A DNA-based nanodevice senses purinergic signaling and drives an immune switch for resolving inflammation

Abstract

Purinergic signaling dysregulation (e.g., excessive extracellular ATP, exATP) plays a critical role in the pathology of inflammatory disorders, but current efforts in drug development for blocking purinergic receptors are unsatisfactory. Here, inspired by natural metabolite sensing/signaling system, we develop a DNA origami-based ATP-sensing nanodevice (ND) for fine-tuning purinergic signaling and immune homeostasis. This ND composes a tubular DNA origami equipped with ATP sensors and the catalytic subunits (ENPP1-CD73 pairs), which can sense high levels of exATP and then expose the catalytic subunits for metabolizing exATP to adenosine, thereby driving an immune switch from exATP-mediated proinflammatory signals to adenosine-mediated immunosuppressive signals. Further surface displaying of ND on the monocytes (ND@Monos) enables its active inflamed site-targeting to restore immunometabolic hemostasis and reduce inflammation in diverse models in vivo. This study highlights that design of metabolite-sensing NDs is a promising strategy for controlling the homeostasis of cell metabolism and the immune response.

Purinergic signalling dysregulation can drive inflammatory disorders, but effective targeted therapy is still lacking. Here, authors develop a smart DNA origami nanodevice that can precisely sense and metabolize high levels of extracellular ATP, thereby fine-tuning purinergic signalling and immune homeostasis.

Introduction

Purines are a class of functional small molecular compounds that play a fundamental role in mediating various biological processes¹. For instance, adenosine triphosphate (ATP) is an indispensable metabolite for the survival and function of most cell types. However, during various forms of tissue injury, large amounts of intracellular ATP are released from stressed, damaged or necrotic cells into the extracellular space, and such excessive extracellular ATPs (exATPs) can serve as damage-associated molecular patterns (DAMPs) to recruit and activate diverse immune cells (e.g., monocytes, macrophages, and neutrophils) by binding to cell-surface purinergic receptors (e.g., P2XR and P2YR), thereby eliciting a proinflammatory immune response and aggravating tissue injury². Although purinergic signaling has been proposed as a promising therapeutic target for treating inflammatory disorders and autoimmune diseases and a few inhibitors targeting purinergic receptors (e.g., P2X7R and A2BR) have been reported, their clinical efficacy is still unsatisfactory owing to their nonspecific tissue distribution, short half-life, off-target effects (disrupting purinergic receptors on normal cells) and severe side effects in vivo². Thus, novel strategies that can restore local purinergic signaling homeostasis at injury sites in a targeted manner are desirable.

The homeostasis of extracellular purinergic signaling and immune homeostasis are tightly regulated by a series of key ectonucleotidases (e.g., ENPP1, CD39, and CD73) expressed on the cell surface, which can sense and metabolize excessive exATP to adenosine diphosphate (ADP), adenosine monophosphate (AMP) and adenosine (ADO, an immunosuppressive metabolite), thereby promoting the resolution of inflammation and tissue injury repair. However, previous studies have shown that the expression and/or activity of ectonucleotidases, such as nucleoside triphosphate diphosphohydrolase 1 (NTPDase1) / CD39 and 5’-nucleotidase ecto (NT5E) / CD73, can be impaired in chronic inflammatory or autoimmune states and that this effect is strongly associated with disease severity^3,4. Thus, instead of purinergic receptor inhibition, targeted restoration of the local purinergic catabolic machinery in injured tissue sites by engineering approaches might be a more precise strategy for treating inflammatory disorders. Inspired by the natural process of purine catabolism-mediated inflammation resolution, we aimed to design a smart nanodevice that can precisely sense exATP and recapitulate the spatial expression pattern and action model of the extracellular ectonucleotidase system. Consequently, this nanodevice can modulate the local immune response through on-demand conversion of excessive harmful exATPs to proresolution metabolites at inflamed sites.

DNA nanotechnology is capable of fabricating predictable and flexible DNA origami nanostructures with homogeneous shapes and sizes on the basis of Watson−Crick base pairing^5–7. The sub-10 nm addressability of DNA origami enables the integration of various functional moieties (e.g., sensing moieties, bioactive molecules) in a precisely controllable and site-specific manner^8–10. Their dynamic reconfigurability allows for the shielding or exposure of diverse bioactive elements based on different physiological and pathological conditions in vivo^11–13. Previous studies have demonstrated that DNA origami, when integrated with sensing moieties (such as pH aptamers and disulfide bonds) and therapeutic payloads, can be used to fabricate intelligent nanodevices to selectively release therapeutics at desired sites and reduce the effects of overdoses^14–16. This precise and elaborate DNA origami-based nanodevice is an ideal tool for recapitulating the natural assembly pattern and mode of action of extracellular ectonucleotidases for controlling local exATP metabolism and the immune response, but it has not yet been reported.

In this study, inspired by natural metabolite sensing and signaling processes, we develop a DNA origami-based ATP-sensing nanodevice (ND) for the regulation of purinergic signaling and immune homeostasis. The ND is generated from a tubular DNA origami system that is equipped with ATP sensors on its hollow sidewall and the catalytic subunits (consisting of ENPP1 and CD73) in its inside cavity. This ND can sense high levels of exATP, followed by rapid deformation to expose the catalytic subunits for the conversion of exATP into ADO, thereby switching ATP-mediated proinflammatory signals to ADO-mediated immunosuppressive signals to restore immune homeostasis. To enable inflamed site-targeted delivery, we further anchor the ND on the monocyte surface and demonstrated that this ND@Monos system can effectively reduce inflammation and tissue injury in diverse models in vivo. This study highlights that tailoring the design of metabolite-sensing NDs is a promising strategy for controlling the homeostasis of cell metabolism and the immune response.

Results

Design and characterization of ATP-sensing DNA origami nanotubes

The operation of a biological system (e.g., cell) requires extensive interaction between the biological machinery and the chemistry of its environment (e.g., metabolites); thus, metabolite sensing is one of the most fundamental biological processes in living organisms¹⁷. Cells can employ a wider spectrum of complicated mechanisms to sense metabolic changes and modulate signaling pathways, and a simplified working model commonly consists of a metabolite sensor, transducer and effector subunit (Fig. 1a). In brief, the sensor protein detects metabolite changes and then transforms the chemical signals into cell signaling pathways via transducer proteins. For example, AMPK (a key cell energy status-sensing protein) is a heterotrimer complex composed of three subunits (α, β and γ), and the γ subunit can sense the AMP/ATP ratio by binding AMP. The AMP-bound γ subunit then induces a conformational change in the AMPK heterotrimer, enabling the exposure of the catalytic pocket of the α subunit, which further transmits signals to its downstream effectors¹⁷.

Fig. 1. Design and characterization of the ATP-sensing NT.

a Model of metabolite sensing and signaling composed of a metabolite sensor, a signal transducer, and an effector. M indicates a metabolite. b Design of the ATP sensing and signaling model. A reconfigurable tubular DNA origami was locked by ATP aptamers (ATP sensors) and could change its conformation to expose two ectonucleotidases (ENPP1/CD73) after the ATP sensors bound to ATP molecules. The exposure of ENPP1/CD73 could subsequently convert ATP into adenosine (ADO), and further activate the anti-inflammatory signals. AMP: adenosine monophosphate. c Fluorescence kinetics of the dissociation of the lock_ATP duplex. The lock_ATP strand was modified with a Cy5-BHQ3 pair. The Cy5 fluorescence was initially quenched by BHQ3 due to the short distance, and could be recovered after the lockATP strand bound to ATP molecules. Cy5: Cyanine 5, BHQ3: black hole quencher 3. d Schematic of the ATP-sensing NT structure. e Principle for monitoring the unlocking process of NTs (top) and the real-time fluorescence kinetics of the unlocking NTs with different ATP concentrations (bottom). f Representative AFM images of the NTs in different states (S1: locking, S2: partial unlocking, S3: partial locking, S4: completed unlocking). L locked state, U unlocked state. Scale bars: 400 nm. The measurement of the height of the NT or NS (bottom). Scale bars: 50 nm. The averages of n = 3 independent experiments are shown in (c, e, f). Source data are provided as a Source Data file.

Inspired by such a natural metabolite sensing and signaling model, we designed a DNA origami-based nanodevice (ND) that includes ATP-sensing, catalytic, and signaling transmit subunits (Fig. 1b), which is a reconfigurable tubular DNA origami that contains ATP sensors (8 pairs) on its hollow sidewall and catalytic subunits (ENPP1 and CD73) inside its cavity. The ATP sensor is an ATP-responsive DNA duplex that contains an ATP aptamer (in green) and its partial complementary strand (in gray). The ATP aptamer has a high affinity for ATP molecules and can change its conformation after binding to ATP, causing dissociation of the duplex. The ATP sensor is also designed as a locker (named lock_ATP), where two ends of the ATP sensor can hybridize with the M13 scaffold on the edge of the DNA origami and maintain the tubular structure of the ND (Fig. 1b). The catalytic and signaling transmit subunits (two ectonucleotidases, ENPP1/CD73) are placed on the lumen side of the tubular DNA origami. In this design, excessive exATP can function as the “key” to unlock the ATP sensor and cause deformation of the ND. Upon ATP binding, the ND can be rapidly unlocked, followed by ENPP1/CD73 exposure and subsequent ATP-ADO conversion to initiate cellular signaling transmission (Fig. 1b).

The lock_ATP duplex was synthesized and then characterized by a fluorescence-quencher pair, where the ATP aptamer was modified with a Cy5 fluorophore and the partial complementary strand was modified with a BHQ3 quencher (Fig. 1c). The ATP sensing and switching of the lock_ATP duplex was monitored by its fluorescence kinetics. The addition of ATP (0-3 mM) induced a gradual increase in fluorescence intensity over time (0-240 min) due to the separation of Cy5 from BHQ3, indicating that the lock_ATP duplex can bind ATP and is dissociated by ATP in a concentration-dependent and time-dependent manner. Next, we prepared tubular DNA origami (Supplementary Fig. 1). A rectangular DNA origami nanosheet (NS) was first assembled by folding an M13 scaffold with predesigned staples and then curled into a tubular structure using 8 pairs of lock_ATP strands, generating an ATP-sensing DNA origami nanotube (NT, Fig. 1d and Supplementary Fig. 2a–c), and the yield of the NT was ~90 ± 2.0% (NT/NS, Supplementary Fig. 2d). The NSs or NTs were then stabilized by coating with PEG-oligolysine as previously reported^8,18. Agarose gel electrophoresis (AGE) analysis revealed that the migration rate of the NTs or NSs decreased (Supplementary Fig. 3a, b), which may have been due to the mass increase in the NTs or NSs after PEG-oligolysine coating. The NTs coated with PEG-oligolysine were stable even after incubation with 10% FBS for 24 hours (Supplementary Fig. 3c). Therefore, all the DNA origami-based structures (NS, NT, and ND) were coated with PEG-oligolysine in subsequent experiments to ensure their stability under physiological conditions. The fluorescence kinetics of the NTs revealed that the NTs exhibit high selectivity and specificity responses to ATP compared to the other structurally related molecules, such as GTP, CTP, and UTP (Supplementary Fig. 4). The unlocking process of NTs by ATP is also concentration-dependent and time-dependent (Fig. 1e). The results collectively verify that NT has high selectivity and specificity to ATP. In contrast, the NT locked by random-sequence DNA double strands, named NT (lock₁), was unable to be unlocked by ATP (Supplementary Fig. 5a–d). Atomic force microscopy (AFM) images revealed that the configuration of the locked NTs could be changed to NS under ATP addition (0.1–6 mM, Fig. 1f). Specifically, NT maintained its tubular structure in the absence of ATP (0 mM, state 1, S1) while it was gradually unlocked into a planar structure, and the percentage of planar structures increased with increasing ATP concentration (from 0.1 to 2 mM, S2, S3). Finally, all the tubular structures switched into planar structures at high concentrations of ATP (6 mM, S4), which is in line with our “key-lock” design.

Fabrication of NDs for controlling purinergic signaling

The ectonucleotidases (e.g., ENPP1 and CD73) are vital functional units in ND for the regulation of exATP levels and purinergic signaling. ExATP can be converted to ADP and AMP via ENPP1, and AMP can be further converted to ADO via CD73 (Fig. 2a). Inspired by the expression pattern and mode of action of ectonucleotidases¹⁹, paired-ENPP1 and CD73 proteins are designed to load into the ND lumen. To achieve this, ENPP1 and CD73 were chemically conjugated with two thiolated DNA single strands (SH-ssDNA1 and SH-ssDNA2, respectively; Supplementary Fig. 6). Ten capture strands were extended from the predesigned location on the NS surface for hybridization with the ssDNA1-ENPP1 conjugate and the ssDNA-CD73 conjugate, allowing the formation of the ENPP1&CD73-NS complex (E&C-NS). Then, the E&C-NSs were annealed with 8 lock_ATP strands and curled into a tubular structure, resulting in the formation of NDs (Fig. 2a). AGE images revealed that the fluorescent bands of ENPP1, CD73 and ND overlapped well, confirming the successful engineering of ENPP1 and CD73 into the ND (Fig. 2b). Importantly, both the number and position of ENPP-1 or CD73 on the NS surface could be precisely controlled, as evidenced by the AFM images, which revealed that five ENPP1 or CD73 molecules were anchored accurately at the predesigned positions on the NS surface (Fig. 2c, Supplementary Fig. 7).

Fig. 2. Synthesis and characterization of the ATP-sensing ND.

a Schematic of ATP catabolism mediated by extracellular ectonucleotidases (top) and the synthesis of ATP-sensing NDs that can recapitulate the spatial expression patterns and working model of these enzymes. b AGE characterization of the ND. ENPP-1 was labeled with Alexa488 (in green), and CD73 was labeled with Alexa647 (in blue). DNA was stained with GelRed (in magenta). E&C: ENPP1 and CD73. c Representative AFM images of NSs, protein-NSs or NDs and their corresponding line profiles across the structures. Scale bars: 50 nm. d The concentration threshold of ATP for triggering ND unlocking. e The consumption of ATP by an ND with two different locks. f The balance of ATP concentration maintained by ND (lock_ATP) (n = 3 independent measurements). g Experimental scheme of the ATP catabolic potency of ND analyzed via liquid chromatography-coupled mass spectrometry (LC‒MS). h LC‒MS chromatograms of ATP, AMP and ADO. i Intensities of ATP, AMP and ADO. The averages of n = 3 independent experiments are shown in (b–f, h). The averages of n = 6 independent experiments are shown in (i). Data are presented as mean ± SD. Source data are provided as a Source Data file.

Next, we assessed whether this ND can sense pathological levels of exATP and then expose ENPP1-CD73 to convert ATP to ADO locally. The concentration of local exATP is relatively low (in the nM range) under physiological conditions, but it can dramatically increase in diseased (e.g., higher than mM in the tumor TME) and injured (up to hundreds of μM) tissues due to the massive release of ATP by stressed or dead cells²⁰, and the activation of purinergic receptors (e.g., P2X7) requires a high concentration (>100 μM) of ATP²¹. Thus, we further optimized the sensitivity of the ND to ATP and yielded a concentration threshold (C_ATP threshold) of 46.9 µM for ATP-unlocking NDs (Fig. 2d), which was below the critical exATP concentration for purinergic receptor activation. The optimized ND with lock_ATP strands (ND-lock_ATP) could consistently hydrolyze ATP and convert it into ADO, whereas the ND locked by random-sequence lock1 strands (ND-lock1) failed to hydrolyze ATP even at millimolar concentrations (Fig. 2e).

To test the capacity of ND-lock_ATP to control purinergic metabolism, we incubated ND-lock_ATP with ATP and simultaneously monitored the dynamic changes in the unlocking process of ND and ATP catabolism in this system. Upon the first round of ATP (1 mM) addition, ND unlocking progressed, and a gradual decline in ATP levels was observed. The process of ND unlocking was halted once the ATP concentration decreased near the C_ATP threshold. After adding the second round of extra ATP (1 mM), the processes of ND unlocking and ATP catabolism continued, and the levels of ATP finally remained low (Fig. 2f). Moreover, the ATP catabolic potency of ND was analyzed via liquid chromatography-coupled mass spectrometry (LC-MS) (Fig. 2g), and ND was able to degrade ATP to produce AMP and ADO (Fig. 2h, i). Collectively, these results indicate that this designed ND might sense exATP in pathological states and control purinergic signaling via ATP-ADO conversion.

To assess whether ND construct alter the loaded proteins’ activity, ATP hydrolysis reactions catalyzed by free ENPP1 or the ENPP1-loaded ND (ND) were performed (Supplementary Fig. 8). The results showed that the catalytic efficiency constant (K_cat/K_m) for ENPP1 and the ND were comparable, indicating that they have approximately the same catalytic efficiency. However, the maximum velocity (V_max) in the ND group was slightly less than that of the ENPP1 group. This might be because, after ENPP1 was loaded onto the DNA origami surface, some of its active sites were covered by the DNA origami due to steric hindrance, which partially affect its maximal catalytic rate. Moreover, the enzymatic activities of free CD73&ENPP1 and CD73&ENPP1-loaded ND were also assayed. Consistently, the enzymatic activities of CD73&ENPP1-loaded ND were similar with those of free CD73&ENPP1 (Supplementary Fig. 9). Together, these results indicate that the ND construct has minor impact on enzymatic activity of the loaded proteins (e.g., ENPP1, CD73).

ND regulates purinergic signaling to drive the immune switch in vitro

Having revealed the ATP-sensing and catabolic effects (metabolizing ATP to ADO) of this ND, we sought to evaluate its biological function in immune cells in vitro. Following tissue injury, high levels of exATP from damaged or dead tissue cells bind to purinergic receptors (e.g., P2X7) and further increase the proinflammatory phenotype and cytokine secretion of diverse types of immune cells (e.g., monocytes, macrophages, and neutrophils) via multiple mechanisms, such as mitochondrial stress (mtROS), the NLRP3 inflammasome, and the nuclear factor-κB (NF-κB) pathway (Fig. 3a)^22–24. Consistent with previous reports^23,24, we found that high concentrations of exATP stimulation induced mitochondrial damage, such as increased mtROS, reduced mitochondrial membrane potential (MMP) and cytokine/chemokine (e.g., IL-1β, IL-6, and MCP-1) expression in LPS-primed M1 macrophages (Fig. 3a–h and Supplementary Fig. 10). In contrast, ND treatments significantly reduced the levels of exATP-induced cytokine and chemokine expression in M1 macrophages, and the anti-inflammatory potency of ND was similar to that of free ENPP1 and CD73 proteins (E&C, at equal concentrations), whereas treatment with NT alone (without ENPP1&CD73) failed to reduce exATP-induced cytokine expression in macrophages (Fig. 3b–h). Collectively, these results demonstrate that this ND can sense exATP and suppress the hyperactivation of immune cells.

Fig. 3. NDs regulate purinergic signaling pathways in immune cells in vitro.

a Experimental scheme for detecting the regulatory effect of NDs on purine signals in immune cells in vitro. b qRT-PCR analysis of the expression of proinflammatory genes (IL-1β, IL-6 and MCP-1) in RAW264.7 cells (***P < 0.001, vs the NC group; ^#P < 0.05, ^##P < 0.01, vs the ATP + PBS group). c Representative micrographs of IL-1β IF staining in RAW 264.7 cells. Scale bar: 50 μm. d Quantification of IL-1β IF staining in RAW 264.7 cells. e Detection of mtROS levels in RAW264.7 cells via flow cytometry. f Mitochondrial membrane potential was assessed by JC-1 staining and measured by flow cytometry. g Quantification of the ratio of JC-1 aggregates/JC-1 monomers. h Western blot analysis and quantification of NLRP3 and IL-1β levels. i Measurement of extracellular ATP level in RAW264.7 cells. j Western blot analysis and quantification of P-creb and P-P65 levels. k Experimental scheme for detecting the regulatory effect of A2aR inhibitor (AB928) on purine signals in immune cells in vitro. l qRT‒PCR analysis of the expression of proinflammatory genes (IL-1β and MCP-1) in RAW264.7 cells m Western blot analysis and quantification of IL-1β, P-creb and P-P65 levels. The averages of n = 3 independent experiments are shown in (b–m). Statistical analysis was performed by one-way ANOVA with Tukey’s multiple comparison test in (b–m). The data are shown as the means ± SD. Source data are provided as a Source data file.

The potent anti-inflammatory mechanism of ND was also explored. We found that ND can also reduce the exATP levels in the culture medium of macrophages, and its ATP-catabolic capacity was comparable to free ENPP1/CD73 group (E&C, at equal concentrations), while NT alone failed to reduce exATP levels (Fig. 3i). In line with data of cell-free tests, these results verify that ND can sense and metabolize high levels of exATP around cells. High levels of exATP can induce P2X7 purinergic receptor activation and its associated cellular potassium efflux and Ca²⁺ influx, resulting in mitochondrial ROS (O₂^-) generation to trigger the NLRP3 inflammasome pathway in macrophages²⁵. mtROS can interact with other proinflammatory signaling pathways, such as NFκB, through IκBα phosphorylation²⁶. In this study, we found that exATP stimulation increased the levels of mitochondrial ROS and mitochondrial damage (indicated by the MMP), as well as the activation of the NLRP3 and NFκB (indicated by its phosphorylation and nuclear translocation) pathways in M1 macrophages, whereas these effects could be attenuated by ND treatment via exATP degradation (Fig. 3e–j, Supplementary Figs. 10, 11). Meanwhile, E&C group but not NT group could inhibit the phosphorylation of NFKB p65 (p-P65) while promote the phosphorylation of CREB (p-CREB), resulting in reduced NLRP3 and cytokien (IL-1β) expression in macrophages (Fig. 3h–j).

On the other hand, ADO can serve as an endogenous modulator of immune actions by binding to its receptors (e.g., A2aR, A2bR) expressed on immune cells, and the activation of such receptors can limit the production of proinflammatory cytokines (e.g., TNF-a) or promote the production of immunoregulatory cytokines (e.g., IL-10) in immune cells (e.g., macrophages)²². The binding of ADO to A2A receptors further activates the intracellular cAMP-protein kinase A (PKA)-transcription factor CREB pathways, thereby directly or indirectly interacting with NF-κB signaling²⁷. Thus, in addition to its direct exATP-lowering effect, the ability of ND to produce ADO may also contribute to its anti-inflammatory effect. Indeed, the inhibitory effect of ND on the expression of cytokines (e.g., IL-1β and TNF-α) was partially reversed by cotreatment with an adenosine receptor inhibitor (AB928, a dual antagonist of A2aR/A2bR) (Fig. 3k–i). The increase in intracellular CREB phosphorylation can reflect the downstream activation of ADO-A2A receptor axis. The macrophages of ND group showed higher levels of p-CREB than those of ATP group or ATP + NT group (Fig. 3j), while this effect was partially reversed by co-treatment with AB928 (Fig. 3m). These results indicate that the anti-inflammatory effect of ND is at least partly dependent on the ADO-A2A receptor-CREB pathways. Interestingly, NT also showed moderate ROS (H₂O₂, •OH) scavenging role in a solution system and cells (Fig. 3e and Supplementary Fig. 12), which might be due to the high sensitivity of DNA molecules to various ROS species^28,29, and this effect may contribute to its beneficial role to some extent. Together, our results indicate that ND can restore immune cell homeostasis by switching exATP-mediated proinflammatory pathways to ADO-mediated immunoregulatory pathways.

Monocyte surface display enables the active delivery of NDs to inflamed sites

Although the ATP-sensing and anti-inflammatory effects of NDs have been verified in vitro, the in vivo therapeutic potency of NDs can be largely limited because of their nonspecific distribution to inflamed sites^14,15. To resolve this issue, it is necessary to develop a delivery system that can actively deliver NDs to inflamed sites in vivo. Monocytes are a major type of innate immune cells that are abundant in blood system. Monocytes have the natural ability to response different forms of tissue injury and infections, which can efficiently cross the inflamed endothelium and then enter the inflamed tissues to exert biological effect³⁰. The unique effects of monocytes, including injured tissue-homing, biological barrier cross and deep tissue penetrating, enabling them as ideal drug delivery system for inflammatory disorders. Despite other innate immune cells, such as neutrophils, also have similar inflamed tissue-homing ability³¹, but neutrophils have shorter half-life and more rapid clearance in vivo³². Based on the unique properties, recent studies have developed materials-based carrier (backpack) to attach the therapeutic payloads on the surface of monocytes for cellular hitchhiking delivery, which has presents a promising method to achieve targeted drug delivery to inflamed tissues³⁰.

Therefore, we sought to develop an active delivery system that can actively target inflamed tissues via the surface display of NDs on monocytes. Recent studies have found that modifying DNA or DNA origami nanostructures with cholesterol renders DNA-lipid conjugates amphiphilic, and this method can effectively prevent the internalization of DNA origami structures when they are anchored on the cell membrane^33–35. Increasing the number of cholesterol molecules can significantly reduce the mobility of DNA origami structures on lipid bilayer and effectively prolong their retention time on the membrane surface³⁶. To this end, single-stranded DNA (ssDNA) was modified with cholesterol (Chol-DNA) and inserted into the monocyte membrane through hydrophobic interactions (Supplementary Fig. 13). ND was subsequently anchored on the monocyte surface through hybridization of its extended strands with Chol-DNA after coincubation at 37 °C for 1 h (Fig. 4a). Fluorescence imaging analysis revealed that the fluorescence of Chol-DNA (in magenta) and ND (in green) overlapped well on the monocyte surface after coincubation (Fig. 4b and Supplementary Fig. 14). The surface modification efficacy of ND on monocytes (ND@Monos) could be optimized by tuning the ND concentration, and over 96% of the ND-positive monocytes could be obtained via the use of 10 nM ND (Fig. 4c), suggesting highly efficient surface display of NDs on monocytes. In addition, the average number of ND on each monocyte was calculated via the Cy5 fluorescence standard curve (Supplementary Fig. 15a). The average number of ND per monocyte was about (9.10 ± 0.12) × 10⁷ when adding 10 nM ND into 1 × 10⁵ cell/mL monocytes (Supplementary Fig. 15b). These results collectively indicate that ND can be a high-efficient loaded onto monocytes.

Fig. 4. Surface display of NDs on monocytes for targeting inflammatory tissue.

a Schematic illustration of the ND anchoring on the monocyte surface. One end (cholesterol) of amphipathic Chol-DNA can insert into the monocyte membrane through hydrophobic interactions, whereas the other end (DNA) is exposed to the monocyte surface and serves as the anchor site for ND loading. b Fluorescence colocalization imaging of ND@Monos. The Chol-DNA was modified with Cy7 (in magenta)), and the ND was labeled with Cy3 (in green), which was the enlarge of Supplementary Fig. S14. Scale bar: 25 µm. c Optimization of ND-positive monocytes by adjusting the ND concentration (0, 2.5, 5, or 10 nM). d The unlocking process of the ND on the monocyte surface in response to ATP (1 mM). Scale bars: 10 µm. e The biodistribution of ND@Monos in the kidney of KI mice or the lung of LI mice was evaluated by ex vivo organ imaging. NDs were labeled with Cy5.5, The biodistribution in other organ was presented in the Supplementary Fig. S22 and S23, NC group indicates normal mice, and PBS indicates mice treated with vehicle. The averages of n = 3 independent experiments are shown in (b–d). n = 4 mice for each group in (e). Statistical analysis were performed with One-way ANOVA with Tukey’s multiple comparison test (c) or unpaired two-tailed Student’s t-test (e). Data are presented as mean ± SD. Source data are provided as a Source Data file.

The stability of ND on the monocyte surface under in vivo-mimic solution conditions was also assayed. The ND was stabilized by coating with PEG-oligolysine and exhibited good stability within 24 h when incubated with 10% FBS. However, more than 40% ND-positive monocytes were observed when ND@Monos were exposed to RPMI 1640 medium (containing 10% FBS) at 37 °C for 12 h (Supplementary Fig. 16). Because the detailed components of serum are complicated, which contains various proteins, enzymes, cytokine, metabolites, lipids, and many other undefined substrates, and these complex factors might affect the stability of ND on monocytes via direct (on DNA) or indirect effects (on cells). Although we can not figure out the exact underlying mechanism right now, one of the possible reasons may be the bare double-stranded DNA (dsDNA) linker. Enzymes may cleave the bare dsDNA, causing the NDs to detach from the monocyte membranes. Thus, coating Chol-dsDNA linkers with PEG-oligolysine may be a possible solution to improve the stability of ND@Monos.

The subcellular distribution of ND in monocytes was assessed via fluorescence colocalization imaging. At different time points after the anchoring of Cy5-labled ND (Cy5-ND) onto the monocytes, the ND anchored monocytes (ND@Monos) were stained with both cell membrane dye (DiO) and intracellular endosome-lysosome dye (LysoTracker). As shown in Supplementary Fig. 17, the fluorescence of ND and DiO overlapped well even at 8 h after anchoring, while the fluorescence colocalization between ND and LysoTracker was much less. These results indicate that the majority of ND remain on the membrane of monocytes, and its internalization by monocytes via endocytosis is minor. However, owing to the inherent complexity of tissue microenvironments in vivo, the anchored ND might undergo gradual degradation or detachment from monocyte membranes after ND@Monos reaching targeted inflamed sites. It is still a challenge to fully avoid the endocytosis of NDs by host cells in vivo.

Next, we used a fluorescence-quencher pair (Cy5-BHQ3) assay to determine whether the ND anchored on the monocyte surface can exert unique properties (ATP-sensing and consequent unlocking processes). The fluorescence of the Cy5-BHQ3 pair on ND is quenched in the locked state of the ND due to the short distance between Cy5 and BHQ3, whereas it can increase in the unlocked state of the ND due to the separation of Cy5 and BHQ3 (Fig. 4d). The fluorescence of ND on momocytes gradually increased overtime (Supplementary Fig. 18), and its half-life (t½) of the unlocked NDs was ~14.86 min. Fluorescence imaging further revealed that the intensity of Cy5-BHQ3-labeled ND anchored on the monocyte surface exhibited a positive correlation with the concentration of ATP (Supplementary Fig. 19). These results demonstrate that ND can retain the ability to sense ATP and unlock after being modified on the monocyte surface. In addition, the ND surface modification process (incubation in buffer solution at 37 °C) is relatively mild and does not induce cell apoptosis (Supplementary Fig. 20a) or the biological properties of monocytes (e.g., chemotaxis). Monocytes can actively migrate into inflamed sites since they can sense the gradient of multiple key chemokines (e.g., MCP-1, CCL5, and CCL7) released from injured tissues via G protein-coupled receptors (such as CCR2 and CCR3)³⁷. Consistently, we observed a comparable chemotaxis ability between monocytes and ND@Monos in response to MCP-1 in a transwell system in vitro (Supplementary Fig. 20b–d).

The in vivo biodistribution of ND@Monos was further evaluated in diverse types of inflammatory injury models, including lung injury (LI) and kidney injury (KI) models. As shown in Supplementary Fig. 21, the injured tissues, including the lung tissues of the LI mice and the kidney tissues of the KI mice, expressed more chemokines (e.g., MCP-1) than did those of the normal group did, suggesting that ND@Monos might actively migrate (home) to inflamed tissues via chemokine-receptor interactions. The ex vivo imaging results revealed signals of Cy5.5-labeled ND and ND@Monos in multiple organs (such as the liver, lung, kidney and spleen) of NC, LI or KI mice at 2 h or 24 h after intravenous (iv) injection, and there was no significant difference in normal kidney or lung tissues between the ND group and the ND@Monos group, whereas signals of the iv injected-labeled ND@Monos group were significantly greater than those of the ND alone group in both injured lung tissues of LI mice and injured kidney tissues of KI mice (Fig. 4e, Supplementary Figs. 22, 23). Additionally, some signals of ND@Monos were observed in other organs/tissues with an abundant vascular system or resident immune cells (e.g., liver, lung, and spleen). Together, these results indicate that the inherent homing ability of monocytes can aid in the active delivery of NDs to inflamed sites in vivo.

The dynamic change in biodistribution of ND@Monos in vivo was also assessed. As shown in Supplementary Fig. 24a, b, the fluorescence intensity in kidneys of ND group or ND@Monos group gradually reduced over time (from 1 h to 24 h), and the injury renal retention of ND@Monos group was significantly higher compared to ND group at each time point. Moreover, we also conducted a quantitative method to analyzed the ND distribution in injury renal tissue in vivo. At different time points after injection of dye tracer (Cy5.5)-labeled ND or ND@Monos, the tracer intensity per tissue amounts was determined. In line with the ex vivo imaging results, the renal signals of ND group or ND@Monos group gradually reduced over time post-injection, and ND@Monos group had higher renal retention compared to ND group (Supplementary Fig. 24c). Together, these results indicate the enrichment of ND@Monos at target inflamed sites in vivo.

Next, we sought to determine if the ND has unlocking ability in the injured tissues using a FRET-lock reporter method. 6 pairs of lock_ATP strands were modified with Cy3-Cy5 fluorophores and then used to lock ND, and then Cy3-Cy5 labeled ND@Monos were iv injected into mice with kidney injury. As shown in Supplementary Fig. 25, the fluorescence signals were mainly observed in the kidney and liver tissue sections, which is in line with our ex vivo imaging results. The kidney section in the injured group showed strong Cy3 fluorescence and weak Cy5 fluorescence, whereas the kidney in the sham group showed weak Cy3 fluorescence and strong Cy5 fluorescence, indicating that most of the ND could be unlocked in the injured kidney and kept in the locked state in the sham kidney. These results are in line with our design. In contrast, a strong Cy5 fluorescence could be observed while weak Cy3 signals were detected in the liver, suggesting that most of the ND could remain in the locked state in the liver, while a small part of the ND might be unlocked due to the unknown factors in complex tissue environments. Together, these results indicate that most of the ND can be ATP-specific unlocked in the injured tissue while retain the locked state in other organ/tissues.

ND@Monos therapy attenuates inflammatory injury in diverse disease models

Having investigated the anti-inflammatory effect of ND in vitro and the homing ability of ND@Monos in vivo, the therapeutic role of ND alone or ND@Monos was further assessed via two different types of tissue injury models, acute lung injury³⁸(ALI) and acute kidney injury (AKI)³⁹ as previously reported. ALI is a serious acute respiratory disease characterized by diffuse lung inflammation and edema; it occurs in ~10.4% of patients in the intensive care unit (ICU) and has high mortality rates⁴⁰. Notably, lung inflammation plays a central role in the pathology of ALI, which is characterized by the recruitment and activation of multiple types of immune cells (e.g., macrophages and neutrophils) and the subsequent release of excessive cytokines and chemokines⁴⁰, and increased exATP has been observed in the lung tissues of an ALI model induced by LPS⁴¹. In line with previous studies³⁸, the lung tissues of ALI mice presented obvious lung edema and thickened alveolar walls, increased levels of immune cell infiltration (e.g., Ly6G⁺ neutrophils), proinflammatory pathways (e.g., NLRP3), and gene expression of cytokines/chemokines (IL-1β, IL-6, TNF-α, and MCP-1) compared with those of normal mice (Fig. 5a–d). In contrast, the severity of inflammatory injury, such as the lung injury score, neutrophil infiltration and cytokine/chemokine expression, in ALI model mice was decreased by ND alone or ND@Monos treatment (iv injection) (Fig. 5a–d). More importantly, the ND@Monos group presented lower lung injury scores, Ly6G⁺ neutrophil infiltration and cytokine/chemokine (e.g., IL-1β, IL-6, and MCP-1) expression than the ND group did (Fig. 5a–d). These results suggest that ND@Monos have superior anti-inflammatory capacity than ND alone in ALI mice.

Fig. 5. ND@Monos attenuate inflammation in diverse disease models.

a Schematic illustration of the animal experiments performed after acute lung injury (ALI). b qRT‒PCR analysis of the IL-1β, IL-6, TNF-α and NLRP3 mRNA levels in lung tissues (n = 6 mice). c Representative micrographs of H&E-stained, LY6G IHC-stained and MCP-1 IF-stained lung tissue. Scale bars: 50 μm. d Quantification of histological injury scores, Ly6G⁺ cells and relative MCP-1 fluorescence in the lung (n = 6 mice). e Schematic illustration of the animal experiments performed after acute kidney injury (AKI). f Serum urea and CREA levels and qRT‒PCR analysis of the Bax and MCP-1 mRNA levels in kidney tissues (n = 6 mice). g Representative micrographs of H&E-stained, Kim-1 and IL-1β IF staining of kidney tissue. Scale bars: 50 μm. h Quantification of histological injury scores, Kim-1⁺ tubules and relative IL-1β fluorescence in the kidney (n = 6 mice). NC group indicates normal mice, and PBS indicates mice treated with vehicle. Data are presented as mean ± SD. Statistical analysis were performed by one-way ANOVA with Tukey’s multiple comparison test. Source data are provided as a Source Data file.

The therapeutic effect of ATP-nonresponsive ND (Nr-ATP) and ATP-responsive ND were also compared in ALI model (Supplementary Fig. 26). Compared with Nr-ND group, ND group showed significantly higher inhibitory effect on NLRP3 pathway, key cytokines (IL-1β, IL-6, TNF-α) and chemokines (e.g., MCP-1) expression in lung tissues of ALI mice. Consistently, pathological examination showed that ND group had lower levels of lung injury score compared to those of Nr-ND group. These results verify that ND have superior anti-inflammatory capacity than Nr-ND in vivo. Interestingly, we found that Nr-ND group showed some reduction in lung injury compared to ALI alone. In this study, we have observed that NT had moderate ROS scavenging role in vitro (Fig. 3e and Supplementary Fig. 12), which might be due to the high sensitivity of DNA molecules to various ROS species^28,29. Together, our results indicate that the anti-inflammatory role of ND mainly depends on its ATP sensing and catabolism, and contribution of the DNA itself is minimal.

Besides ALI, AKI is another common and severe syndrome characterized by a rapid decline in kidney function caused by various insults (e.g., trauma or toxic drugs), and its morbidity rate remains high (~21.6% per year); however, current clinical treatments for AKI are still unsatisfactory⁴². Renal inflammation, such as hyperactivation of immune cells and massive cytokine release, is recognized as a critical mediator of the pathology of AKI, and a toxic renal drug (cisplatin) is capable of inducing cell death and excessive exATP release⁴³. Cisplatin is mainly cleared by the kidney via glomerular filtration and tubular excretion, which leads to a higher concentration of this drug in the kidney than in other organs. The pathology of cisplatin-induced AKI is complex, and multiple key factors, such as DNA damage, cell necrosis, and inflammation, have been involved⁴³. The therapeutic effects of ND or ND@Monos were assessed in a mouse model of AKI induced by cisplatin. Compared with normal controls, AKI model mice presented renal function injury (increase in UREA and CREA), increased kidney injury scores and markers (Kim-1), proapoptotic signals (BAX), and cytokines/chemokines (e.g., IL-1β and MCP-1) (Fig. 5e–h). Similarly, ND or ND@Monos treatments (iv injection) could reduce renal injury (reduced levels of UREA and CREA) and inflammation in AKI mice (Fig. 5e–h), and the ND@Monos group further showed better renoprotective potency, as evidenced by lower levels of UREA, kidney injury score, Kim-1⁺ injured tubules, Bax mRNA levels, and cytokine/chemokine (IL-1β, MCP-1) expression than the ND group did (Fig. 5e–h). Collectively, our results suggest that compared with NDs, ND@Monos have greater anti-inflammatory effects on adverse types of tissue injury.

Notably, monocyte/macrophage are major types of innate immune cell that regulate renal injury and repair, and increase in proinflamamtory macrophages has been found in cisplatin-induced AKI⁴⁴. Because of the heterogenous nature of immune cells, recent studies have found different macrophage subsets in the kidneys post-injury and they may play distinct roles in different conditions⁴⁵. For example, a recent single-cell RNAseq study found that Ccl2/Ccr2^high macrophage subsets drive renal inflammation and tubular injury in cisplatin-induced AKI, whereas Cx3cr1⁺ macrophage subsets exhibit protective role in AKI regression⁴⁶. However, previous studies that use clodronate liposome to systemically deplete macrophages can result in controversial outcomes (unchanged or even delayed renal repair)^44,47. This effect might be due to clodronate liposomes can deplete both proinflamamtory macrophages and protective macrophages in the kidney. Together with our results, these findings suggest that, due to the macrophage heterogeneity, strategies that can specifically regulate macrophage function (e.g., ND), rather than bulk macrophage inhibition or depletion, may advance tissue injury repair.

It has been reported that the administration of an ectonucleotidase protein might also have potent anti-inflammatory effects in other types of tissue injury in vivo. For example, intraarticular injection of recombinant mouse CD73 protein decreased the expression of proinflammatory (M1) markers (iNOS, TNF-α and IL-1β) and increased the expression of the M2 marker Arg-1 in the tibiae of aseptic loosening (AL) model mice⁴⁸. However, the therapeutic potency of recombinant proteins is generally limited because of their short half-life, rapid clearance and unspecific tissue distribution in vivo. In this study, we also assessed the therapeutic effect of ENPP1 and CD73 proteins (E&C) relative to ND alone in ALI models. E&C treatments moderately reduced the lung injury score, immune cell (Ly6G⁺ neutrophil and F4/80⁺ macrophage) infiltration, and cytokine (IL-1β, IL-6, and TNF-α) expression in the lung tissues of ALI mice, and the overall severity of lung inflammatory injury was greater in the E&C group than in the ND group (Supplementary Fig. 27). Together, our results suggest that ND@Monos have an overall greater anti-inflammatory effect than ND or ectonucleotidases alone in vivo.

ND@Monos exerts tissue-protective effects by restoring metabolic and immune homeostasis in vivo

Next, the possible mechanism underlying the beneficial effects of ND@Monos was explored via RNA-seq of lung tissues (Fig. 6a). A PCA plot revealed a clear separation among the groups (NC, ALI, ALI + ND@Monos) (Supplementary Fig. 28a), indicating that they presented distinct gene expression patterns. Furthermore, volcano plots and heatmaps revealed altered gene expression profiles between the ALI group and the NC group (Supplementary Fig. 28b, c). In line with previous reports⁴⁹, the upregulated DEGs in the ALI group (vs. the NC group) were enriched in various inflammatory processes, such as leukocyte migration, cytokine production, cell chemotaxis, and the regulation of the inflammatory response (Supplementary Fig. 28d). The global changes in gene expression patterns among the three groups were analyzed via cluster analysis on the basis of total genes, and there were 6 different gene clusters between the groups (Fig. 6b). In cluster 1, some genes were upregulated in the ALI groups and ALI + ND@Monos groups, and GO enrichment analysis revealed that these genes were enriched mainly in purine metabolism, such as purine nucleotide metabolism, ATP/ADP metabolism and purine nucleotide biosynthesis (Fig. 6c), which again indicated a critical role of purinergic signaling disorders in lung inflammation⁵⁰. In cluster 3, many genes whose expression decreased in the ALI group (vs. the NC group) were rescued by ND@Monos treatment, and these genes were enriched mainly in cellular metabolic processes, such as energy (ATP) synthesis, mitochondrial electron transport, oxidative phosphorylation, and the tricarboxylic acid (TCA) cycle (Fig. 6d). Previous studies have indicated that mitochondrial dysfunction and energy depletion are key mediators of cell death and inflammation in various types of tissue injury, including lung diseases⁵¹. Thus, these results suggest that mitochondrial damage and metabolic disorders in the lung tissues of ALI mice can be reversed by ND@Monos treatment.

Fig. 6. NDs play an immunostimulatory role by mediating purinergic pathways.

a Schematic illustration of the RNA-seq and metabolomics analyses. b Gene cluster trend in different groups (n = 3 mice). c GO enrichment analysis of genes in cluster 1. d GO enrichment analysis of genes in cluster 3. e GO enrichment analysis of genes in cluster 4. GO enrichment analysis was performed using a one-sided hypergeo metric test. Terms with a p < 0.05 were considered significantly enriched. f Heatmap of the relative abundance of proinflammatory cytokines and purine metabolic genes in the three groups (n = 3 mice). g PCA score plot of different groups subjected to metabolic analysis. h Pathway enrichment analysis of the DEMs enriched in ALI mice treated with ND@Monos (vs. the ALI group). Significance of enrichment was determined by a hypergeometric test with an FDR-corrected p-value (q-value) threshold of 0.05. i, j Western blot analysis and quantification of IL-1β, P-creb and P-P65 levels (n = 5 mice). Data are presented as mean ± SD. NC group indicates normal mice, and PBS indicates mice treated with vehicle. Statistical analysis was performed by one-way ANOVA with Tukey’s multiple comparison test. Source data are provided as a Source Data file.

In contrast, in cluster 4, some abnormally elevated genes in the ALI group (vs. the NC group) were suppressed by ND@Monos treatment (Fig. 6e). These genes were enriched mainly in biological processes related to cytokine production, the inflammatory response, the immune response, and cell apoptosis (Fig. 6e). In addition, some proinflammatory cytokines, such as IL18, IDo1, Tlrs, Cd300ld, and Tnfrsf14, which were elevated in the ALI groups, were suppressed by ND@Monos treatment (Fig. 6f). Together with the in vivo findings, these results verified that ND@Monos treatment could reduce inflammatory injury and cell death in inflamed lung tissues. During tissue injury, cells can release ATP into the extracellular space through multiple pathways, such as exocytosis and connexin channels (also known as gap junction channels), thereby amplifying inflammatory injury¹⁹. We found that multiple genes involved in regulating connexin channel formation (e.g., Gja5, Gjc1, Gjb5, and P2ry1) were upregulated in the ALI group but further suppressed by ND@Monos treatment (Fig. 6f), suggesting that channel-mediated ATP release was inhibited by ND@Monos. Moreover, some metabolic enzyme-encoding genes involved in purine catabolism (Nme1, Alp1, and Acp3) and ADO receptors (e.g., Ada, Adora1, Adora2a, and Adora3) were downregulated in the ALI group (vs. the NC group), and the expression of these genes was rescued by ND@Monos treatment (Fig. 6f). Further metabolomics analyses revealed distinct metabolite profiles among the three groups (Fig. 6g and Supplementary Fig. 29). Compared with those in the ALI group, the significantly changed metabolites in the ALI + ND@Monos group were also enriched in metabolic pathways related to mitochondrial function (e.g., the TCA cycle and pyruvate metabolism), redox status (glutathione metabolism) and purine metabolism (Fig. 6h), which again indicated that ND@Monos could attenuate cellular metabolic disorders in inflamed lung tissues.

Moreover, the key downstream intracellular signaling pathways (e.g., NFκB, CREB) that mediate the beneficial effect of ND were verified in vivo. The protein expression levels of phosphorylated NFκB p65 and CREB, as well as the downstream pro-inflammatory cytokines (IL-1β) were detected in lung tissues of ALI model. As shown in the Fig. 6i, j, compared with ALI group, ND@Monos group had significantly lower protein levels of p-NFKB p65 and IL-1β while increased protein levels of p-CREB in the lung tissues, which was in line with the cell tests and omics analysis results. Collectively, these results indicate that ND@Monos exert anti-inflammatory in vivo via regulating the NFκB and CREB signaling pathways. Together, these results suggest that ND@Monos exerts a tissue protective effect by restoring metabolic and immune homeostasis in vivo.

Given the promising anti-inflammatory effects and mechanism of action of ND@Monos, the biosafety and immunogenicity of ND and ND@Monos were assessed via in vitro and in vivo tests, respectively. In vitro, ND treatment did not impair cell viability or increase LDH release in diverse types of tissue cells, including renal tubular epithelial cells (HK-2), lung epithelial cells (MLE12) and vascular endothelial cells (HUVECs) (Supplementary Fig. 30). In vivo, neither ND nor ND@Monos treatment affected liver function (ALT and AST) nor renal function (UREA and CREA) in normal mice on day 3 after iv injection (Supplementary Fig. 31a, b). In addition, there were no obvious histological lesions in major organs/tissues, including the heart, liver, spleen, lung, and kidney, of normal mice receiving ND or ND@Monos (Supplementary Fig. 31c). Moreover, the effects of ND or ND@Monos on the immune response and the release of key proinflammatory cytokines were assessed in normal mice. Compared with those in the NC group, the levels of cytokines (IL-1, IL-6 and TNF-α) in blood samples and multiple organ/tissue (including muscle, kidney, heart, liver, spleen and lung) samples were comparable between the ND and ND@Monos groups (Supplementary Fig. 32). Collectively, these results indicate the high biosafety, biocompatibility and low immunogenicity of ND and ND@Monos in vivo. In summary, DNA origami-based ATP-sensing NDs can regulate purinergic signaling, and ND@Monos can actively target inflamed tissue and thus restore immunometabolic homeostasis in vivo. This study highlights that tailoring the design of metabolite-sensing NDs is a promising strategy for controlling the homeostasis of cell metabolism and the immune response.

Discussion

Notably, purinergic signaling plays critical roles in controlling the homeostasis of cell metabolism and the immune response. However, it can be disrupted in inflammatory disorders because excessive exATP is released by injured or dying cells. Inspiring by the working mode of the natural metabolite sensing and signaling system (e.g., AMPK, Fig. 1a), we designed the DNA nanodevice to sense and metabolize high levels of exATP. The addressability of DNA origami enables the restoration of the spatial expression pattern and action model of the extracellular ectonucleotidase system by precisely manipulating ENPP1-CD73 pairs on the DNA origami surface in a controllable and site-specific manner. During initial attempts, we also tested the possibility of conjugating CD39 to ND, but assembly of the ssDNA-CD39 complex with the DNA origami unexpectedly compromised its structural stability, for reasons that remain unclear. Given that ENPP1 have similar ATP-catabolizing activities, the ENPP1-CD73 pair was constructed and used in this study. Next, the in vitro and in vivo data have demonstrated that CD73&ENPP1-loaded ND can effectively sense high levels of exATP and then convert them to ADO, thereby switching the high-exATP-mediated proinflammatory pathways (e.g., mtROS, NLRP3 and NF-κB) to ADO-mediated immunoregulatory pathways (CREB). These results were aligned with other reports that ENPP1 or CD73 overexpression or exogenous proteins can also reduce tissue inflammation in disease models⁵². Together with our results, these findings suggest that metabolite-sensing ND is a promising strategy to regulate local immune microenvironment.

Although these results are encouraging, many important questions are still not completely known and more studies are required before future translational study and clinical trial. For example, despite dramatic increase in exATP concentrations and its strong correlation with inflammation, the exact exATP concentrations in diverse forms of tissue injury remain elusive, as in site detection of ATP is a technical challenge. Also, the stability of the ND on monocytes in blood is still not satisfied and should be improved. In vivo, besides the injured tissues, some of the systemically administrated ND@Monos also distributed in certain organ/tissues with abundant vascular system and immune cells (e.g., liver). As the high concentrations of exATP may easily unlock the ND and the co-existing of other pro-inflammatory substrates in the injured tissues, future studies may consider integrating several sensors (locks) that can simultaneously respond to more types of pro-inflammatory substrates on ND, or develop logic-gated ND to improve its sensing accuracy. Owing to the inherent complexity of tissue environments, fully avoiding the endocytosis of NDs is difficult, and the responses of cells to extracellular stimulus (e.g., ND) is also complicated. It is possible that some of ND might be taken up by other cells out of inflamed tissue (e.g, liver), which might also affect the intracellular signals (e.g., ATP, ENPP1/CD73) via unknown crosstalk and/or feedback effects. Thus, more in-depth researches are needed to address these questions. Nevertheless, this study highlights that intelligent DNA nanodevices are robust tools for controlling endogenous metabolic and immune actions and might hold promise as advanced therapeutic strategies for various diseases.

Methods

Ethical Statement

All animal experiments were approved by the Animal Care and Use Committee of West China Hospital, Sichuan University (Permit No. 20240605004) and followed the guidelines of the National Institutes of Health (NIH). The mice were purchased from Byrness Weil Biotech Ltd. (Chengdu, China), fed standard chow and sterile water ad libitum and housed at a controlled temperature and humidity with a 12 h light/dark cycle.

The fluorescence kinetics of unlocking the lock_ATP duplex

The lock_ATP duplex (see ‘DNA oligo sequences of the lock_ATP duplex.xlsx’ in the Supplementary Data 1) The 5 µM ATP aptamer strand was mixed with its partially complementary strand at a molar ratio of 1:1 in 1×TAE-Mg²⁺ buffer solution (containing 40 mM Tris, 20 mM acetic acid, 2 mM EDTA, and 12.5 mM magnesium acetate, pH = 8.0). The mixture was then annealed from 95 °C to 25 °C (Supplementary Data 1) to form the lock_ATP duplex. To investigate the ATP-responsive process, the lock_ATP duplex was labeled with a pair of fluorescence quenchers (Cy5-BHQ3), where the ATP aptamer strand was modified with a Cy5 fluorophore and its partial complementary strand was modified with a BHQ3 quencher. The Cy5-BHQ3-labeled lock_ATP duplex (500 nM) was mixed with different ATP concentrations (0, 0.05, 0.5, 1, 2, or 3 mM), and the fluorescence intensity of Cy5 was monitored via a microplate reader (BioTek, SYNERGY H1) at 37 °C for 4 h.

Preparation and purification of ATP-responsive DNA origami nanotubes

A rectangular DNA origami nanosheet (NS) (see ‘DNA oligo sequences of the DNA origami.xlsx’ in the Supplementary Data 1) was prepared by annealing 10 nM of the M13 strand with 50 nM of staple strands in 1 × TAE-Mg²⁺ buffer solution by gradually reducing the temperature from 95 °C to 4 °C (Supplementary Data 1). Next, the NSs were purified via PEG precipitation as described previously⁵³. In brief, an equal volume of PEG buffer (containing 15% PEG8000, 5 mM Tris, 1 mM EDTA, and 505 mM NaCl) was added to the mixture. The resulting mixture was centrifuged at 10,000 × g for 15 min. After the supernatant was discarded, the pellet was resuspended in 1 × TAE-Mg²⁺ buffer and shaken at 40 °C and 400 rpm for 8 hours to obtain purified NSs. A NanoDrop instrument (Thermo Scientific) was used to determine the mass concentration (C_mass) of the obtained NSs. The final molar concentration (C_molar) of the NSs can be calculated through a formula of experience (C_molar = C_mass/4.478 nM). After that, 10 nM purified NSs were mixed with 100 nM lock_ATP duplex in 1 × TAE-Mg²⁺ buffer and annealed from 60 °C to 36 °C (Supplementary Data 1). Finally, the mixture was purified via PEG precipitation to obtain ATP-responsive DNA origami nanotubes (NTs). The final concentration of the NTs was also calculated via the formula.

Coating DNA origami-based structures with K₁₀-PEG_5K

To increase the stability of DNA origami-based structures under physical conditions, all the DNA origami-based structures (including NS, NT or ND) were stabilized by coating with K₁₀-PEG_5K oligolysine as described previously¹⁸. Briefly, K₁₀-PEG_5k (mPEG_5K-b-PLKC₁₀, 5,800–7,500 Da) was purchased from Alamanda polymers. The DNA origami-based structures (20 nM, 10 µL) were mixed with 1 mL of K₁₀-PEG_5k at an N:P ratio of 0.5:1 (the ratio of nitrogen to amines:phosphates in DNA) at room temperature for 30 min.

The fluorescence kinetics of the ATP-responsive NT

The ATP-responsive NT was locked by the Cy5-BHQ3-modified lock_ATP strand (Supplementary Data 1) and purified via PEG buffer. Then, 20 nM ATP-responsive NT (Cy5-BHQ3) was mixed with different ATP concentrations (0, 0.1, 1, 2, or 6 mM), and the fluorescence intensity of Cy5 was monitored via a microplate reader (BioTek, SYNERGY H1) at 37 °C for 4 h.

To evaluate the selectivity and specificity of the NT, 20 nM ATP-responsive NT (Cy5-BHQ3) was mixed with ATP (3 mM), GTP (3 mM), CTP (3 mM), and UTP (3 mM), and the fluorescence intensity of Cy5 was monitored via a microplate reader (BioTek, SYNERGY H1) at 37 °C for 110 min.

Modification of ENPP1 and CD73 with single-stranded DNA

ENPP1 powder and CD73 powder were dissolved in 1 × PBS. A 50-fold excess of 3-(2-pyridyldithio) propionic acid N-hydroxysuccinimide ester (SPDP) was mixed with 4 μM ENPP-1 in 1 × PBS buffer (pH 8.5) and shaken at 25 °C and 200 rpm for 2 hours. After excess SPDP was removed via Zeba spin desalting columns (7 K MWCO, 0.5 mL), 4 μM SPDP-ENPP-1 was conjugated with a 3-fold excess of thiol-modified DNA strands (SH-ssDNA1) in 1 × PBS buffer (pH 7.4) and shaken at 25 °C and 400 rpm for 8 h. Excess SH-ssDNA strands were removed through ultrafiltration (3000 × g, 10 min) with 30 kDa cutoff filters (Amicon) three times in 1 × PBS (pH = 7.4). The conjugation of CD73 with SH-ssDNA2 was performed via the same process.

The assembly of ENPP-1 and CD73 with NSs

The complementary DNA strands (Supplementary Data 1) of ssDNA1 or ssDNA2 were extended from the designed site of the NS (capture-NS). A 10-fold excess of ssDNA1-ENPP-1 and ssDNA2-CD73 was mixed with 15 nM capture-NS in 1 × TAE-Mg²⁺ buffer at 37 °C for 1 h.

AFM imaging

Briefly, 5 µL of the sample (2 nM) was deposited onto the mica surface and allowed to adsorb for 5 min. Prior to scanning in tapping mode, 50 µL of 1 × TAE-Ma²⁺ buffer was added. Imaging was performed via a J scanner on a multimode Nanoscope IIIa AFM (Veeco/Digital Instruments) with a silicon nitride cantilever featuring a sharp pyramidal tip (OMCL-TR400PSA, Olympus).

LC-M/MS assay of ATP conversion to ADO by ND

To assess ND enzymatic activity, 50 μl of ND (30 μg/ml CD73 or 30 μg/ml ENPP1) was transferred to 150 μl of ATP solution (0.6 mM) and incubated for 1 h at 37 °C. To extract purines, 100 µl of 8% HClO₄ and 42.5 μL of H₂O were added to 157.5 μL of solution, vortexed for 30 sec and centrifuged at 12000 × g for 20 min at 4 °C. Then, 47.3 μl of K₂CO₃ (0.5 mM) was added to 250 µl of the supernatant, which was subsequently centrifuged at 14,000 × g for 30 min at 4 °C. The supernatant was subjected to a SCIEX ExionLC UHPLC system coupled with a Triple Quad 5500+ mass spectrometer (AB Sciex, Framingham, MA). Chromatographic separation was achieved on an ACQUITY UPLC BEH amide column (100 × 2.1 mm, 1.7 μm; Waters, Milford, MA, USA) maintained at 40 °C. A gradient elution program was conducted with buffer A (100% H₂O, 10 mM ammonium acetate, and 5 μM medronic acid) and buffer B (100% H₂O, 10 mM ammonium acetate, and 5 μM medronic acid), which were delivered at a flow rate of 0.3 mL/min. The gradient program was as follows: 0–1.5 min, 90% B, 1.5–3 min, linear decrease to 70% B, 3–5 min, linear decrease to 40% B, 5–9 min hold at 40% B; 9–9.1 min, increase to 90% B, 9.1–14 min, hold at 90% B. The mass spectra were acquired via electrospray ionization in negative-ion mode. The MS parameters were configured as follows: curtain gas (CUR): 30.0, collision gas (CAD): 8.0, ion spray voltage (IS): -4000 V. The temperature was set to 550.0 °C, ion source gas 1 (GS1): 40.0, and ion source gas 2 (GS2): 50.0. The samples were analyzed by multiple reaction monitoring (MRM) of the transition m/z 506.2 > 159.1 for ATP, 346.2 > 211.1 for AMP, and 426.2 > 159.1 for ADP. The ATP, AMP, and ADP collision energies were −35.06, −23.18, and −34.08, respectively. Data acquisition was performed via Analyst 1.7.2™ software, data processing was conducted via SCIEX OS (version 2.1.6.59781) software (AB Sciex, Framingham, MA), and the concentrations of ATP, AMP, and ADO in the sample were obtained in reference to a standard curve for adenosine.

Quantification of the ATP concentration

The ATP concentration was measured with an ATP Assay Kit (Beyotime, Cat: S0026) according to the manufacturer’s instructions.

ATP hydrolysis reactions

45 nM of ENPP1 or ND (loaded with equal ENPP1) was incubated with different concerntration of ATP (0.82 mM, 1.02 mM, 1.43 mM, 1.58 mM, 1.7 mM) in 1 × PBS (contains 12.5 mM Mg²⁺) buffer at 37 °C. The ATP concentration was measured 1 minute after the reaction began with an ATP Assay Kit (Beyotime, Cat: S0026) according to the manufacturer’s instructions. The initial reaction rate (v₀) was calculated with the equation: $v_0=\frac{d\left[S\right]}{dt}$. [S] represents ATP concentrtion. The data were fitted with the equation: $\frac{1}{v_0}=\frac{K_m}{v_{\max }}*\frac{1}{\left[S\right]}+\frac{1}{v_{\max }}$. v_max: maximum velocity, K_m: Michaelis constant, K_cat: catalytic constant.

Cell culture and treatments

The mouse macrophage line (RAW 264.7, AW-CM0088) and human monocyte cell line (THP-1, AW-CH0359) were purchased from AnWei-sci Biotechnology (Shanghai, China) and cultured in RPMI 1640 medium (Gibco, CA, USA) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Excel Biological Technology Co., Ltd., Shanghai, China), penicillin (50 U/mL; Beyotime Biotechnology, Shanghai, China) and streptomycin (50 μg/mL; Beyotime Biotechnology) in a humidified atmosphere at 37 °C with 5% CO₂. An ATP-dependent proinflammatory phenotype was induced as previously reported with slight modifications²⁴. Briefly, RAW 264.7 cells were treated with LPS (40 ng/ml, Sigma-Aldrich) for 4 h, followed by stimulation with ATP (5 mM, MedChemExpress, MCE, New Jersey, USA) for 15 min.

Anchor of ND on monocyte surface

2 × 10⁶ cells THP-1 suspended in 100 μL of serum-free medium, followed by mixing with 5 μL of 100 μM Chol-DNA (Supplementary Data 1). Subsequently, the mixture was incubated on a thermostatic oscillator at 37 °C, with shaking at a speed of 400 rpm for 10 min. After incubation, excess Chol-DNA was removed via two successive centrifugation steps, each performed at 1300 g for 3 min. The Chol-DNA-modified THP-1 cells were resuspended in 60 μL of serum-free medium. 2 × 10⁵ Chol-DNA-modified THP-1 cells were incubated with 2.5 nM ND with 10 extended ssDNAs at 37 °C for 1 hour. After incubation, excess ND was removed via one time centrifugation (1300 g, 3 min).

The fluorescence colocalization of ND@Monos with different dyes

The ND was labeled with Cy5 and the Cy5-ND@Monos was obtained via above method. 1 × 10⁵ Cy5-ND@Monos was incubated with DiO (5 μM, Beyotime) and LysoTracker (200 nM, Beyotime). After that, the Cy5-ND@Monos was subjected for fluorescence imaging using confocal microscopy (Stellaris STED, Leica, Wetzlar, Germany).

RNA extraction and real-time PCR

Total RNA was extracted from cell or tissue samples via TRIzol reagent (Thermo Fisher Scientific, Waltham, MA, USA) and reverse-transcribed into cDNA via an iScript cDNA synthesis kit (Vazyme Biotech, Nanjing, China). Real-time polymerase chain reaction (PCR) was performed on a CFX96 real-time PCR detection system (Bio-Rad, Hercules, CA, USA) with SYBR Green (Vazyme Biotech). The primers used in this study are listed in Supplementary Data 1. The PCR data were analyzed via Bio-Rad CFX Manager software, and the relative change in mRNA was calculated via the delta‒delta Ct method with Rps18 as a reference gene.

Immunofluorescence (IF) staining

The slides or tissue sections were fixed with 4% paraformaldehyde in PBS for 10 min at room temperature and then permeabilized with 0.03% Triton X-100 (Sigma‒Aldrich) for 10 min. After blocking with 1% bovine serum albumin (BSA, BioFroxx, Einhausen, Germany) for 1 h, the cells or tissue sections were incubated with rabbit anti-IL-1β (A1112, ABclonal Wuhan, China), goat anti-Kim-1 (R&D Systems, AF1817, Minneapolis, MN, USA), and rabbit anti-MCP-1 (10137S, Cell Signaling Technology) antibodies overnight at 4 °C. After being washed with PBS, the samples were incubated with the corresponding goat anti-rabbit (Alexa Fluor® 647, ab150079, Abcam), donkey anti-goat (Alexa Fluor® 647, ab150131, Abcam) or donkey anti-rabbit (DyLight® 550, ab96892, Abcam) secondary antibodies at 37 °C for 1 h. The nuclei were visualized by staining with DAPI (Sigma‒Aldrich, St. Louis, MO, USA). A dragonfly confocal microscope (Dragonfly 200, Andor, Andor Technology, Belfast, Northern Ireland, UK) was used to observe the stained samples.

Western blotting

The cell or tissue samples were lysed in radioimmunoprecipitation assay (RIPA) buffer (Beyotime Biotechnology, Shanghai, China) containing protease and phosphatase inhibitors. The protein concentration of each sample was quantified via a BCA protein assay kit (CWBIO, Beijing, China). Equal amounts of proteins were electrophoresed on 12% sodium dodecyl sulfate‒polyacrylamide gels (SDS‒PAGE), followed by protein transfer from the gel to polyvinylidene difluoride membranes (PVDF, Merck Millipore, Billerica, MA, USA). The PVDF membrane was blocked in 5% nonfat milk and incubated with primary rabbit anti-IL-1β (A1112, ABclonal, Boston, MA, USA), rabbit anti-NLRP3 (15101, Cell Signaling Technology), rabbit anti-NFκB (10745-1-AP, Proteintech, Rosemont, IL, USA), rabbit anti-P-NFκB (3033 s, Cell Signaling Technology, CST, Beverly, MA, USA), rabbit anti CREB1 (A11989, ABclonal, Boston, MA, USA), rabbit anti Phospho-CREB1 (AP0333, ABclonal, Boston, MA, USA) and rabbit anti-β-actin (AC026, ABclonal) antibodies at 4 °C overnight. The PVDF membranes were washed and incubated with the corresponding horseradish peroxidase-conjugated secondary antibody (ZB2301; Zhongshanjinqiao Biotechnology, Beijing, China) at 37 °C for 1 h. Protein signals were detected with a chemiluminescence kit (Millipore). Quantification of the signal intensity of the bands was carried out via ImageJ software (NIH).

Mitochondrial membrane potential (Δψm) and mtROS assays

To measure the mtROS level and mitochondrial membrane potential (Δψm), the cells were incubated with MitoSOX Red (2.5 μM, Thermo Fisher Scientific, Sunnyvale, CA, USA) or JC-1 (5 nM, AAT Bioquest, Sunnyvale, CA, USA) for 15 min at 37 °C. After incubation, the mtROS level was assessed via confocal microscopy, and the Δψm was analyzed via flow cytometry (LSRFortessa, BD Biosciences, USA).

The chol-DNA inserts on monocytes

A total of 1 × 10⁶ cells were suspended in 100 μL of serum-free medium and mixed with 2.5 μL of 100 μM chol-DNA. After incubation with shaking on a thermostatic oscillator (400 rpm) for 10 min at 37 °C, excess chol-DNA was removed by two rounds of centrifugation (1300 × g, 3 min).

Fluorescence imaging

A 100 μL aliquot of the sample was placed onto a glass-bottom dish (D29-10-1.5-N, Cellvis) and allowed to incubate for 20 min. Following incubation, fluorescence images were captured via a fluorescence microscope (Leica Thunder Imager DMI8, Germany).

Mouse inflammatory disease models and treatments

Two different types of inflammatory disease models, ALI and AKI, were used to assess the in vivo therapeutic effects of ND or ND@Monos and were induced as previously reported with slight modifications^38,39,54. Briefly, the mouse model of ALI was induced by intraperitoneal (i.p.) injection of LPS (10 mg/kg, Sigma‒Aldrich). The mouse model of AKI was induced via the intraperitoneal injection of cisplatin (20 mg/kg, MedChemExpress), and renal injury degree was assessed at day 2 after modeling. The dose of cisplatin used in this study mainly induce renal injury while has rare effect on other major organs (e.g., heart, liver, spleen and lung) in mice.

Mouse unilateral ureteral obstruction (UUO) model

The mouse kidney injury model induced by UUO was performed as previously reported⁵⁵. Briefly, the mice were anesthetized with sodium pentobarbital, and UUO was performed via ligation of the left ureter with 7–0 silk through a left flank incision. The sham kidney was subjected to the same procedure without ureter ligation. Three days after surgery, the mice were subjected to a biodistribution assay.

In vivo biodistribution of NDs or ND@Monos

Cy5.5-labeled ND (10 nM in 100 μL of PBS, 1 mM Mg²⁺) or an equal amount of Cy5.5-labeled ND displayed on monocytes (ND@Monos, 10 nM ND in 100 μL of PBS, 1 mM Mg²⁺, 1 × 10⁴ monocytes) was intravenously (i.v.) injected into normal mice, ALI mice or UUO mice via the tail vein. At 2 h or 24 h after EV injection, the mice (n = 3 per group) were sacrificed by an overdose of anesthesia, and the major organs (heart, lung, liver, kidneys, and spleen) were collected and detected on a In Vivo Imaging Instruments (Kino, Spectral Instruments Imaging, Tucson, Arizona USA). The fluorescence intensity of the NDs in each organ was quantified via Analyze 12.0 software (PerkinElmer).

Mouse treatment with ND or ND@Monos

For ALI mouse treatment, male BALB/c mice (20–25 g) were randomly divided into 5 groups (n = 6): control, ALI, ALI + CD73 + ENPP1, ALI + Nanodevice, and ALI + ND@Monos. For treatments, CD73 (50 nM) + ENPP1 (50 nM), ND (10 nM ND, ENPP1: 50 nM, CD73: 50 nM) or ND@Monos (10 nM ND, ENPP1: 50 nM, CD73: 50 nM in 100 μL of PBS) were intravenously (i.v.) injected into the mice via the tail vein after LPS injection, and the mice in the ALI alone group received 100 μL of PBS. At 24 h after ALI, the mice were sacrificed, and serum and lung samples were collected.

For AKI mouse treatment, male C57BL/6 mice (20–25 g) were randomly divided into 4 groups (n = 6 mice/group): control, AKI, AKI + ND, and AKI + ND@Monos. For treatments, ND or ND@Monos (10 nM ND, ENPP1: 50 nM, CD73: 50 nM in 100 μL of PBS) were intravenously (i.v.) injected into the mice, and the control mice received 100 μL of PBS. On day 2 after AKI, the mice were euthanized for collection of serum and kidney samples.

FRET-lock reporter for imaging tissue sections

6 pairs of lock_ATP strands were modified with Cy3-Cy5 fluorophores and then used to lock ND. After obtaining Cy3-Cy5 labeled ND@Monos, 100 μL of 15 nM the ND@Monos were intravenously injected into UUO mice. The different major organ/tissues, including heart, liver, spleen, lung, and kidney, were collected from mice sacrificed at 6 h post-injection, and then were subjected to freezing sectioning. Different organ sections were imaged with Cy3-Cy5 FRET channel (Ex: 554 nm, Em 670 nm) and Cy3 channel (Ex: 554 nm, Em 570 nm), respectively.

Histological examination

Tissue samples fixed with 4% paraformaldehyde were embedded in paraffin and cut into sections (5 μm thick). The tissue sections were stained with hematoxylin and eosin (H&E). To analyze pathological alterations, images of the stained sections were captured with a light microscope (Zeiss, AX10 imager A2, 180 Oberkochen, Germany). Three sections per mouse and six fields of each section were used. The percentages of damaged tubules were calculated as follows: 0, 0–10%; 1, 10–25%; 2, 26–50%; 3, 51–75%; and 4, ≥75%³. The percentage of injured lung tissue was determined via pathological scoring under a microscope. The histological assessor was blinded to the group information. For immunohistochemistry (IHC) staining, tissue sections were fixed with 4% paraformaldehyde, incubated with the following primary antibodies: mouse anti-F4/80 (70076 s, CST) and rabbit anti-Ly6G (GB11229, Servicebio) at 4 °C overnight, incubated with the secondary antibody at 37 °C for 2 h, washed three times, and observed under a light microscope. Three sections per mouse and six fields from each section were observed and quantified. The immunohistochemistry results were analyzed via ImageJ software.

Biochemical measurements

The levels of biochemical parameters, including creatinine (CREA), BUN, ALT and AST, in the serum samples of the mice were analyzed with a Cobas 6000 biochemistry analyzer (Roche Diagnostics, Switzerland) with appropriate kits.

RNA-seq analysis of lung tissues

Total RNA from lung tissue was extracted via TRIzol and then treated with DNase I (Takara, Shiga, Japan) to deplete the genomic DNA according to the manufacturer’s instructions. Library quality was evaluated with a 2100 Bioanalyzer (Agilent, USA). An ND-2000 (NanoDrop Technologies, USA) was used to construct the sequencing library. The RNA sequencing transcriptome library was generated with a TruSeq^TM RNA sample preparation kit from Illumina (San Diego, CA). cDNA was synthesized via a SuperScript double-stranded cDNA synthesis kit (Invitrogen) with random hexamer primers (Illumina). Libraries were size-selected for cDNA target fragments of 300 bp on 2% low-range Ultra agarose followed by PCR amplification via Phusion DNA polymerase (New England Biolabs, Ipswich, MA, USA). The expression abundance of each gene was determined via TPM, and the differentially expressed genes (DEGs) were identified via DESeq2 on the basis of read counts (fold change [FC] > 1.2 and p-value < 0.05). The results of the bioinformatic analysis of the omics data, such as PCA, heatmaps, volcano plots, and GO and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment, were analyzed and visualized via an online platform (https://www.bioinformatics.com.cn).

Untargeted metabolomics analysis

LC‒MS/MS-based untargeted metabolomics analysis of lung tissue was performed on a Thermo UHPLC-Q Exactive HF-X system equipped with an ACQUITY HSS T3 column (100 mm × 2.1 mm i.d., 1.8 μm; Waters, USA) at Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). In brief, 50 mg of sample was added to a 2 mL centrifuge tube, and a 6 mm diameter grinding bead was added. A total of 400 μL of extraction mixture (methanol:water = 4:1 (v:v)) containing 0.02 mg/mL internal standard (L-2-chlorophenylalanine) was used for metabolite extraction. The samples were ground with a Wonbio-96c (Shanghai Wanbo Biotechnology Co., Ltd.) frozen tissue grinder for 6 min (−10 °C, 50 Hz), followed by low-temperature ultrasonic extraction for 30 min (5 °C, 40 kHz). The samples were left at −20 °C for 30 min and then centrifuged for 15 min (4 °C, 13,000 × g), after which the supernatant was transferred to an injection vial for LC‒MS/MS analysis. The pretreatment of the raw LC‒MS data was performed via Progenesis QI (Waters Corporation, Milford, USA) software, and a three-dimensional data matrix in CSV format was exported. The significantly changed metabolites (DEMs, FC > 1.2, and VIP > 1) were considered significant. GO analysis of the DEGs and significantly changed metabolites was performed via the online tool MetaboAnalyst (https://www.metaboanalyst.ca/MetaboAnalyst/).

Intracellular ROS assay

The intracellular ROS level was determined with a dihydroethidium (DHE) probe (Beyotime Biotechnology, Shanghai, China) per the manufacturer’s protocols. The treated cells were washed 3 times in PBS and incubated with the mixture of DHE at 37 °C for 30 min in the dark. Then, the red fluorescence was observed under a fluorescence microscope (Zeiss, Imager Z2), and the intracellular oxidative levels were measured via ImageJ.

Mitochondrial membrane potential (Δψm) assessment

The Δψm was evaluated by JC-1 (5 nM) for 30 min at 37 °C and then visualized, and images were acquired via fluorescence microscopy (Zeiss, Imager Z2). The Δψm was analyzed via ImageJ, and the values are expressed as the fold increase in magenta/green fluorescence compared with that in normal control (NC) cells.

Cell viability assay

Cell viability was determined with a CCK-8 assay kit (Dojindo, Kumamoto, Japan) according to the manufacturer’s protocol. Briefly, HK-2 (human proximal tubular epithelial cell line), EA.hy926L (human endothelial hybrid cell line) and MLE-12 (murine lung epithelial cell line) cells were seeded into 96-well plates at a density of 1 × 10⁴/well and treated with NT for 24 h. Then, CCK-8 solution (10 μl/well) was added to each well, and the plate was incubated at 37 °C for an additional 2 h. The absorbance was measured using a microplate reader (BioTek, Biotek Winooski, Vermont, USA) at a wavelength of 450 nm. The cell viability was normalized to that of normal control (NC) cells.

LDH release assay

Cellular toxicity was assayed by measuring the release of lactate dehydrogenase (LDH) with an LDH Assay Kit (Beyotime) according to the manufacturer’s instructions. Briefly, cells were seeded in 96-well plates at a density of 1 × 10⁴/well and treated with NT for 24 h. Then, 25 μL of the LDH leakage reagent was added to a well of a 96-well plate as the maximum enzyme activity control. After 1 h, the 96-well plate was centrifuged for 5 min at 400 × g, and then, 120 μL of the suspension was transferred to a new 96-well plate. After adding 60 μL of reaction mixture and incubating at 37 °C for 30 min, the absorbance of each well was determined at wavelengths of 490 nm and 620 nm via a microplate reader (BioTek Instruments).

Monocyte migration assay

The effect on monocyte (THP-1) migration was assayed via a transwell system. In brief, 24-well inserts (3 μm pore sizes, Corning Costar, Cambridge, MA, USA) were used. A total of 106 THP-1 cells or ND@THP-1 cells in 200 μl of serum-free RPMI 1640 were loaded into the upper chamber of the transwell insert. RPMI medium (600 μl) containing 10 ng/ml MCP-1 (ABClonal, Boston, MA, USA) was added to the lower chamber. The cells were then allowed to migrate for 8 h. The migrated cells in the lower compartment were fixed with 10% paraformaldehyde at room temperature for 15 min and then stained with 0.1% crystal violet for 30 min (Beyotime Institute of Biotechnology, China). Images of the stained cells were captured via light microscopy (Zeiss, AX10 imager A2, Oberkochen, Germany).

Cell apoptosis assay

Cell apoptosis was analyzed via an Annexin V/PI kit (BD Pharmingen, San Diego, CA, USA) following the manufacturer’s instructions. Briefly, after being washed with PBS, the cells were collected and stained with annexin V and propidium iodide (PI) for 15 min in the dark, and the stained cells were analyzed via flow cytometry (LSRFortessa, BD Biosciences, USA).

In Vivo biosafety and immunogenicity of NDs and ND@Monos in Mice

Male mice were randomly divided into three groups: the NC group (n = 6), the ND group (n = 6) and the ND@Monos group (n = 6): an intravenous injection of PBS solution or ND or ND@Monos. On day 3 after treatment, the mice were sacrificed, and the blood and organs were collected for further analysis. Cytokine (TNF-α and IL-6) levels in mouse serum samples were analyzed via commercial ELISA kits (Dakewe, Beijing, China) according to the manufacturer’s instructions. The serum levels of biochemical parameters (ALT, AST, CREA, and BUN) were also analyzed. Cytokine (IL-1β, IL-6, and TNF-α) mRNA levels and the histopathology of major organs (heart, lung, liver, spleen, and kidney) were evaluated via qRT‒PCR and H&E staining, respectively.

Statistics and reproducibility

All experiments requiring statistical analysis were performed at least three times with similar results. The numbers of samples were decided based on the previous publications⁵⁶. Data from animal and cell studies were collected in a randomized and blinded fashion, and no data were excluded during the statistical analysis. Data distribution was assumed to be normal, but this was not formally tested. All data are presented as mean ± standard deviation (SD). Statistical analyses were conducted using GraphPad Prism 8.0 software, with one-way analysis of variance (ANOVA) employed for multiple group comparisons and two-tailed Student’s t-test applied for pairwise comparisons. Details of the statistical methods used for each dataset are provided in the figure legends. A p-value < 0.05 was considered statistically significant.

Reporting summary

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

Author contributions

W.L. and S.L. contributed equally. W.L., S.L., C.Z., and J.L. designed the study. W.L., S.L., X.Z., K.L., and F.L. performed the experiments. W.L., S.L., H.C., R.L., J.X., Y.M., and H.Y. analyzed the data. W.L., S.L., C.Z., and J.L. wrote and revised the manuscript.

Peer review

Peer review information

Nature Communications thanks Chunhai Fan, Simon Robson, who co-reviewed with Justin Lunderberg and Yihang Qi, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data from this study are fully available within the article, Supplementary Information or Source Data file. Raw data of the transcriptome can be accessed from the Genome Sequence Archive accession number CRA028070. Raw data of metabolomics in this study have been deposited in the OMIX, China National Center for Bioinformation / Beijing Institute of Genomics, Chinese Academy of Sciences and can be accessed via accession code OMIX014329. Any additional requests for information can be directed to, and will be fulfilled by, the corresponding authors. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68879-2.