IntroductionIonic liquids (ILs) are salts composed of bulky asymmetric cations and counterion anions, with melting temperatures typically below 100 °C. In addition, they show rich structural diversity through the combinations of various cations and anions1,2,3. Due to their various outstanding physicochemical properties, such as negligible vapor pressure4,5, inherent charge6,7, good stability and solubility8,9, and high structure tunability1,10, previous generations of ILs have diverse applications in synthesis11,12, catalysis13,14, electrochemistry7,15, etc16,17. Moreover, scientists have also realized the potential of ILs to promote drug development and have conducted a few pilot studies18,19,20,21. For example, choline and geranic acid-based ILs have been explored as base ingredient of tumor ablation agents to improve the killing effects of doxorubicin22.Despite their promise, the biomedical applications of ILs are still largely uncharted. This is mainly due to the current lack of a systematic and in-depth understanding of the fundamental rules that govern the biosafety profiles of ILs, particularly concerning their biocompatibility and toxicity23. Current studies have predominantly explored a specific subset of ILs types (such as choline-based ILs), providing preliminary insights into their potential role in the interaction on lipid bilayer models24,25, but highlighting the need for an extended selection range with the assist of cohesive knowledge. Furthermore, simulation data have shown that IL nanoaggregates form within the solvent6,26. However, previous biological impacts have focused solely on individual IL molecules, disregarding the potential role of IL nanoaggregates in the functionality. In addition, knowledge regarding as-yet-unknown characteristics (e.g., in vivo distribution, cellular internalization, and subcellular interaction) is eagerly awaited to stimulate ideas for the next-generation of ILs. Therefore, there is an urgent need to comprehensively elucidate the biocompatibility and toxicity rules and the underlying “black box” mechanisms.In this work, we initially design a modular IL library to capture their structural effects on cell viability, which contains diverse combinations (61 types) of their structural modules (cationic side chain, cationic head, and anion). Thereafter, we demonstrate that the viabilities of cell lines, cell spheroids, and patient-derived organoids all decrease as the cationic alkyl chain lengthens in vitro. To investigate the mechanisms at the cellular level, we work with two representative IL types, one with short cationic alkyl chains (scILs) and the other with long cationic alkyl chains (lcILs). Through a combination of simulations and experiments, it is revealed that ILs interact with cells in their nanoaggregate form rather than as individual molecules, and the divergent intracellular trafficking patterns of scILs and lcILs result in contrasting biological effects. The distinct biological issue regarding scIL and lcILs has also been verified in murine and canine models, and a better tolerance is revealed for oral (p.o.) administration. In addition to the superior compliance and tolerance of p.o. route, enhanced bioavailability of the insoluble drug (megestrol acetate, a semi-synthetic progestin with a long history of use in treating breast cancer27) is demonstrated with scIL nanoaggregates as drug carriers, compared to the commercial oral tablet in dogs. Such a representative formulation case thus manifests the merit of ILs in the pharmaceutical engineering scenario. Overall, our study contributes to a systematic and deep understanding of IL nanoaggregates in terms of their structure-based biocompatibility and toxicity, paving the way for their safe and rational biomedical applications.ResultsEffects of structurally diverse ILs on cell viabilityGiven that the IL structure is highly tunable and previous cytotoxic studies focused on specific IL molecules, the reported case-by-case data might fail to provide a cohesive relationship between structure and biocompatibility/toxicity. To solve this problem, we herein established a library that contained 61 types of ILs, which varied in a partially combinatorial fashion of three structural modules, including the cationic side chain (C), the cationic head (H), and the anion (A) (Fig. 1a and Supplementary Fig. 1a). Equipped with these ILs, we started our experiment with the viability assay in three cell lines, including mouse brain endothelial (bEnd.3) cells, mouse breast cancer (4T1) cells, and human hepatocellular carcinoma (HepG2) cells. Briefly, the cells were incubated with each IL at gradient concentrations (25, 100, 400, and 1600 μM) by a cell counting kit-8 (CCK-8). The viabilities at a typical 24 h exposure time were selected for determination, as a time-dependent response was found for the lower concentration (Supplementary Fig. 2).Fig. 1: Effects of structurally diverse ILs on the viabilities of cell lines, cell spheroids, and patient-derived organoids.a Representative example showing the typical structures of ILs. Note that each IL comprises the following modules: cationic side chain (C), cationic head (H), and anion (A). b–d Heatmap showing the impact of the indicated ILs on the viabilities of diverse cell lines (bEnd.3, 4T1, and HepG2) using the CCK-8 assay. The types of modules (C, H, and A) varied in (b, c) and (d) respectively, while at least one module remained constant. e Cell viability analysis of machine learning model towards iteration for prediction upon HepG2 cell treatments with ILs of different modules and concentrations. The curves in the x-z plane represent the correlation between the different modules (C, H, or A) of ILs and cell viability. The curves in the y-z plane represent the correlation between the concentration of ILs and cell viability. The all investigated structures of three modules in the IL library are shown in Supplementary Fig. 1a. f Representative bright-field/confocal laser scanning microscopy (CLSM) images (left) and viability quantification (right) of cell spheroids treated with PBS (Conn), 400 μM C3MIMCl (C3), or 400 µM C12MIMCl (C12) for 24 h. Green, calcein-AM (live cells); red, propidium iodide (PI, dead cells). The cell viability was presented as the proportion of calcein-positive cells to the sum of calcein-positive and PI-positive cells. g, h Representative bright-field/CLSM images (g) and viability quantification (h) of patient-derived liver cancer organoids treated with PBS or 400 μM ILs for 24 h. Live and dead cells were stained with calcein-AM and PI, respectively. Cell viability was determined by CellTiter-Glo (CTG) assay and normalized to the control organoids. The viability data of the heat map in (b–d) represent the mean of three biologically independent samples. Data in (f) and (h) represent the mean ± s.e.m., n = 6 biologically independent samples. Statistical significance was calculated via one-way ANOVA (f, h). Source data were provided as a Source Data file.Full size imageFor all cell lines tested, our library screening data showed that the cell viabilities were dependent on the IL dosage (Fig. 1b–d and Supplementary Fig. 1). Notably, the “C” module of ILs had a much greater impact on the cell viabilities than the “H” or “A” module (Fig. 1b–d and Supplementary Fig. 1b–e). Specifically, the cell viabilities decreased as the number of carbons in the alkyl chain increased, while altering the “H” and “A” modules caused little impact. Moreover, we conducted principal component analysis (PCA) on the cell viability database involving different module ILs, concentrations, and cell lines (Supplementary Fig. 3). The formation of distinguishable clusters did reveal the dominant role of the “C” module in ILs. In addition, we developed a machine learning model (i.e., a feed-forward neural network model) using the normalized and high-quality database sourced from our experimental data to predict cell viability (Fig. 1e and Supplementary Fig. 4). An extended IL library with more diverse combinations of IL modules (e.g., CxH2A2, x = 1–16; C1HxA3, x = 1–8; C5H3Ax, x = 1–15) was obtained. Multivariate analysis again unveiled an overall trend that C1-C4 ILs exhibited almost no cytotoxicity, while a dramatic increase in cytotoxicity was observed for ILs with the “C” module that contained 8 or more carbons.For further verification, we subsequently performed a live/dead assay with three-dimensional (3D) cell spheroids cultured from HepG2 cells. Briefly, C3MIMCl (1-propyl-3-methylimidazolium chloride) was examined as a representative scIL type, while C12MIMCl (1-dodecyl-3-methylimidazolium chloride) was the representative lcIL type, and an IL concentration of 400 μM was deployed for treating the cell spheroids. C12MIMCl was found obviously more toxic to the cell spheroids than C3MIMCl (Fig. 1f and Supplementary Fig. 5), which was consistent with the trend detected in the two-dimensional (2D) cell screening of the aforementioned cell lines and other primary cells (breast cancer cell in Supplementary Fig. 6). Specifically, minimal differences were observed in the spheroid morphology between the phosphate-buffered saline (PBS) control and C3MIMCl-exposed samples, with a cell viability of approximately 100%. In sharp contrast, the C12MIMCl-exposed spheroids appeared internally loose, and the spheroid boundary was blurred, along with a cell viability of less than 5%. Additional assays examining patient-derived liver cancer organoids revealed the same trend: exposure to C12MIMCl but not C3MIMCl caused obvious toxicity, with dead cells major in the cell population (Fig. 1g, h and Supplementary Fig. 7). Collectively, these results unveiled that the cytotoxicity of ILs was correlated with the length of the cationic alkyl chain rather than other tested structural modules.Evidence of IL nanoaggregates and their interactions with cellsBefore investigating the underlying mechanisms for the above distinct cytotoxicity of scILs and lcILs, we were interested in the morphology of ILs interacting with cells. Previous simulation studies have proposed that ILs exposed to an aqueous environment might form aggregates. To provide substantial experimental evidence, we characterized representative ILs in an aqueous solution using cryogenic transmission electron microscopy (Cryo-TEM). Indeed, we detected the existence of nanoaggregates in both the C3MIMCl and C12MIMCl samples (Fig. 2a). Analysis of 800 nanoaggregates from the Cryo-TEM images for each IL type showed that the frequency size of the C3MIMCl nanoaggregates was ~5 nm, while this value increased to ~12.5 nm for the C12MIMCl nanoaggregates. To gain deeper insight into nanoaggregate formation, we further conducted molecular dynamics (MD) simulations using a Martini coarse-grained (CG) force field for C3MIMCl and C12MIMCl in the aqueous phase. Consistent with the Cryo-TEM images, although the long cationic alkyl chain of C12MIMCl contributed to a large size, both C3MIMCl and C12MIMCl could form nanoaggregates (Fig. 2b and Supplementary Fig. 8). Moreover, the cationic alkyl chains (indicated by magenta) were embedded inside the cationic head (indicated by yellow) paired with anions (indicated by blue) in the aqueous phase, indicating that amphiphilicity was the driving force for nanoaggregate formation. In addition to the “C” module, we further investigated the influence of the other two modules in ILs on the morphology of nanoaggregates, in which the “H” module and “A” module were replaced with pyridinium head (PY) and alanine anion (Ala), respectively. Notably, neither the “H” nor the “A” module accounted for the morphology of the ILs (Fig. 2a, b).Fig. 2: Evidence of IL nanoaggregates and their interactions with cells.a Representative Cryo-TEM images (left) and size (diameter) quantification (right) of different IL nanoaggregates. b Snapshots of IL nanoaggregates in aqueous solution using MD simulations. Magenta, yellow, and blue indicate the “C”, “H”, and “A” modules, respectively. c Representative CLSM images (left) and mean fluorescence intensity (MFI) quantification (right) of C3MIMCl (C3) and C12MIMCl (C12) nanoaggregates after internalization by HepG2 cells. Green, cell membranes; blue, nuclei; red, Cy5 labeled ILs. d Representative TEM images of C3MIMCl (trapped in the vesicle, indicated by the yellow dashed line) or C12MIMCl (scattered in the cytosol) nanoaggregates after internalization by cells. e STED microscopy images captured at the indicated time points, showing the accumulation trend of Cy5-C12MIMCl nanoaggregates to mitochondria. Green, MitoBright LT; red, Cy5-C12MIMCl. f TEM images (top) and cartoon illustrations (bottom) of the mitochondrial changes in cells that were treated with C12MIMCl nanoaggregates (yellow arrows). The experiments in (e, f) were repeated three times independently with similar results. Size data in (a) are sourced from 800 particles for each IL type. Data in (c) represent the mean ± s.e.m., n = 3 biologically independent samples. Statistical significance was calculated via two-tailed Student’s t test (a, c). Source data were provided as a Source Data file.Full size imageAfter confirming the nanoaggregate form of ILs in the aqueous environment, we focused on the interactions between these nanoaggregates and cells. Considering that internalization was the prerequisite for exerting the cytotoxic impact, we initially investigated HepG2 cells exposed to Cy5-C3MIMCl (C3MIMCl labeled with Cy5) nanoaggregates or Cy5-C12MIMCl (C12MIMCl labeled with Cy5) nanoaggregates for 2 h. The amount of C3MIMCl and C12MIMCl internalized in the cells was approximately equivalent (Fig. 2c). In addition to excluding the size effect of IL nanoaggregates on their internalization, this finding also indicated that C3MIMCl and C12MIMCl nanoaggregates might exhibit other intracellular behavior, as these materials exerted distinct cytotoxicity. As shown in Fig. 2d, C3MIMCl nanoaggregates were detected within the membrane-wrapped vesicles, sharing a common fate with most exogenous substances internalized by cells. In sharp contrast, C12MIMCl nanoaggregates with positive surface charge facilitated transmembrane transport (Supplement Figs. 9 and 10), and scattered in the cytosol. Therefore, we envisioned that C12MIMCl nanoaggregates might further traffick in the cytosol and elicit more potent toxicity via unforeseen mechanism.Promoted by the above findings, we utilized stimulated emission depletion (STED) microscopy to monitor the intracellular trafficking of Cy5-C12MIMCl nanoaggregates (Fig. 2e). After 30 min of incubation, we found that the C12MIMCl nanoaggregates tended to accumulate around the mitochondria. After the incubation time was extended to 120 min, more C12MIMCl nanoaggregates colocalized with mitochondria, and the distribution pattern of mitochondria transformed to the abnormal state. For high resolution, we continued to use TEM to observe the interaction process between C12MIMCl nanoaggregates and mitochondria. Initially, a few nanoaggregates attached to the outer membrane surface of mitochondria (Fig. 2f). Following the accumulation of numerous nanoaggregates, notable defects were detected in the mitochondrial membranes. This was accompanied by a loss of cristae morphology, ultimately leading to the infiltration of nanoaggregates and substantial deconstruction of the mitochondria. Therefore, it was plausible that the damage of mitochondria caused by C12MIMCl nanoaggregates might highly correlate with the prominent cytotoxicity.Cytotoxic mechanism of mitophagy and apoptosis induced by C12MIMCl nanoaggregates but not C3MIMCl nanoaggregatesUpon discovering that the mitochondria were the intracellular targets for C12MIMCl nanoaggregates rather than C3MIMCl nanoaggregates, we investigated the underlying cytotoxic mechanism. To this end, we conducted RNA sequencing to identify differentially expressed genes in the transcriptomes of HepG2 cells upon treatment with PBS, C3MIMCl, or C12MIMCl (ILs at a concentration of 100 μM) (Fig. 3a). Compared with the PBS and C3MIMCl groups, the C12MIMCl group exhibited increased expression of mitophagy-related genes (such as Sqstm1, Map1lc3b, and Sesn2), apoptosis-related genes (such as Ddit3, Aen, and Bbc3), and oxidative stress-related genes (such as Oser1 and Atf4). Accordingly, gene set enrichment analysis (GSEA) revealed the significant enrichment of upregulated genes in both mitophagy and apoptosis pathways in the C12MIMCl group.Fig. 3: Cytotoxic mechanism regarding mitophagy and apoptosis induced by C12MIMCl nanoaggregates but not C3MIMCl nanoaggregates.a Heatmap (left) and GSEA (right) of RNA sequencing data from HepG2 cells treated with PBS (Conn), 100 μM C3MIMCl (C3), or 100 µM C12MIMCl (C12) for 2 h. b Representative CLSM images of cells incubated with PBS or ILs for 0.5 h by JC-1 staining to detect mitochondrial membrane potential. Red fluorescence indicates the appearance of JC-1 aggregates in normal mitochondria, whereas green fluorescence indicates the appearance of JC-1 monomers following mitochondrial membrane depolarization. Blue fluorescence indicates the nuclei. c Representative images (left) and quantification (right) of ROS in cells treated with PBS or ILs for 0.5 h. Green, ROS; blue, nuclei. d Representative images of cells incubated with PBS or ILs for 2 h and stained with Mitobright LT dye (green) and Mtphagy dye (red) to detect mitochondrial morphology and mitophagy, respectively. e MFI quantification of Mtphagy. f Immunoblotting quantification of the mitophagy indicators p62, PINK1, and LC3B from different treatment groups. g Immunoblotting quantification of the apoptotic indicator Caspase 9 from different treatment groups. Data in (f) and (g) were normalized to the levels of the control cells. h Flow cytometry analysis of cells after treatment with PBS or ILs for 2 h using the Annexin V/PI double staining assay. i Schematic illustration for the cytotoxic mechanism of C12MIMCl nanoaggregates. After cellular internalization, C12MIMCl nanoaggregates gathered to mitochondria and then induced them to depolarize. Subsequently, C12MIMCl nanoaggregates upregulated the expression of proteins involved in mitophagy and apoptosis and finally resulted in cell death. Data in (c, e, f) and (g) represent the mean ± s.e.m., n = 3 biologically independent samples. Statistical significance was calculated via one-way ANOVA (c, e–g). Source data were provided as a Source Data file.Full size imageIntracellular confinement around mitochondria and RNA sequencing data provided primary clues for the cytotoxic mechanism issue. Subsequently, we assessed a typical index of mitochondrial function using a JC-1 probe and found that C12MIMCl nanoaggregates resulted in an obvious reduction in mitochondrial membrane potential, while almost no reduction was detected upon exposure to C3MIMCl nanoaggregates compared with PBS (Fig. 3b and Supplementary Fig. 11a). Consistent with the mitochondrial disruption and increased expression of oxidative stress-related genes, the depolarization observed in the C12MIMCl group further led to the elevated production of reactive oxygen species (ROS), exhibiting approximately 2.5-fold higher levels than those in the PBS and C3MIMCl groups (Fig. 3c).The upregulated genes and GSEA data also prompted us to investigate the mitophagy and apoptosis issue, which could be considered as the consequence of the above mitochondrial depolarization and ROS production. On the one hand, C12MIMCl nanoaggregates caused a notable mitophagy phenomenon as indicated by the red fluorescence signal of Mtphagy dye, which was not observed in the PBS or C3MIMCl group (Fig. 3d, e). Immunoblotting analysis also confirmed that mitophagy indicators (p62, PINK1, and LC3B) were specifically upregulated by C12MIMCl nanoaggregate exposure (Fig. 3f and Supplementary Fig. 11b). On the other hand, Caspase 9 immunoblotting results suggested robust apoptosis occurred exclusively in the cells treated with C12MIMCl nanoaggregates (Fig. 3g and Supplementary Fig. 11b). Upon quantification by flow cytometry, the C12MIMCl group manifested the highest apoptotic rate of 81.4%, in comparison with the low rates (around 10%) in PBS and C3MIMCl groups (Fig. 3h and Supplementary Fig. 11c). Collectively, these results revealed a mechanistic explanation for the observed distinct cytotoxicity of scILs and lcILs: Although both C3MIMCl and C12MIMCl formed nanoaggregates and could be internalized by cells, the C12MIMCl nanoaggregates were able to target and disrupt mitochondria, leading to the depolarization of mitochondria and downstream responses (mitophagy and apoptosis) (Fig. 3i).Disruption mechanism of mitochondrial membranes induced by C12MIMCl nanoaggregatesAfter reviewing the whole process for cytotoxicity of C12MIMCl nanoaggregates, we proposed that the membrane disruption of mitochondria should be the initiating event. With this in mind, we were interested in how the C12MIMCl nanoaggregates interacted with mitochondrial membranes and intruded the mitochondria. Following MD simulations for IL nanoaggregates, we investigated interaction between the IL nanoaggregates and this unique cellular organelle with double membranes. The outer mitochondrial membrane (OMM) and inner mitochondrial membrane (IMM) with sophisticated lipid components and proportions28 were incorporated for a more accurate simulation (Fig. 4a). Considering the charged groups of ILs and the well-known mitochondrial membrane potential, we assumed a force was accountable for this issue. Upon this simulation, the C12MIMCl nanoaggregate was adsorbed on the OMM surface and inserted the OMM (Fig. 4b and Supplementary video 1). Subsequently, the C12MIMCl nanoaggregate stretched the lipid molecules of the OMM, penetrated the OMM to the intermembrane space, and further surmounted the energy barrier to cross the IMM, exhibiting a dual-barrier crossing capacity (Fig. 4b and Supplementary video 2).Fig. 4: Simulations of the C12MIMCl nanoaggregates that interacted with the mitochondrial membranes.a Schematic illustrations of the components of the models and driving force using MD simulations. The models of different lipid types and C12MIMCl (C12) nanoaggregate are shown in the left panel. The overall compositions of the OMM and IMM are specified in the right panel. Abbreviations are as follows: POPC, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine; POPE, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine; SAPE, 1-stearoyl-2-arachidonoyl-sn-glycero-3-phosphoethanolamine; CHOL, cholesterol; SAPI, 1-stearoyl-2-arachidonoyl-sn-glycero-3-phosphoinositol; CRDL, cardiolipin. b Snapshots of C12MIMCl nanoaggregate migrating directly across the mitochondrial membranes within 100 ns. c Quantitative analysis of the interaction energy between the C12MIMCl nanoaggregate and OMM/IMM. d Number density of molecules forming C12MIMCl nanoaggregate at the OMM (40 ns) and IMM (80 ns). The magnified views show the fluorescent green lipid molecules that were dragged into the C12MIMCl nanoaggregate. e Counts of C12MIMCl molecules sequestered in the OMM/IMM and lipid molecules that were dragged out from the OMM/IMM upon the different numbers of C12MIMCl nanoaggregates migrating across membranes. f Radial distribution functions for the outer leaflet liquid POPC of the OMM (40–45 ns) and IMM (80–85 ns). g Tail angles of the lipid molecules in the OMM (left, 40–45 ns) and IMM (right, 80–85 ns). h Counts of flip-flop lipid molecules in mitochondrial membranes. Data in (g) represent the mean ± s.e.m., n = 26 frames from 5 ns postmigration. Statistical significance was calculated via one-way ANOVA (g). Source data were provided as a Source Data file.Full size imageWe next calculated the interaction energy between the nanoaggregate and the membranes as the nanoaggregate migrated across the OMM and IMM (Fig. 4c). There were evident decreases in the interaction energy, particularly with a strong negative peak in the Lennard-Jones (LJ) potential in comparison to the Coulomb (Coul) potential, which indicated an important role of intermolecular force in this process. Moreover, a larger reduction in the LJ potential was observed when the C12MIMCl nanoaggregate migrated across the OMM. Thus, the aforementioned dual-barrier crossing event was attributed not only to the driving force derived from mitochondrial potential but also to the lipophilic interaction between the long alkyl chains of IL nanoaggregates and membrane lipids. Upon this event, it was noting that a reduction was found in the number density of the C12MIMCl nanoaggregate (Fig. 4d). Quantitatively, the number of C12MIMCl molecules residual in the OMM and IMM was seven and three, respectively (Fig. 4e and Supplementary Table 1). During this process, two lipid molecules from the OMM and one lipid molecule from the IMM were incorporated into the nanoaggregate. Meanwhile, two molecules from the OMM were sequestered in the IMM. These findings clearly supported that partial C12MIMCl molecules in the nanoaggregate had exchanged with the lipid molecules of mitochondrial membranes during transmembrane migration. The faster translocation observed, compared to previous studies29, could be attributed to the incorporation of a driving force to simulate the influence driven by the mitochondrial membrane potential, effectively compensating for the limitations of CG models under high-stress conditions. The simulations microscopically and mechanistically also validated prior macroscopic observations regarding the interaction between the lipid bilayer model and ILs25.In light of the TEM images depicting the presence of several C12MIMCl nanoaggregates surrounding a mitochondrion, we carried out simulations with an increased number of nanoaggregates (Supplementary Fig. 12), which might reveal a cumulative effect on the mitochondrial membranes. Compared with a single nanoaggregate, the migration of four C12MIMCl nanoaggregates caused a > 4-fold increase in the counts of residual C12MIMCl molecules (39 for OMM and 18 for IMM) and lipid molecules that were dragged out from the membranes (27 for OMM and 18 for IMM) (Fig. 4e and Supplementary Table 2). Hence, the detrimental impact of C12MIMCl nanoaggregates on mitochondrial membranes manifested not merely as a simple additive effect but also involved a synergistic augmentation of lipid destruction. This event could potentially result in an imbalance of lipid rearrangement during this exchange process. After analyzing radial distribution functions for the lipids, we further discovered that exposure to C12MIMCl nanoaggregates led to a tighter arrangement of lipid molecules, which was more pronounced in the presence of four nanoaggregates (Fig. 4f and Supplementary Fig. 13). Moreover, the lipid molecules also exhibited an increasing tail angle orientation and a higher flip-flop frequency as the number of nanoaggregates increased (Fig. 4g, h). The above abnormalities in lipid arrangement, tail angles, and flip-flop frequency collectively offered a microscopic explanation for the disruption in mitochondrial membranes observed upon interaction with C12MIMCl nanoaggregates.Effects of scILs and lcILs on mice upon diverse administration routesTo assess whether the distinct cytotoxicity of scILs and lcILs observed in vitro was also evident in vivo, C57BL/6 mice were grouped and administered C3MIMCl or C12MIMCl via three clinically relevant administration routes, after which we conducted a series of toxicology investigations (Fig. 5a). Upon oral administration, C12MIMCl was approximately 80-fold more toxic than C3MIMCl, as supported by the remarkably different minimum lethal doses. Specifically, four mice died at a 0.4 μmol/g C12MIMCl dose, while only one mouse died at a 20 μmol/g C3MIMCl dose (Fig. 5b and Supplementary Fig. 14). Cy5-C3MIMCl or Cy5-C12MIMCl was orally administered to mice, marking the first instance of IL distribution assessment through in vivo fluorescence imaging over the time course. C3MIMCl and C12MIMCl shared similar features, as a highly intense fluorescence signal was detected in the abdominal region and gradually disappeared within 24 h (Fig. 5c). For better visualization, we further resected the major mouse organs, revealing substantial accumulation of C3MIMCl and C12MIMCl in the stomachs and intestines, whereas other organs showed a very slight signal of these two ILs (Fig. 5d and Supplementary Fig. 15).Fig. 5: Assessments of IL biocompatibility and toxicity to mice upon three administration routes.a Experimental design for assessing the biocompatibility and toxicity of C57BL/6 mice through different routes of PBS (Conn), C3MIMCl (C3) or C12MIMCl (C12) administration. b Survival number of mice within 14 days after p.o. administration of C3MIMCl or C12MIMCl. c Representative in vivo fluorescence images of mice after p.o. administration of Cy5-C3MIMCl or Cy5-C12MIMCl at the indicated time points. d Representative ex vivo fluorescence images of the excised organs after 6 h p.o. administration of Cy5 labeled ILs. e Immunofluorescence images of histological sections after 24 h p.o. administration of ILs. Blue, DAPI (nuclei); magenta, LC3B (mitophagy); green, TUNEL (apoptosis). f Representative haematoxylin and eosin (H&E) staining images of organ sections (from excised stomachs and intestines) after 24 h p.o. administration of ILs. g Routine blood test for Mon% and Lym% (in white blood cells) after 24 h p.o. route of PBS or ILs. h Serum ALT, ALP, AST, BUN, and LDH levels of mice after 24 h p.o. route of PBS or ILs. The shaded gray area represents the ranges of reference data of healthy mice. Abbreviations are as follows: ALP, alkaline phosphatase; AST, aspartate aminotransferase; LDH, lactate dehydrogenase. i Survival number of mice within 14 days after i.m. (top) or i.v. (bottom) route of the indicated doses of C3MIMCl or C12MIMCl. j Representative ex vivo fluorescence images and MFI of the excised organs at 5 min after i.m. (top) or i.v. (bottom) route. k Immunofluorescence images of histological sections of mice after 24 h i.m. (top) or i.v. (bottom) route. Blue, DAPI (nuclei); magenta, LC3B (mitophagy); green, TUNEL (apoptosis). According to half of the minimum lethal doses, the doses for toxicity assessments upon p.o., i.m. and i.v. routes were 0.2 μmol/g, 0.015 μmol/g and 0.01 μmol/g, respectively. Data in (g, h) and (j) represent the mean ± s.e.m., n = 3 biologically independent mice per group. Statistical significance was compared via one-way ANOVA (g) or two-tailed Student’s t test (j). Source data were provided as a Source Data file.Full size imageRecalling the cytotoxic mechanism, we continued to perform immunofluorescence analysis to assess potential mitophagy (LC3B, magenta) and apoptosis (TUNEL, green) in the major organs after 24 h of treatments (Fig. 5e and Supplementary Fig. 16). As expected, C3MIMCl group exhibited no sign of mitophagy or apoptotic signal, whereas robust signals for both indicators were evident in the stomach and intestine of the C12MIMCl group. Correspondingly, shedding of gastric epithelial cells and damage to intestinal villi were only observed in the C12MIMCl group (Fig. 5f). Given that severe damage in the gastrointestinal tract might induce a systemic response, we herein assessed the abnormalities in the blood. There was no significant difference between the PBS and the C3MIMCl-exposed mice in blood cell analysis, whereas C12MIMCl caused significant anomalies in the percentage of monocytes (Mon%) and percentage of lymphocytes (Lym%) in white blood cells (Fig. 5g). In addition, the C12MIMCl-exposed mice displayed significantly elevated values of alanine aminotransferase (ALT) and blood urea nitrogen (BUN) (Fig. 5h), accompanied by the evident infiltration of the inflammatory cells (Supplementary Fig. 17) and increase of cytokine levels in serum and urine (Supplementary Fig. 18). Although C12MIMCl mainly accumulated in the stomach and intestine, the dysfunctions in other organs (e.g., liver and kidney) might be attributed to the systemic inflammation.Upon further examinations with intramuscular (i.m.) and intravenous (i.v.) routes, we revealed that the minimum lethal doses for both C3MIMCl and C12MIMCl decreased in the sequence of the p.o., i.m., and i.v. routes, and C3MIMCl again demonstrated much higher minimum lethal doses than C12MIMCl for the i.m. and i.v. routes (Fig. 5i and Supplementary Fig. 14). Specifically, the minimum lethal doses for the i.m. route of C3MIMCl and C12MIMCl were 2 μmol/g and 0.03 μmol/g, respectively, while these values for the i.v. route further decreased to 0.5 μmol/g and 0.02 μmol/g, respectively. Moreover, the distributions of C3MIMCl and C12MIMCl were similar when administered via i.m. and i.v. routes, with signals present in most examined organs (livers, lungs, kidneys, stomachs, intestines, and bladders) (Fig. 5j). Although such wide distributions via the i.m. and i.v. routes were distinct from those of the p.o. route, we detected mitophagy and apoptosis exclusively in the organs that C12MIMCl had accumulated in (Fig. 5j, k). Notably, there was a positive correlation between the fluorescent distribution signal and mitophagy/apoptotic signals (Supplementary Fig. 19), indicating that the amount of locally exposed C12MIMCl was associated with the level of potential tissue damage (Supplementary Figs. 20, 21). Moreover, toxicological investigations were performed with ILs in which the anions were replaced with Ala or acetic anion (Ac), and we observed that scILs (C3MIMAla and C3MIMAc) again exhibited superior biocompatibility compared with lcILs (C12MIMAla and C12MIMAc) across all tested administration routes (Supplementary Figs. 22, 23), emphasizing that the length of the cationic alkyl chain was dominant in the toxic effects of ILs in vivo.Effects of scILs and lcILs on canines upon oral administration routeTo advance our basic insights closer to clinical relevance, we extended our studies to beagles, a common nonrodent model used for toxicology studies of large animals. Considering the compliance and superior tolerance to ILs upon the p.o. route, we performed toxicity assessments of beagles after p.o. administration of PBS, C3MIMCl, or C12MIMCl and developed a representative case of ILs in the realm of drug delivery (Fig. 6a). One day before and after the administration, we conducted routine blood testing and recorded the physical and mental status of the dogs. Moreover, endoscopic imaging was employed one day after administration, and the tissue sections were analysed after dissection.Fig. 6: Assessments of IL biocompatibility and toxicity to canines upon p.o. administration.a Schematic illustration of the experimental design for p.o. administration of PBS (Conn), C3MIMCl (C3), or C12MIMCl (C12) to dogs (beagles) and corresponding analysis. b Table summarizing the occurrence of emesis, diarrhea, hematochezia, and listlessness of the dogs after the p.o. administration of PBS or ILs. c Relative serum Na+ and Cl− levels of dogs after 24 h of different treatments. Data were normalized to the levels of the untreated dogs. d H&E staining images of the stomachs and intestines of dogs receiving different treatments. e Routine blood test for the relative Mon% and Lym% of dogs after 24 h of different treatments. Data were normalized to the levels of the untreated dogs. f Urea and Scr levels of dogs after 24 h of different treatments. g Plasma concentration-time curves of insoluble drug (megestrol acetate) under two conditions: commercial tablet or C3MIMCl formulation following p.o. administration at 40 mg. h Pharmacokinetic analysis of systemic exposure (AUC0-inf) and peak concentration (Cmax) following p.o. administration of megestrol acetate. AUC0-inf is estimated by computing the integral of the concentration-time curve from time zero to infinity. Cmax is recorded as the maximum observed drug concentration. Data in (c, e–g) and (h) represent the mean ± s.e.m., n = 3 biologically independent dogs per group. Statistical significance was calculated via one-way ANOVA (c, e) or two-tailed Student’s t test (f, g, h). Source data were provided as a Source Data file.Full size imageWe observed no abnormalities in the physical or mental status of the dogs in the control and C3MIMCl groups. In sharp contrast, the C12MIMCl group exhibited symptoms of emesis and diarrhea shortly after oral administration. Additionally, the animals appeared hematochezia and listless, often assuming a prone position, during the observation period (Fig. 6b). Specifically, regarding the acute gastrointestinal symptoms after C12MIMCl administration, the first dog vomited (with a pink foam appearance) and passed a yellow watery stool within 1 h (Supplementary Table 3). The second dog vomited white foam at 10 min, 30 min, and 50–60 min (three times consecutively) and passed a yellow watery stool at 50–60 min. The third dog experienced vomiting three times consecutively within 10–20 min. The above emesis and diarrhea also caused electrolyte imbalance, which was supported by the abnormal serum Na+ and Cl− levels in the C12MIMCl group (Fig. 6c).Further endoscopic images showed that dogs in the C12MIMCl group displayed obvious gastrointestinal bleeding in the esophagus, cardia, gastric body, pylorus, and intestine, whereas these lesions were not observed in the control or C3MIMCl group (Supplementary Fig. 24). The histological analysis also revealed typical signs of gastric submucosa edema, along with vasodilatation, congestion, and hemorrhage in the lamina propria of the intestinal mucosa exclusively in the C12MIMCl group (Fig. 6d). The toxicity results in mice also prompted us to investigate the potential systemic inflammation and associated damage to other organs. There was a significant increase in the Mon% of the C12MIMCl group, while a decline was observed in the Lym% (Fig. 6e). Consequently, inflammatory cells infiltrated in the renal interstitium (Supplementary Fig. 25), with urea and serum creatinine (Scr) levels beyond the normal range (Fig. 6f). The results obtained from the dogs further reinforced that C12MIMCl with a long side chain exhibited pronounced toxicity, while C3MIMCl with a short side chain showed good biocompatibility.The aforementioned results indicated that scILs provided valuable insights into direct exploitation for drug delivery applications. Additionally, the oral route offered convenience and compliance. Building on these findings and the capacity of the hydrophobic core to accommodate insoluble drugs, we incorporated C3MIMCl as a nanocarrier for megestrol acetate. Despite expanding therapeutic applications (e.g., breast cancer and anorexia-cachexia syndrome30), the efficacy of megestrol acetate was still limited by its low bioavailability. The pharmacokinetic profiles following a single p.o. administration of 40 mg exhibited considerable differences in the concentration-time curves between the commercial tablet and C3MIMCl formulation (Fig. 6g). In comparison to the commercial tablet, both the peak concentration (Cmax) and the area under concentration-time curve from time zero to extrapolated infinity (AUC0-inf) for C3MIMCl group were enhanced (3.7-fold and 3.1-fold, respectively) (Fig. 6h).Despite the intricate excipient components in the commercial tablet, the enhanced absorption performance could be attributed to the substantially improved drug dispersity (smaller and more uniform size distribution), facilitated by C3MIMCl nanoaggregates (Supplementary Fig. 26). In addition, superior biocompatibility of C3MIMCl nanoaggregates over C12MIMCl nanoaggregates, along with their enhanced bioavailability relative to commercial tablet, strengthened their potential as carriers for insoluble drugs. Beyond this representative case, extensive prospects of ILs in other areas of pharmaceutical engineering are also awaited, upon the aforementioned spectrum of ILs-bio interactions.DiscussionThe possibility of ILs forming ionic aggregates has been proposed since 200531. While this concept shows potential, the characteristics of ILs, such as low contrast and striking difference before and after dehydration, pose challenges for traditional techniques/methods to obtain the pristine state of IL aggregates. Herein, we constructed a unique Cryo-TEM protocol that involved an optimum ice layer and was free of staining to ensure that ILs remained pristine, which prevented the collapse of 3D aggregations into 2D planes and eliminated the false large signals caused by IL stacking. Upon the successful establishment of this observation technique, we substantially demonstrated the nanoaggregate structure for authentic ILs. This nanostructure was exhibited in aqueous solution regardless of the type of C/H/A module. In this sense, the evidence also revealed an intriguing prerequisite for ILs to be exploited as nano-delivery carriers.Upon designing a modular IL library, we drew fundamental rules governing their biocompatibility and cytotoxicity, in which the dominant role of ILs was the “C” module, and ILs with longer cationic alkyl chains induced decreased cell viability. Specifically, lcIL nanoaggregates could enter cells and directly target mitochondria, disrupting their function and leading to mitophagy and apoptosis. The relative “black box” mechanism was further disclosed by computational chemistry (i.e., MD simulations), as lcILs could intrude the membranes and exhibited a dual-barrier crossing capacity, leading to the abnormality of membrane lipids. In contrast, the scIL nanoaggregates were sequestered in intracellular vesicles without cross-membrane or targeting capacity, resulting in uncompromised cell viability. This distinct toxicity issue was also verified via in vivo assessments, and the minimum lethal doses of ILs via different administration routes decreased in the sequence of the p.o., i.m., and i.v. routes to mice. ScILs displayed good biocompatibility, whereas the mitophagy and apoptotic signals were positively correlated with the tissue distribution signal of lcILs, which corresponded with the cellular mechanisms. The results of the long-term evaluations, such as the 36-h cell viability assay and the 7-day cytokine level analysis, provided further confirmation of the safety profile of scILs. For more similarity to the human physiological environment, we extended the scope to the canine and validated that only lcILs triggered discomfort in dogs, such as vomiting, diarrhea, and the prone position.The results of this study bring to light several significant merits that warrant discussion. Primarily, the conclusions drawn from the IL library are more reliable compared to datasets with limited types, potentially expediting the advancement of the next generation of ILs. Moreover, based on our normalized and high-quality database, a neural network model was developed for the accurate prediction of cell viability, which eliminated the problems related to heterogeneous data arising from different laboratories. In addition to the foundational understanding of ILs-bio interaction at the molecule level24,32, present findings advanced the knowledge that ILs could modulate cellular responses at the nano-scale, opening a new era in harnessing the vast IL family for nano-bio interfacial applications. Rather than relying on sporadic data, our current evidence has been pressure-tested through a whole-chain system (e.g., structural design, genetic alteration, subcellular targeting, and in vivo pharmacology/toxicology), which is more likely to withstand the test of time. Upon the comprehensive overview of ILs-bio interactions (even mimick the clinical scenarios) and clarified mechanisms, biosafety concerns arising from ILs involved in the biomedical application could be resolved, and the efficacy of ILs could be exploited rather than being seen as a problem.Our findings also offer insights into the potential of ILs as pharmaceutical candidates while uncovering the associated risks that should be avoided. In particular, special attention should be given to the distinct biocompatibility and toxicity of ILs arising from the variation in length of the cationic alkyl chain. Nevertheless, the revealed nanoaggregate structure and the cell entry capacity equip the ILs with unique drug loading and intracellular function/killing attribute, respectively. On the one hand, with ~90% of drugs suffering from poor solubility, which largely precludes their further clinical applications33,34, the superior biocompatibility and high bioavailability of scILs-based carrier provide an avenue to resolve this handicap. Beyond improving oral absorption efficacy, the structural versatility of ILs permits hydrophilic modifications (e.g., hydroxyl groups or polyethylene glycol chains) on cation head groups or anion moieties to advance targeted delivery to specific lesions. On the other hand, the confinement of charged lcIL groups around the mitochondria accelerates the membrane depolarization, which endows lcILs with promising anti-tumor properties through serious energy disruption. To fully utilize this merit, rational chemical modification (e.g., incorporating targeted ligands and photo- or pH-responsive blocks) is imperative for selective accumulation/exposure to tumor cells while minimizing the non-specific effects. Therefore, present study not only fills knowledge gaps regarding IL nanoaggregate and biosystem interactions, facilitating their safe and sustainable applications, but also ignites a spark of stimulating numerous biomedical hotspots working with ILs in the future.MethodsCell lines, organoids, and animalsBEnd.3 and HepG2 cells were obtained from the American Type Culture Collection. 4T1 cells were purchased from Procell Life Science & Technology Co., Ltd. The patient-derived primary breast cancer cells and liver cancer organoids were supplied by K2 Oncology Co., Ltd. and received approval from National Cancer Center/Cancer Hospital, Chinese Academy of Medical Science, and Peking Union Medical College (approval ID, 18-012/1641) and Yixing People’s Hospital (approval ID, 2020R060). The written informed consents were obtained from the participants (female volunteer who provided breast tissue and a male volunteer who donated liver tissue). Since sex and gender of participants are not relevant to this analysis, no sex or gender analysis was conducted in this study.C57BL/6 mice were purchased from Beijing Vital River Laboratories. Beagle dogs were purchased from Nanjing Chaimen Biotechnology Co., Ltd. and housed at EverPro Medical Co., Ltd. Dog experiments were approved by the Animal Ethics Committee of Hunan University of Chinese Medicine (approval ID, LL2022021901, LL2022021902, and HNUCM21-2312-32). Male animals were used due to their availability at the time of the experiment. The biological sex of the donor animals is not anticipated to affect the outcomes or interpretation of our findings, as the experimental design and endpoints were independent of sex. Mice and dogs were housed in ventilated cages in a 12 h light-dark cycle (8:00–20:00 light; 20:00–8:00 dark) and had access to food and water ad libitum with relatively constant temperature (21–23 °C) and humidity (55 ± 5%). C57BL/6 mice experiments were reviewed and approved by the Animal Ethics Committee of the Institute of Process Engineering (approval ID, IPEAECA2019102 and IPEAECA20241110). This study was performed in strict accordance with the Regulations for the Care and Use of Laboratory Animals and Guideline for Ethical Review of Animals (China, GB/T 35892-2018).ReagentsCxH1A1, CxH1A2, CxH1A3, CxH1A4, CxH1A5, and CxH1A6 were purchased from Shanghai Cheng Jie Chemical Co., Ltd. C12H2A1, C12H3A1, C12H4A1, and C12H5A1 were purchased from Shenyang Seddon Technology Co., Ltd. Other ILs were obtained from the Beijing Key Laboratory of Ionic Liquids Clean Process.In vitro analysis of cell viabilities in cell lines and patient-derived primary breast cancer cellsThe bEnd.3 and HepG2 cell lines were cultured in high-glucose DMEM supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin, while the 4T1 cell lines were maintained in RPMI 1640 medium with 10% FBS and 1% penicillin/streptomycin. Cells were seeded into culture dishes or flasks and incubated at 37 °C in a humidified atmosphere with 5% CO₂. Passaging was performed using a 0.25% trypsin solution for cell detachment. BEnd.3, 4T1, and HepG2 cells were seeded in 96-well plates and incubated with ILs at different concentrations and different time points (12, 24, and 36 h for HepG2 cells). Cell viability was determined by the CCK-8 assay (Beyotime) and detected using the Infinite M200 microplate spectrophotometer (TECAN).The patient-derived primary breast cancer cells (8 × 105)-Matrigel suspension were seeded onto a 96-well plate (K2OS-96-M, K2 Oncology) and cultured for 24 h in an OrganoPro™ medium (K2O-M-BR, K2 Oncology) at 37 °C and 5% CO₂. Subsequently, after incubation with PBS or ILs for 24 h, the viability of primary breast cancer cells was assessed using the CTG (Promega) assay.Machine learning predictionA fully connected feed-forward neural network was implemented using TensorFlow35,36 to predict IL-induced cell viability based on molecular and experimental features. The model architecture consisted of three layers. Five features-cationic side chain (C) module, cationic head (H) module, anion (A) module, cell type, and concentration were extracted from the dataset as input, with cell viability as the target variable. The dataset was stratified and split into training and testing sets (80:20 ratio). The model was compiled using a mean squared error (MSE) loss function and optimized with the Adam algorithm. Training was conducted, and performance was validated on the testing set. Predictions were further evaluated using synthetic datasets to test the model’s generalizability under controlled input variations. This approach allowed for systematic exploration of the relationship between IL structural features and toxicity.In vitro analysis of the viabilities of spheroids and organoidsCell spheroids were performed in a 1% agarose-coated 96-well plate with ~1000 cells per well. The cell spheroids were cultured for 4 days to reach ~300 μm in diameter and incubated with PBS or ILs for 24 h. Then, the cell spheroids were collected and stained with a calcein-AM/PI Live-Dead Cell Staining Kit (Solarbio). The samples were detected by CLSM (Nikon) using NIS Elements AR 5.20. The data were analysed by ImageJ.Following thawing from liquid nitrogen, liver cancer organoids were cultured in 24-well plates using organoid growth medium (GAS medium, K2O-CML-01801, K2 Oncology) until they were suitable for passaging. Subsequently, 50 μl of an organoid-Matrigel suspension (1:1, v/v) was injected into 96-well plates to form Matrigel droplets, which were incubated at 37 °C and 5% CO₂ for 30 min to facilitate Matrigel gelation. After adding 40 μl of GAS medium, the organoids were cultured in a 37 °C incubator for 2 days, after which tumor organoid formation and growth were monitored. After incubation with PBS or ILs for 24 h, the organoids were stained with a calcein-AM/PI Live-Dead Cell Staining Kit (Solarbio). The samples were detected by CLSM (Nikon) using NIS Elements AR 5.20. The data were analysed by ImageJ.The organoids were incubated with PBS or ILs in 96-well plates for 7 days, evaluated using CTG (Promega) assay, and detected by a FLUOstar Omega plate reader (BMG LABTECH).Characterization of IL nanoaggregatesILs were dissolved in water at a concentration of 0.04 mM. The samples were prepared by directly dropping the IL aqueous solutions onto the ultrathin carbon film. After ~1 min, the water present on the surface was carefully removed, leaving behind a thin residual layer. Subsequently, the samples were transferred to a Gatan 914 cryo-transfer specimen holder, and imaging was performed using a JEM-2100 TEM (JEOL) operating at 200 kV. Observations were conducted in a 1 h timeframe to prevent the 3D structure from collapsing into a 2D shape due to solvent evaporation under vacuum. All Cryo-TEM images were collected using a charge-coupled device camera. The size data were analysed by ImageJ.Martini coarse-grained force field37, insane.py38, and GROMACS software39 (version 2019.6) were used for MD simulations to capture the self-assembly processes of ILs. In the self-assembly systems, 4000 IL molecules were initially evenly distributed in a simulation box of 56 × 56 × 56 nm3. The box was solvated with water molecules for an IL-water ratio of 1:116.The zeta potential of IL nanoaggregates in aqueous solution was measured via a ZetaSizer (NANO ZS).Cellular internalization and subcellular interaction analysisHepG2 cells were incubated with the Cy5-ILs (chemical modification, Fanbo Biochemicals Co., Ltd.) for 2 h. After three washes with PBS, the cells were fixed in 4% paraformaldehyde for 15 min. Subsequently, the cells were stained with FITC-phalloidin (Solarbio) for 30 min and 4′,6-diamidino-2-phenylindole (DAPI, Solarbio) for 10 min and detected by CLSM (Nikon) using NIS Elements AR 5.20. The data were analysed by ImageJ.To delineate the endocytic entry mechanism of IL nanoaggregates, cells were pretreated with macropinocytosis inhibitor (amiloride hydrochloride (150 μM)) at 37 °C for 1 h. Then, ILs were added into the cells with or without inhibitors for 2 h incubation. The cells were detected by flow cytometry (CytoFLEX LX, Beckman Colter).1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC, Sigma-Aldrich) lipids were initially dissolved, evaporated, and dried. PBS was then added, and lipid vesicles were prepared via sonication40,41. The lipid vesicles were stained with the fluorescent dye 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI, Solarbio) for 10 min. After washes, the samples were incubated with either PBS or Cy5-labeled ILs for 30 min. The samples were subsequently imaged using CLSM (Nikon) with NIS Elements AR 5.20 software.HepG2 cells were cultivated in 6-well plates and incubated with PBS/ILs for 0.5 or 2 h. Then, cells were collected and fixed in glutaraldehyde solution at 4 °C overnight. For TEM observation, the samples were stained, dehydrated, and observed by H-7650B (HITACHI) operating at 80 kV. The typical treatment for the sample sections included dehydration, standing, and other processes that involved metal solutions (e.g., osmic acid, lead citrate, and uranyl acetate), which might cause a slightly larger signal for ILs.The mitochondria of cells were stained with MitoBright LT Red (DOJINDO). Then, the cells were incubated without or with Cy5-C12MIMCl nanoaggregates for 30 or 120 min and fixed in 4% paraformaldehyde. Colocalization was visualized using STED microscopy (SP8, Leica).RNA sequencing analysisCells treated with PBS or ILs for 2 h were collected, quickly frozen, and detected by Novogene Co., Ltd. Briefly, after total RNA extraction, mRNA was purified using Oligo (dT) beads, fragmented by adding fragmentation buffer, and reverse-transcribed into cDNA according to the manufacturer’s recommendations. Second-strand cDNA synthesis was subsequently conducted using DNA Polymerase I and RNase H. Following polymerase chain reaction (PCR) amplification, cDNA products were sequenced using an Illumina NovaSeq 6000, and the quality was assessed on the Agilent Bioanalyzer 2100 system. Differential expression analysis was performed using the DESeq2 R package. Genes with an adjusted P