Supplementary Fig. 7). The cells modified with Seq1/Seq2 retained a stable fluorescence distribution on the cell membrane after 12 h of incubation (Supplementary Fig. 8), indicating good stability of the synthetic DNA receptors under physiological conditions. As a result, Seq1 and Seq2 strands on the cell surface served as signal receptors, which could receive external stimuli to induce downstream signal output.

Fig. 1: Working principle of the synthetic transmembrane DNA receptors for signal transduction.
Fig. 1: Working principle of the synthetic transmembrane DNA receptors for signal transduction.The alternative text for this image may have been generated using AI.
Full size image

a The design of the DNA-Mediated transmembrane structure (Seq1, Seq2), which contains an extracellular part for signal recognition (L1, L2), a transmembrane part for anchoring (C1, C2), and an intracellular part for downstream regulation (Dz1, Dz2). Strand L1 is complementary to strand L2. Strand Dz1 and strand Dz2 are components of a Mn2+-dependent DNAzyme. Seq1 and Seq2 were integrated into the membrane of liposomes, respectively. Experiments were repeated three times independently with similar results. b The vesicle-cell fusion process between DNA-integrated liposomes and cell membranes. c Details of the DNA-mediated transmembrane signal transduction via cell surface dimerization. Upon the activation of the DNAzyme (Dz1, Dz2) by Mn2+, the fluorescence (Cy5) of the hairpin substrates (H-Mn) can be turned ON. d CLSM images of H-Mn loaded HeLa cells (top), H-Mn loaded and S1, S2 integrated HeLa cells (middle), and H-Mn loaded and S1, S2 integrated HeLa cells (bottom) after Mn2+ treatment. BF: bight-field. Atto 425: blue fluorescence. FAM: green fluorescence. Cy5: red fluorescence. Merge: mixed green and red channel. The FL intensities of different fluorochromes along the white line in the overlay images were shown in the histograms on the right. Scale bar: 10 μm. e FL intensity comparison histogram of different fluorochromes per cell in d. f PAGE analysis of the Mn2+-dependent activation of the DNAzyme (Dz1, Dz2) for substrate cleavage. Lane 1 – 7: ladder (lane 1), Seq1 (lane 2), Seq2 (lane 3), Sub1 (lane 4), a mixture of Seq1 and Seq2 (lane 5), a mixture of Seq1, Seq2, and Sub1 (lane 6), a mixture of Seq1, Seq2 and Sub1 upon Mn2+ treatment (lane 7). Experiments were repeated three times independently with similar results. Error bars in (e) represent the standard deviation of six independent experimental repeats, and the measure of the center represents their corresponding mean value. Source data are provided as a Source Data file.

To determine whether Seq1 and Seq2 could respond to external stimuli, a hairpin strand (H-Mn) with its loop part designed as the corresponding substrate sequence of the Mn2+-dependent DNAzyme, was delivered into Seq1/Seq2 integrated HeLa cells via liposomes (Fig. 1c). Initially, the red fluorescence (FL) of Cy5 tagged at the 3’-end of H-Mn was quenched by BHQ3 (quencher) tagged at the 5’-end (Supplementary Fig. 9). Triggered by the hybridization of two linkers (L1 tagged with blue fluorochrome and L2 tagged with green fluorochrome), separated Seq1 and Seq2 could form dimers, resulting in close spatial proximity between the Seq1 and Seq2 sequences. The dimerization could transmit the signals from the outside to the inside of the cell, which resulted in the generation of the complete structure of the Mn2+-dependent DNAzyme (composed of the intracellular parts of Seq1 and Seq2). In the presence of H-Mn, the specific recognition between the DNAzyme and H-Mn triggered the opening of the loop part of H-Mn, generating red FL on the cell membrane. Upon the addition of Mn2+ ions, the activated DNAzyme could catalyze the cleavage of H-Mn, resulting in the relocation of red FL from the cell membrane to the cytoplasm. As shown in confocal laser scanning microscopy (CLSM) images (Fig. 1d), obvious green and blue FL signals could be observed on the surface of HeLa cells after the treatment of lipo-Seq1 and lipo-Seq2, indicating successful integration of Seq1 and Seq2. After the addition of Mn2+ ions, the red FL can be observed in HeLa cells, which indicates the cleavage of H-Mn. The addition volume of lipo-Seq1 and lipo-Seq2, and the incubation time were optimized using HeLa cells (Supplementary Figs. 10 & 11). The average FL intensity of different fluorochromes per cell and 2D color-coded red channel intensity corresponding to HeLa cells incubated with transmembrane DNA receptors in Fig. 1d was shown in Fig. 1e and Supplementary Fig. 12. A negative control was performed using transmembrane DNA strands without a C12 spacer. As shown in Supplementary Fig. 13, negligible green and blue FL signals can be seen, which indicates the important role of the C12 spacer in integrating DNA strands into the cell membrane. Another control experiment using a pair of DNA strands without complementary domains was also performed. As shown in Supplementary Fig. 14, a negligible red FL signal was observed, indicating that no dimer was formed. Moreover, Seq1 and Seq2 with inactive DNAzyme were performed to form dimerization, and the experimental results showed no recovery of red FL, indicating no hybridization between DNAzyme and H-Mn, confirming the recognition efficiency of our strategy (Supplementary Fig. 15). Furthermore, the configuration changes of Seq1 and Seq2 including dimerization, substrate hybridization, and DNAzyme-controlled substrate cleavage in the presence of Mn2+ were confirmed by polyacrylamide gel electrophoresis (PAGE). As shown in Fig. 1f, a band with larger molecular weight in lane 5 emerged, indicating the formation of Seq1/Seq2 dimers. Upon the addition of H-Mn, a new higher band was observed, which corresponded to the hybridized structure of the Seq1/Seq2 dimer and H-Mn (lane 6). The addition of Mn2+ ions resulted in the generation of a new lower band, indicating the substrate cleavage (lane 7).

Together, these results demonstrated that the DNA-mediated artificial transmembrane signal transduction system was successfully constructed, which was feasible for signal recognition and transduction.

Design of logic gate-responsive DNA receptors

Previous results have proved that the dimerization of synthetic transmembrane DNA receptors facilitated the implementation of a dynamic cascade signal transduction process in response to external molecular signals. Next, we developed the aptamer-locked design rules to tune the extracellular signal reception part of the synthetic DNA receptors, to achieve more delicate control of signal input, as different logic functions are implemented (AND, OR). As shown in Fig. 2a, to perform the AND logic operation on the target cell membrane, a pair of AND logic-responsive DNA receptors (A-Seq1 and A-Seq2) were designed (Supplementary Table 2). The extracellular parts of A-Seq1 and A-Seq2 were designed to be aptamer sequences specifically targeting MUC 1 and Met, respectively (A-apt1 tagged with blue fluorochrome and A-apt2 tagged with green fluorochrome)49,50,51,52. Here we chose MUC 1 and Met as the target proteins, which are overexpressed on the surface of cancer cells53,54. The FL of A-apt1 and A-apt2 were quenched by corresponding complementary DNA strands (L-A1 and L-A2) via the Förster resonance energy transfer (FRET) effect. C12 spacers were used for membrane anchoring. The intracellular parts of A-Seq1 and A-Seq2 (A-C1 and A-C2) were designed to be complementary to form dimers and were tagged by a pair of fluorochrome/quencher (Supplementary Fig. 16). There are two steps for the implementation of the AND logic operation. For step 1, A-Seq1 was first integrated into the target cell membrane. For next step 2, A-Seq2 was integrated into the target cell membrane to form dimers with A-Seq1. Upon the dimerization between A-Seq1 and A-Seq2 due to the hybridization of A-C1 and A-C2, the red FL was quenched. Here, the inputs were two selected cell membrane proteins, MUC 1 and Met, (1 for presence, 0 for absence), while the output was the logic operation of the blue and green FL (1 for presence, 0 for absence). It was only in the presence of both MUC 1 and Met that the FL of A-apt1 and A-apt2 could be both recovered since the corresponding complementary DNA strands (L-A1 and L-A2) could be removed upon the recognition between the aptamer sequences and their corresponding proteins (MUC 1 and Met). Similarly, to perform the OR logic operation (Fig. 2b), a pair of OR logic-responsive DNA receptors (O-Seq1 and O-Seq2) was designed (Supplementary Table 3). Just similar to the design of A-Seq1 and A-Seq2, the extracellular parts, O-apt1 and O-apt2 were designed to be aptamer sequences specifically targeting MUC 1 and Met, respectively (tagged with blue/green fluorochrome). The FL of blue/green fluorochrome was quenched by partially complementary DNA strands (L-O1 and L-O2). C12 spacers were used for membrane anchoring. The intracellular parts, O-C1 and O-C2 were designed to be the same as A-C1 and A-C2. Notably, the partially complementary DNA strands, L-O1 and L-O2 were designed to be complementary with each other, thus the departure of either one of them could lead to the dissociation of the other (Supplementary Fig. 17). In this case, both blue and green FL could be seen in the presence of either MUC 1 or Met, thus the OR logic operation could be obtained.

Fig. 2: Molecular design and operational principles of logic gate-responsive DNA receptors.
Fig. 2: Molecular design and operational principles of logic gate-responsive DNA receptors.The alternative text for this image may have been generated using AI.
Full size image

a The design of AND logic gate-responsive DNA receptors (A-Seq1 and A-Seq2) on the cell surface for precise cell identification. Two recognition aptamers (A-apt1 and A-apt2) used for specifically targeting MUC 1 and Met, which are blocked by partially complementary DNAs (L-A1 and L-A2), two C12 spacers used for anchoring live cells bound with aptamers, and intracellular parts of A-Seq1 and A-Seq2 (A-C1 and A-C2) conjugated with dyes for monitoring DNA dimerization. b The design of OR logic gate-responsive DNA receptors (O-Seq1 and O-Seq2) on the cell surface for precise cell identification. Two recognition aptamers (O-apt1 and O-apt2) used for specifically targeting MUC 1 and Met, which are blocked by partially complementary DNAs (L-O1 and L-O2), two C12 spacers used for anchoring live cells bound with aptamers, and intracellular parts of O-Seq1 and O-Seq2 (O-C1 and O-C2) conjugated with dyes for monitoring DNA dimerization. c, d CLSM images of four cell lines incubated with AND and OR logic gate-responsive signal receptors at 37 C for 2 h. Step 1: the recognition between AptMUC 1 and MUC 1 after adding A-Seq1/O-Seq1; step 2: the recognition between AptMet and Met after adding A-Seq2/O-Seq2 and the dimerization between A-Seq1/O-Seq1 and A-Seq2/O-Seq1. Scale bar: 10 μm. e, f FL intensity comparison histogram of different fluorochromes per cell in (c, d). g, h PAGE analysis of the stepwise assembly of logic gate-responsive signal receptors. Lane 1 to 7 in AND logic gate: ladder (lane 1), A-Seq1 (lane 2), A-Seq2 (lane 3), L-A1 (lane 4), L-A2 (lane 5), a mixture of A-Seq1 and L-A1 (lane 6), a mixture of A-Seq2 and L-A2 (lane 7). Lane 1 to 7 in OR logic gate: ladder (lane 1), O-Seq1 (lane 2), O-Seq2 (lane 3), L-O1 (lane 4), L-O2 (lane 5), a mixture of O-Seq1 and L-O1 (lane 6), a mixture of O-Seq2 and L-O2 (lane 7). Experiments were repeated three times independently with similar results. Error bars in (e, f) represent the standard deviation of three independent experimental repeats, and the measure of the center represents their corresponding mean value. Source data are provided as a Source Data file.

Afterwards, the AND and OR logic gate-responsive DNA receptors were tested using four different cell lines, HeLa, T47d, HepG2, and MDA-MB-453. According to previous reports, both MUC 1 and Met are highly expressed on the surface of HeLa cells, while T47d overexpressed MUC 1 and HepG2 overexpressed Met. For MDA-MB-453, neither of the two proteins is expressed55. As shown in CLSM images (Fig. 2c), obvious blue and green FL signals could be observed on the surface of HeLa cells after step 2. Only blue/green FL could be seen on the surface of T47d/HepG2 cells. No FL was observed on the surface of MDA-MB-453 cells. These results confirmed the successful implementation of the AND logic operation. Figure 2d showed the CLSM images of different cells upon the treatment of OR logic gate-responsive DNA receptors. As expected, obvious blue and green FL signals were observed on the surface of HeLa, T47d, and HepG2 cells, while no FL was observed on the surface of MDA-MB-453 cells, which confirmed the successful implementation of the OR logic operation. The average FL intensity of different fluorochromes per cell in Fig. 2c & d was shown in Fig. 2e & f and Supplementary Fig. 18. A control experiment using A-Seq1/A-Seq2 and O-Seq1/O-Seq2 with mutated MUC 1 aptamer confirmed the targeting ability of the aptamer part, as the cell surface fluorescence was significantly reduced (Supplementary Fig. 19). Furthermore, the configuration changes of the logic gate-responsive DNA receptors including dimerization, aptamer binding, and protein-induced departure of complementary DNA strands were confirmed by PAGE (Fig. 2g, h).

Together, these results established a set of design rules for building synthetic transmembrane DNA receptors for more complex logic gate computation.

Design of DNAzyme-based dynamic cascade signal transduction

The successful operation of logic gate-responsive DNA receptors was then applied to implement more complex regulation of cell activities, in which three DNA-based modules could be replaced by other functional motifs. As shown in Fig. 3, a more delicate design was injected here to endow the system with the capacity to regulate both inter- and intracellular signal transduction, including T cell/cancer cell interactions and PD-L1 mRNA expression in cancer cells56,57,58,59,60. Fig. 3a depicted the designing details of the AND logic DNAzyme-based dynamic cascade signal transduction (AND logic DCST, DA-Seq1 and DA-Seq2), which contained two selective parts (DA-apt1 for the specific recognition of MUC 1 proteins and DA-apt2 for the specific recognition of Met proteins), two recognition parts (DA-L1 and DA-L2, complementary strands for dimerization), two anchoring parts (DA-C1 and DA-C2 for membrane anchoring), Mg2+-dependent DNAzyme (DA-R1 tagged with blue fluorochrome and DA-R2 tagged with green fluorochrome), and Mn2+-dependent DNAzyme (DA-Dz1 and DA-Dz2). The FL of DA-R1 and DA-R2 were quenched by L-DA1 and L-DA2 correspondingly via the FRET effect (Supplementary Table 4). The successful design of the AND logic DCST was first validated in vitro (Supplementary Fig. 20). Next, the AND logic DCST was tested using four different cell lines, HeLa, T47d, HepG2, and MDA-MB-453. According to the previous design, the AND logic DCST was set to activate the T cell/cancer cell interactions and the following PD-L1 mRNA suppression in HeLa cells which expressed both MUC 1 and Met on the cell membrane (1, 1, 1: the first 1 for the presence of MUC 1, the second 1 for the presence of Met, the third 1 for the activation of signal cascades after adding metal ions), while the other three cell lines could not be activated (Fig. 3b). There were three steps for the implementation of the AND logic DCST. For step 1, DA-Seq1 and DA-Seq2 were first integrated into the target cell membrane to form dimers. After encountering the target proteins (MUC 1 and Met), L-DA1 and L-DA2 could be removed and the FL of DA-apt1 and DA-apt2 could be recovered. For next step 2, the integrated cancer cells and T cells (modified with substrate chains on the cell membrane, tagged with BHQ1 quencher) interactions could be achieved through the hybridization between the complete Mg2+-dependent DNAzyme and its corresponding substrate, which quenched the green FL of DA-R2. Meanwhile, the complete Mn2+-dependent DNAzyme formed inside the cancer cells could hybridize with PD-L1 mRNA. For final step 3, T cells were released by metal ion (Mg2+) activated DNAzyme cleavage, enabling a quick release of T-cells from treated cancer cells (Fig. 3c). The green FL was recovered upon the Mg2+-induced cleavage of the substrate. In cancer cells, Mn2+ ions could activate the complete Mn2+-dependent DNAzyme for the cleavage of PD-L1 mRNA, thus reducing the expression level of PD-L1. As shown in CLSM images (Fig. 3d, step 1), obvious blue and green FL signals could be observed on the surface of HeLa cells. After the T cell/cancer cell interactions, the green FL of FAM can be quenched by BHQ1 at the 5’-end of the substrate, thus showing lower green fluorescence intensity (Fig. 3d, step 2), indicating the successful insertion and hybridization between Mg2+-dependent DNAzyme (DA-R1 and DA-R2) and substrate chains. Upon the addition of Mg2+ and Mn2+, substrate chains were cleaved from the ribonucleotide cleavage site, causing immediate T cell/cancer cell disassembly as well as the recovery of the FL of FAM (Fig. 3d, step 3). The corresponding average FL intensity (Fig. 3e) further confirmed the successful implementation of the above AND logic DCST process. In addition, the protein level of PD-L1 in DNA-treated cancer cells was measured by Western blotting (Supplementary Fig. 21). The results showed a significant downregulation of PD-L1 was observed in HeLa cells. These results demonstrated that the AND logic DCST treatment could induce PD-L1 inhibition in HeLa cells expressing both MUC 1 and Met. The performance of AND logic DCST was also confirmed by PAGE (Supplementary Fig. 22).

Fig. 3: Design of DNAzyme-based signal transduction system (DCST) for AND operation.
Fig. 3: Design of DNAzyme-based signal transduction system (DCST) for AND operation.The alternative text for this image may have been generated using AI.
Full size image

a The design of DNAzyme-based signal transduction construction for AND operation (AND logic DCST, DA-Seq1, and DA-Seq2). Two selective parts (DA-apt1 and DA-apt2) used for the specific recognition of MUC 1 and Met, two recognition parts (DA-L1 and DA-L2) as complementary strands used for dimerization, two anchoring parts (DA-C1 and DA-C2) used for membrane anchoring, Mg2+-dependent DNAzyme (DA-R1 and DA-R2) used for recognizing the corresponding substrate chains modified on the surface of T cells, and Mn2+-dependent DNAzyme (DA-Dz1 and DA-Dz2) used for the cleavage of PD-L1 mRNA. b Schematic illustration of AND logic DNAzyme-based signal transduction process for cell-selective activation. The DCST process was tested on four different cell lines: HeLa (1, 1, 1), T47d (1, 0, 0), HepG2 (0, 1, 0), and MDA-MB-453 (0, 0, 0). Here, the first two encodings represented the presence of MUC 1 and Met proteins as input (1 for presence, 0 for absence), while the third encoding represented the activation of the following signal cascades as output (1 for activation, 0 for nonactivated). c Schematic diagram of AND logic DCST treatment inducing PD-L1 inhibition in HeLa cells expressing MUC 1 and Met. d CLSM images showing the outputs of the AND operation under different cancer cells (different protein recognition sites). Scale bar: 20 μm. e Quantified fluorescence intensities corresponding to cell fluorescent images in (d). Average red and green channel intensity corresponding to Steps 1–3 of the outputs of the AND operation under different cancer cells. Step 1: the integration of DA-Seq1 and DA-Seq2 into the target cell membrane to form dimers, followed by the specific targeting of proteins (MUC 1 and Met); step 2: the recognition between cancer cells and T cells; step 3: the dissociation between integrated cancer cells and T cells in the presence of Mg2+ ions and the cleavage of PD-L1 mRNA in the presence of Mn2+ ions. Error bars represent the standard deviation of three independent experimental repeats, and the measure of the center represents their corresponding mean value. Source data are provided as a Source Data file.

Similarly, the design of OR logic DNAzyme-based dynamic cascade signal transduction (OR logic DCST, DO-Seq1 and DO-Seq2) was illustrated in Fig. 4a, which contained two selective parts (DO-apt1 and DO-apt2 for the specific targeting of the MUC 1 and Met, respectively), two recognition parts (DO-L1 and DO-L2 as complementary strands for dimerization), two anchoring parts (DO-C1 and DO-C2 for membrane anchoring, Mg2+-dependent DNAzyme (DO-R1 tagged with blue fluorochrome and DO-R2 tagged with green fluorochrome), and Mn2+-dependent DNAzyme (DO-Dz1 and DO-Dz2) (Supplementary Table 5). The FL of DO-R1 and DO-R2 were quenched by corresponding partially complementary DNA strands (L-DO1 and L-DO2) via the FRET effect. Notably, the loop parts of the partially complementary DNA strands L-DO1 and L-DO2 are designed to be fully complementary to DO-apt2 and DO-apt1, thus the departure of either one of them could lead to the dissociation of the other. Before cell experiments, the performance of the OR logic DCST was first validated in vitro (Supplementary Fig. 23). Next, the OR logic DCST was also tested using four different cell lines, HeLa, T47d, HepG2, and MDA-MB-453. According to the previous design, the OR logic DCST was set to activate the T cell/cancer cell interactions and the following PD-L1 mRNA suppression in cancer cells which expressed either MUC 1 or Met on cell membrane (HeLa: expressing MUC 1 and Met; T47d: expressing MUC 1; HepG2: expressing Met), while the MDA-MB-453 cells (without expressing MUC 1 or Met) could not be activated (Fig. 4b). There were three steps for the implementation of the OR logic DCST. For step 1, DO-Seq1 and DO-Seq2 were first integrated into the target cell membrane to form dimers. After encountering the target proteins (MUC 1 or Met), L-DO1/L-DO2 could be removed and then the released L-DO1/L-DO2 could induce the recovery of both blue and green FL. For next step 2, the integrated cancer cells and T cells (modified with substrate chains on the cell membrane, tagged with BHQ1) interactions could be achieved through the hybridization between the complete Mg2+-dependent DNAzyme and its corresponding substrate, which quenched the green FL of DO-R2. Meanwhile, the complete Mn2+-dependent DNAzyme formed inside the cancer cells could hybridize with PD-L1 mRNA. For final step 3, T cells were released by metal ion (Mg2+) activated DNAzyme cleavage, enabling a quick release of T-cells from treated cancer cells (Fig. 4c). The green FL was recovered upon the Mg2+-induced cleavage of the substrate. In cancer cells, Mn2+ ions could activate the complete Mn2+-dependent DNAzyme for the cleavage of PD-L1 mRNA, thus down-regulated the expression of PD-L1. As shown in CLSM images (Fig. 4d, step 1), obvious blue and green FL signals could be observed on the surface of HeLa, T47d, and HepG2 cells. After the T cell/cancer cell interactions, the green FL of FAM was quenched by BHQ1 at the 5’-end of the substrate (Fig. 4d, step 2), indicating the successful insertion and hybridization of these chains. Upon the addition of Mg2+ and Mn2+, substrate chains were cleaved from the ribonucleotide cleavage site, causing immediate T cell/cancer cell disassembly as well as the recovery of the FL of FAM (Fig. 4d, step 3). However, this process could not be observed in MDA-MB-453 cells. The corresponding average FL intensity (Fig. 4e) further confirmed the successful implementation of the above OR logic DCST process. The protein level of PD-L1 in HeLa cells treated with OR logic DCST was measured by Western blotting (Supplementary Fig. 21), which showed certain inhibition of PD-L1 in different cell types. These results demonstrated that the OR logic DCST treatment could induce PD-L1 inhibition in cancer cells (HeLa, T47d, and HepG2) expressing either MUC 1 or Met. The successful assembly of OR logic DCST was validated by PAGE analysis (Supplementary Fig. 24).

Fig. 4: Design of DNAzyme-based signal transduction system (DCST) for OR operation.
Fig. 4: Design of DNAzyme-based signal transduction system (DCST) for OR operation.The alternative text for this image may have been generated using AI.
Full size image

a The design of DNAzyme-based signal transduction construction for OR operation (OR logic DCST, DO-Seq1 and DO-Seq2). Two selective parts (DO-apt1 and DO-apt2) used for the specific recognition of MUC 1 and Met), two recognition parts (DO-L1 and DO-L2) as complementary strands used for dimerization, two anchoring parts (DO-C1 and DO-C2) used for membrane anchoring, Mg2+-dependent DNAzyme (DO-R1 and DO-R2) used for recognizing the corresponding substrate chains modified on the surface of T cells, and Mn2+-dependent DNAzyme (DO-M1 and DO-M2) used for the cleavage of PD-L1 mRNA. b, Schematic illustration of OR logic DNAzyme-based signal transduction process for cell-selective activation. The DCST process was tested on four different cell lines: HeLa (1, 1, 1), T47d (1, 0, 0), HepG2 (0, 1, 0), and MDA-MB-453 (0, 0, 0). Here, the first two encodings represented the presence of MUC 1 and Met proteins as input (1 for presence, 0 for absence), while the third encoding represented the activation of the following signal cascades as output (1 for activation, 0 for nonactivated). c Schematic diagram of OR logic DCST treatment inducing PD-L1 inhibition in cancer cells (HeLa, T47d, and HepG2) expressing MUC 1 or Met. d CLSM images showing the outputs of the OR operation under different cancer cells (different protein recognition sites). e Quantified fluorescence intensities corresponding to cell fluorescent images in (d). Average red and green channel intensity corresponding to Step 1–3 of the outputs of the OR operation for different cancer cells. Step 1: the integration of DO-Seq1 and DO-Seq2 into the target cell membrane to form dimers, followed by the specific targeting of proteins (MUC 1 and Met); step 2: the recognition between cancer cells and T cells; step 3: the dissociation between integrated cancer cells and T cells in the presence of Mg2+ ions and the cleavage of PD-L1 mRNA in the presence of Mn2+ ions. Error bars represent the standard deviation of three independent experimental repeats, and the measure of the center represents their corresponding mean value. Source data are provided as a Source Data file.

Together, these results showed that the successful operation of logic gate-responsive DNA receptors can be further applied to implement targeted regulation of cell activities.

Therapeutic efficacy of DNAzyme-based dynamic cascade signal transduction (DCST)

We next applied DCST to treat cancer cells. The therapeutic efficacy of the logic DCST strategy was tested under the co-incubation of cancer cells and T cells for 48 h (Fig. 5a, Supplementary Table 6). First, we investigated the targeted killing ability of logic DCST strategy by in vitro scratch assay to mimic wound healing behavior treated by AND and OR logic DCST. As shown in Supplementary Fig. 25, the wound closure of HeLa cells was significantly impeded compared with the other three cell lines upon the treatment of AND logic DCST. As expected, the treatment of OR logic DCST could inhibit the wound closure of HeLa, T47d, and HepG2, while MDA-MB-453 was not affected. These results demonstrated that the logic DCST strategy could cause the targeted killing of cancer cells based on logical operation results. Next, a T cell mediated cancer cell killing assay was performed. Activated T-cells (PBMCs) were modified with substrate strands and then cocultured with cancer cells with or without DCST treatment. Afterward, T-cells were removed, and the cancer cells were analyzed by CCK8 test (Fig. 5b) and Annexin-v-FITC/PI staining. From the CLSM images of Annexin-v-FITC/PI-stained cancer cells upon the treatment of AND and OR logic DCST (Fig. 5c), only HeLa cells showed obvious Annexin-v-FITC positive and PI positive signal, which indicates apoptosis upon the treatment of AND logic DCST, while HeLa, HepG2, and T47d cells all showed obvious Annexin-v-FITC positive and PI positive signal upon the treatment of OR logic DCST. In contrast, control experiments did not show such obvious treatment effect, which demonstrated the effective design of the DCST strategy to enhance T-cell infiltration (Fig. 5d). Corresponding average FL intensity further confirmed the above results (Fig. 5e). The cell viability experiments showed that the cancer cells remained alive after liposome vesicles treatment of 8 h (Supplementary Fig. 26), excluding the toxicity of liposome vesicles. These results suggested that the DNAzyme-based dynamic cascade signal transduction system could successfully achieve targeted cell analysis and precise synergistic killing, which demonstrated the feasibility of operating the systems in biological environments to be potentially applied in the diagnosis and treatment of diseases.

Fig. 5: Selective cancer cell killing by DCST.
Fig. 5: Selective cancer cell killing by DCST.The alternative text for this image may have been generated using AI.
Full size image

a Schematic illustration of DNAzyme-based signal transduction process on the cell surface for precise cell identification and targeted killing. Steps 1-3: recognition, dimerization/linkage, and disassembly. b Examination of the programmed cell death induced by different logic DCST by apoptosis assay using CCK-8 assay of different cell lines (HeLa, T47d, HepG2, and MDA-MB-453). To determine the cancer cell killing effect, AND/OR logic DCST were incubated with HeLa, T47d, HepG2, and MDA-MB-453 at 37 °C for 2 h. Then cells were washed and co-incubated with T cells to incubate for another 46 h, then removed T cells and replaced with fresh DMEM medium before testing with 10% CCK-8. c CLSM images showing the programmed cell death induced by different logic DCST by apoptosis assay using Annexin-v-FITC (green)/PI (red) double staining of different cell lines (HeLa, T47d, HepG2, and MDA-MB-453). Scale bar: 30 μm. d The heat map of SUDA induced selective cell death of different cell lines, which was determined by adding the sum of dead cells (late apoptosis, PI + /Annexin-v +) and dying cells (early apoptosis, PI-/Annexin-v +). e FL intensity comparison histogram of cancer cell-killing under different treatments in (c). Error bars in (b) represent the standard deviation of three independent experimental repeats, and the measure of the center represents their corresponding mean value. Error bars in (e) represent the standard deviation of ten independent experimental repeats, and the measure of the center represents their corresponding mean value. Source data are provided as a Source Data file.

Discussion

In this study, we show that the DNA-mediated artificial transmembrane signal transduction system can be systematically engineered and actuated for customized signal recognition and transduction. The synthetic DNA receptors are composed of three parts, including an extracellular signal reception part, a lipophilic transmembrane anchoring part, and an intracellular signal output part. Upon the input of external signals, the synthetic DNA receptors can form dimers on the cell surface, leading to a series of downstream cascade events. We first demonstrated the dimerization of synthetic transmembrane DNA receptors on living cell membrane. Next, we developed the aptamer-locked design rules to tune the extracellular signal reception part of the artificial DNA receptors, enabling more delicate control of logic gate computation. The successful operation of logic gate-responsive DNA receptors was further applied to implement more complex regulation of cell activities, in which three DNA-based modules can be replaced by various functional motifs. To further highlight the synthetic DNA receptors’ capability for practical applications, we developed a DNAzyme-based dynamic cascade signal transduction (DCST) system that can be used for synergistic cancer cell killing. We believe this strategy opens the door for other user-defined sensing and actuation effects that is broadly applicable to implement sophisticated biological functions. With its specificity and versatility, this approach is well-suited for identifying and targeting specific cell types within heterogeneous environments, which is crucial for personalized medicine. Our future research will focus on expanding the reach of this strategy to multiple applications in live organisms and clinical environments, such as non-invasive diagnostics and targeted therapies.

Ethical statement

Our research does not include any studies involving humans or animals.

Methods

Materials

HeLa, T47d, HepG2, MDA-MB-453, and T-cells were purchased from bluefbio (Shanghai) Biology Technology Development Co., Ltd. (Shanghai, China). Lipo 2000, DNase I, and Annexin-v FITC/PI were obtained from ThermoFisher Scientific Inc (Waltham, MA, USA). Cell Counting Kit8 (CCK-8) was obtained from Energy Chemical. MnCl2 and MgCl2 were purchased from Adamas-beta® (Shanghai, China). DNA markers were purchased from ZOMANBIO. Cholesterol was purchased from J&K Scientific. 1,2-dipalmitoylphosphatidylcholine (DPPC), 1,2-dioleoyl-sn-glycero-3-phosphocholine lipid monomer (DOPC), 1,2-dioleoyl-3-trimethylammonium-propane (DOTAP) were purchased from Shanghai Macklin Biochemical Co., Ltd. β Actin antibody, PD-L1 antibody, and Goat Anti-Rabbit IgG (HRP) were purchased from Shanghai Ebiwei Biotechnology Co., Ltd. All other chemical reagents were of analytical grade and purchased from Sigma-Aldrich, Inc. (USA). All solutions were prepared using ultrapure water obtained through a Millipore Milli-Q water purification system (Billerica, MA, USA), with an electric resistance of >18.2 ΜΩ. All DNA strands were synthesized and purified by Sangon Biotech (Shanghai) unless otherwise specified.

All DNA products were quantified and stored in ultrapure water (Milli-Q) for subsequent experiments. The information of DNA sequences is given in Supplementary Table. 1–5.

Apparatus

All the fluorescence images were obtained using a water dipping objective (60×) on a laser scanning confocal microscope (Nikon AlR., Japan). All materials were weighed by an analytical balance (ME 104, METTLER TOLEDO), and all reagents were centrifuged by a centrifuge (Centrifuge 5430, Eppendorf). UV-Vis absorbance spectra were recorded on a USB 2000+ spectrometer (Ocean Optics Inc., USA). Fluorescence spectra were measured using an F97pro FL spectrophotometer (Lengguang, Shanghai).

Making DNA-integrated liposome vesicles

Cholesterol, 1,2-dipalmitoylphosphatidylcholine (DPPC), 1,2-dioleoyl-sn-glycero-3-phosphocholine lipid monomer (DOPC), 1,2-dioleoyl-3-trimethylammonium-propane (DOTAP) were utilized in this study. DOPC, DPPC, cholesterol, and DOTAP were dissolved in chloroform at a molar ratio of 1:1:2:0.2. Liposome preparation followed the film hydration method, where the solvent was completely removed under vacuum, and the resulting lipid film was hydrated with 1 mL of PBS buffer at 37 °C. Transmembrane DNA chains were then combined with the lipid mixture and incubated in 1× PBS buffer (pH 7.4) at 37 °C for 4 h. To form DNA-integrated liposome vesicles, the solution was extruded through 200 nm polycarbonate membranes (Whatman; Diegem, Belgium) using a mini-extruder (Avanti Polar Lipids). The mixture was gently vortexed and left to incubate at room temperature overnight45. The particle size distribution and surface charge were subsequently characterized using dynamic light scattering (DLS) and zeta potential analysis (Supplementary Fig. 2). Two different DNA-liposomes were used in separate fusion steps to ensure greater controllability, stability, flexibility, and adaptability. This approach minimizes dimer formation, reduces structural risks, and allows for system customization to suit diverse applications. To assess the DNA transmembrane efficiency on vesicle surfaces, DNA-integrated liposome vesicles were prepared using transmembrane DNA strands labeled with fluorescent dyes at both termini (Atto 425 and Cy5) (Supplementary Fig. 3a). The Atto 425 and Cy5 FL intensity was measured after removing unreacted DNA strands by ultrafiltration using Amicon 30k at 10900 x g in 15 min. The DNA-integrated liposome vesicles were then incubated with 20 U/mL DNase I at 37 °C for 15 min to degrade any DNA exposed on the vesicle surface or the external portion of the transmembrane DNA. DNase I-treated DNA-integrated liposome vesicles were centrifuged by ultrafiltration using Amicon 30k at 10900 x g in 15 min. The FL intensity of remained solution was measured. Based on the change in FL intensity (Supplementary Fig. 3b), the transmembrane efficiency of DNA on the liposome surface was determined to be 95%.

Preparation of H-Mn chains encapsulated liposome

2 μL, 100 μM of H-Mn chains were mixed with 20 μL of serum-free medium, while 1.0 μL of Lipo-2000 was dissolved in 20 μL of serum-free medium and incubated for 10 min at room temperature. After the two solutions were mixed and incubated for 30 min at room temperature, H-Mn chains encapsulated liposomes were obtained.

Gel electrophoresis analysis

To study the assembly of synthetic transmembrane DNA receptors, agarose gel electrophoresis (AGE) was carried out in 1× TBE buffer at 75 V for 45 min. All samples were mixed with 6 x loading buffer (GenStar Biotech.) before gel electrophoresis. Fluorescence-free gels were stained by Super GelRed (US Everbright Inc.). All gels were imaged using a Bio-Radmolecular imager under blue light.

Cell culture

HeLa (cervical cancer cell), HepG2 (liver cancer cell), and T47d (breast cancer cell) were cultured in DMEM media supplemented with 10 % fetal bovine serum, 100 U/mL penicillin, and 100 mg/mL streptomycin at 37 °C in 5 % O2 atmosphere. CD8 + T cells were maintained in CD8 + T cells complete medium obtained from bluefbio (Shanghai) Biology Technology Development Co., Ltd. Trypsin (Beijing T&L Biological Technology Co., Ltd.) was used to digest these cells.

DNA modification on cell membrane

For the DNA-integrated vesicle-cell fusion process, cancer cells were incubated with 40 µL DNA-integrated liposome vesicles in DMEM solution at room temperature for 2 h, washed (300 x g, 3 min) twice with pH 7.5 0.01 M PBS buffer. As shown in Supplementary Fig. 4, we quantitated Seq1/Seq2 with the signal recognition part or downstream signal output part at the outer surface of the cell membrane by treating HeLa cells with DNase I49,61. The DNA chains exposed outside the membrane could be digested by DNase I, resulting in reduced FL signals. The percentage of Seq1/Seq2 with the signal recognition part at the outer surface of the cell membrane were calculated to be about ~ 50 %.

Characterizations of DNA strands on HeLa cells

To calculate the fusion efficiency and quantify DNA on cell membranes, DNA-integrated liposome vesicles were prepared and the FL intensities of supernatant after fusion were measured.

First, the fusion efficiency of two different DNA-liposomes (lipo-Seq1 and lipo-Seq2) was evaluated by fluorescence spectra. At low concentrations, the FL intensity of the solution is proportional to the concentration of the fluorescent molecules. Vesicle fusion efficiency η was calculated using F0-F/F0 by measuring the FL intensity of lipo-Seq1/Seq2 (F0) and the culture medium (F) after the vesicles were co-incubated with cells. As shown in Supplementary Fig. 5, approximately 12% and 14% of lipo-Seq1 and lipo-Seq2, respectively, were removed from the supernatant, indicating their fusion efficiency and successful incorporation into cells.

To quantify DNA modifications on the cell membrane, we measured the FL intensity of the two types of DNA sequences at different concentrations to establish the linear relationship between the concentration and the FL intensity (Supplementary Fig. 6). Next, the fluorescence of lipo-Seq1/Seq2 and the supernatant after ultrafiltration to remove vesicles was then measured. The fluorescence of the supernatant after vesicle removal corresponds to a DNA concentration close to zero on the standard curve, allowing the amount of DNA bound to the vesicles to be determined. During the preparation of lipo-Seq1/Seq2, the volume of DNA used for the synthesis was 5 μL (100 μM). Thus, the average number of DNA molecules (N) on each cell can be determined according to the following formula:

$$N=\frac{{{{\rm{n}}}}{{{\rm{\eta }}}}{{{{\rm{N}}}}}_{A}}{\rho V}$$
(1)

Where n is the amount of DNA carried on the vesicle, η is the fusion efficiency of liposomes, NA is Avogadro’s constant, ρ is the cell density, and V is the volume of cell medium. According to the above formula, the number of Seq1 and Seq2 modified on cell surface was calculated to be 2.9 × 109 per cell and 3.4 × 109 per cell, respectively.

Fluorescence measurement

All fluorescence measurements were performed on a G9800 Agilent fluorescence spectrometer (America). The A-Seq1/L-A1, O-Seq1/L-O1, O-Seq2/L-O2, DA-Seq1/DA-L1, DA-Seq2/DA-L1, DO-Seq1/DO-L1, and DO-Seq2/DO-L1 were annealed separately at 95 oC for 5 min and cooled slowly to room temperature. The fluorescence spectra of 200 nM DNA chains in PBS buffer solution were measured at an excitation wavelength of 436 nm, 494 nm, 565 nm, and 643 nm.

Confocal laser scanning microscopy imaging

Confocal fluorescence microscopy imaging was performed using a Nikon A1 confocal fluorescence microscope (Japan) with × 60 oil objective lens. The collection wavelength ranges were 450 − 500 nm for the FAM channel excited at 405 nm, 500 − 550 nm for the FAM channel excited at 488 nm, 570 − 620 nm for the TAMRA channel excited at 561 nm, and 663 − 738 nm for the Cy5 channel excited at 561 nm.

Selective cancer cell killing by AND/OR logic DCST

To determine the selective cell-killing effect of the AND/OR logic DNAzyme-based dynamic cascade signal transduction (DCST), experiments were designed to leverage the specific surface markers MUC 1 and Met expressed on different cancer cells. Two DNA strands, DA-Seq1 and DA-Seq2, were engineered for AND/OR logic operations. Each strand comprised an extracellular aptamer domain targeting MUC 1 or Met (DA-apt1 and DA-apt2, respectively), a transmembrane domain, and an intracellular DNAzyme domain (DA-Dz1 and DA-Dz2). Complementary quenching strands (DA-L1 and DA-L2) were hybridized to the aptamer domains to suppress activation until specific markers were present. Four cancer cell lines (HeLa, T47d, HepG2, and MDA-MB-453) were selected for evaluation based on their marker expression profiles. HeLa cells express both MUC 1 and Met, enabling AND logic activation, while T47d and HepG2 cells express MUC 1 or Met, respectively, supporting OR logic activation. MDA-MB-453 cells, with minimal expression of both markers, served as a negative control.

For the experimental protocol, the cells were first incubated with DA-Seq1 and DA-Seq2 liposomes at 37 °C for 2 h to allow fusion of DCST components onto the cell membrane. After washing with PBS, the cells were treated with Mn2+ ions to activate the DNAzyme. In the AND logic DCST, hybridization of DA-L1 and DA-L2 occurred only when both MUC 1 and Met were present, while in the OR logic DCST, activation required either MUC 1 or Met. Following signal activation, the cells were co-incubated with T cells for 46 h at 37 °C, during which the activated DNAzyme suppressed PD-L1 expression, enhancing T-cell-mediated apoptosis. The effectiveness of this process was assessed using multiple approaches, including confocal fluorescence imaging to monitor DNAzyme activation and PD-L1 suppression, Annexin V/PI staining to confirm apoptosis, and western blot to quantify PD-L1 expression levels.

Cytotoxicity measurements

All cytotoxicity measurements were tested with 10 % CCK-8 by collecting the absorbance at 450 nm using a Synergy 2 Multi-Mode Microplate Reader (Bio-Tek). All cells (a total of 4 × 103) were plated in a 96-well plate and incubated overnight for adherence before treatment in DMEM medium. To determine the cancer cell killing effect, AND/OR logic DCST were incubated with HeLa, T47d, HepG2, and MDA-MB-453 at 37 °C for 2 h. Then cells were washed and co-incubated with T cells to incubate for another 46 h, then removed T cells were and replaced with fresh DMEM medium before testing with 10 % CCK-8.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.