A Virus-Inspired Inhalable Liponanogel Induces Potent Antitumor Immunity and Regression in Metastatic Lung Tumors

The lungs are the most common sites of metastases for cancer cells (1, 2). While surgery has been widely used to remove primary tumors and prolong survival, a general method that can efficiently suppress lung metastases remains lacking (3, 4). Recent breakthroughs in immunotherapy (e.g., immune checkpoint blockade) have created exciting opportunities to improve the therapeutic outcome of lung cancers, including lung metastases (5–7). However, only a small subset of patients (10%–30%) can respond to immune checkpoint blockers (ICB; ref. 8). Further analyses indicate that the absence of CD8⁺ T cells in the tumor microenvironment is an important reason for the low response rate (9). Pulmonary delivery of immunostimulatory agents to directly remodel the lung tumor microenvironment represents a promising strategy to activate the antitumor CD8⁺ T-cell immunity and suppress lung metastases (10–12). For example, poly IC (pIC), a double-stranded RNA, can act on multiple pattern recognition receptors such as TLR3 and MDA5/RIG-I, which are responsible for sensing the invasion of RNA viruses and inducing potent T-cell responses (13–15). Recent studies have shown that the cytosolic MDA5/RIG-I is more efficient than the endosomal TLR3 in inducing the secretion of type I interferons such as IFNβ, which is critical for promoting the activation of T-cell immunity (16–18). However, following pulmonary delivery, pIC suffers from rapid degradation and poor endosomal escape (19), thus limiting the efficiency of pIC to reach MDA5/RIG-I and compromising type I interferon secretion.

Lipid nanoparticles (LNP) have been widely used to deliver nucleic acids such as siRNA and mRNA (20–23). The PEGylated surface of LNP has also been shown to facilitate mucus penetration and enhance its interaction with target cells (24, 25), so using LNP to deliver pIC holds great promise to unleash the adjuvant potential of pIC. Despite this, the cytosolic delivery efficiency of LNP mainly relies on the lipid structure, which may limit the potency of pIC. A strategy that can complement the feature of LNP to tune the cytosolic delivery efficiency would be beneficial to amplify the potency of pIC further.

RNA viruses evolve to have sophisticated structures to efficiently pack and deliver their RNA into the cytosol of host cells for replication. However, potential safety concerns can limit the use of viral delivery systems for therapeutic RNA delivery (26). This motivates us to mimic the viral systems to design a safe and efficient nonviral delivery system that complements the feature of LNP. Notably, SARS-CoV-2 utilizes ribonucleoproteins to efficiently tune the loading and delivery of viral RNA (27). In particular, the unmodified viral nucleocapsid (N) protein forms gel-like condensates with viral RNA based on multivalent RNA–protein and protein–protein interactions, thus facilitating the nucleocapsid assembly. Early in infection, the serine–arginine (SR)–rich sequence of N protein is rapidly phosphorylated at multiple sites by cytoplasmic kinases. Phosphorylation reduces the interactions between RNA and protein, generating a more liquid-like droplet, thus facilitating viral genome processing (28). Inspired by this phenomenon, we aim to use hyaluronic acid (HA) modified with bisphosphonate (HA-BP) within LNP to tune RNA delivery. HA-BP within the LNP is designed to complex with RNA to form a nanogel (NG) in the presence of Ca²⁺. Following cellular uptake, the anionic NG is protonated within acidic endosomes, causing an influx of anionic ions such as Cl⁻. This, in turn, increases the osmotic pressure and endosomal membrane permeability, ultimately leading to efficient RNA delivery to the cytosol.

Here, we show that a liponanogel (LNG) consisting of an anionic polymer hyaluronic acid–based NG coated by a lipid shell is successfully prepared using a microfluidic device. The lipid shell serves as a protective layer to stabilize the NG core and facilitates mucus penetration in the lungs. Moreover, the NG core can be protonated within acidic endosomes to enhance the endosomal membrane permeability and cytosolic delivery of pIC. LNG–pIC induces more IFNβ than pIC and LNP-pIC in the lungs after pulmonary delivery. Furthermore, LNG-pIC induces potent CD8⁺ T-cell immunity and regresses melanoma and breast cancer lung metastases. Notably, LNG-pIC can be easily incorporated with PD-L1 siRNA to enhance further the therapeutic effect without the systemic use of ICB. Our study sheds light on the importance of using an NG core within a lipid shell for enhanced cytosolic delivery and may inspire the development of new strategies for cancer immunotherapy.

The HA-BP was synthesized by conjugating the amine group of bisphosphonic acid to the carboxyl group of hyaluronic acid (29). Briefly, HA (1 g, 2.5 mmol) was dissolved in 100 mL of deionized water. 1-(3-Dimethylaminopropy1)-3-ethylcarbodiimide hydrochloride (EDC·HCl; 959 mg, 3.75 mmol) and N-hydroxysuccinimide (NHS; 575 mg, 3.75 mmol) were slowly added to the HA solution. The solution was adjusted to pH 4 to 5 and stirred for 45 minutes at room temperature before adding alendronate sodium (813 mg, 2.5 mmol). The reaction mixture was adjusted to pH 6 to 7 and stirred for 3 days at room temperature. The product was dialyzed using a dialysis bag with a molecular weight cut-off of 8 kDa in deionized water to remove the impurities. The structure was confirmed by ¹H-NMR spectrometry.

NG-pIC was fabricated by a droplet-confined nanoprecipitation in water-in-oil (w/o) microemulsion (30–32). Briefly, 300 µL of 500-mmol/L CaCl₂ with 100 µL of 3-mg/mL pIC was dispersed in 15-mL cyclohexane/Igepal CO-520 (70/30 V/V) solution to form a well-dispersed water-in-oil reverse microemulsion A. Then, 300 µL of 25-mmol/L Na₂HPO₄ (pH = 9.0) with 200 µL of 2-mg/mL HA-BP was dispersed in a 15-mL oil phase to obtain microemulsion B. A and B were mixed and incubated for 30 minutes at room temperature to allow for the formation of NG. The mixture was washed twice with 30 mL of absolute ethanol and centrifuged at 12,000 g for 20 minutes to remove cyclohexane and surfactant. The pellet was dispersed in 1 mL of deionized water to obtain NG-pIC.

LNG-pIC was prepared using a microfluidic system (INano E, Micro&Nano). Briefly, 1,2-dilinoleyloxy-3-dimethylaminopropane (DLin-MC3-DMA, AVT), cholesterol (AVT), 1,2-Distearoyl-sn-glycero-3-phosphocholine (DSPC; AVT), and 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy (polyethylene glycol)-2000; DSPE-PEG2000; AVT] were dissolved in pure ethanol at molar ratios of 47:38.5:9.5:5 with a final lipid concentration of 10 mg/mL. In some experiments, DSPE-PEG2000 was replaced with an equal amount of 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-PEG2000) to evaluate the effect of the lipid tail. The lipid solution was then mixed with a sodium hydrogen citrate buffer containing HA-BP (0.45 mg/mL) and pIC (0.5 mg/mL) using the microfluidic system. The LNG-pIC was dialyzed against 1% sodium chloride buffer overnight at 4°C. Then, 100 µL of CaCl₂ solution (500 mmol/L, pH 7.4) was added to 1 mL of LNG-pIC solution and incubated at room temperature for 30 minutes. The solution was dialyzed against 0.9% sodium chloride buffer for 6 hours at 4°C. The size and zeta potential were determined by dynamic laser scattering on a Malvern Zetasizer. The morphological characterization was carried out by cryo-electron microscopy (cryo-EM). For cryo-EM, 4 µL of each sample was dropped on a glow-discharged copper grid coated with holey carbon (R 2/2; Quantifoil), incubated for 2 minutes, blotted for 3.5 to 4.5 s, and then plunged into liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific). The samples were loaded on a 200-kV FEI Talos Arctica transmission electron microscope (Thermo Fisher Scientific) equipped with a K2 direct electron detector. For each image, a movie consisting of 32 frames was recorded in counting mode at nominal magnifications of 17,500× (pixel size, 2.412 Å) and at a defocus of −3.0 µm in SerialEM (33). The electron beam-induced motion was corrected by MotionCor2 (34). The pIC encapsulation efficiency was analyzed via gel electrophoresis (1% agarose gel, 120 mV, 15 minutes) and quantified by ImageJ.

Bronchoalveolar lavage fluid (BALF) was collected by sequentially lavaging the lungs three times with 1 mL of PBS, and the recovered fluids were pooled for each animal. Then, a 20-µL NP containing 10-µg/mL pIC was mixed with 500-µL BALF and incubated at room temperature. At 2 and 24 hours after incubation, hydrodynamic diameters were determined by dynamic laser scattering (25). To learn the stability of NP, 50-µL NP containing 25-µg/mL pIC was incubated with 1-mL BALF at 37°C under gentle shaking (120 rpm). At 24 hours, samples were collected, and the amount of remaining pIC was analyzed via gel electrophoresis (1% agarose gel, 120 mV, 15 minutes).

B16F10, 4T1-Luc, and DC2.4 cell lines were purchased from National Infrastructure of Cell-Line Resource, China. The cells were maintained in DMEM (Gibco) containing 1% penicillin/streptomycin (Gibco) and 10% FBS. The cell lines were cultured in a 37°C incubator with 5% CO₂. Throughout the studies, all cells were used as received and tested negative for Mycoplasma contamination and rodent pathogens.

Bone marrow-derived dendritic cells (BMDC) were generated by culturing bone marrow cells flushed from the femurs of C57BL/6 mice in a BMDC medium. The culture medium was half-replaced every 3 days, and the nonadherent and loosely adherent immature dendritic cells (DC) were collected on day 6. BMDCs were seeded in confocal dishes at a density of 1 × 10⁵ cells/well for 24 hours, and the medium was replaced with 1-mL fresh medium containing 8-µg/mL pIC in the form of NG, LNP, or LNG. After incubation for 24 hours, the cells were washed twice with PBS and stained with 5-µmol/L acridine orange (AO) solution in culture medium for 30 minutes. Then, the cells were observed with a confocal laser scanning microscope (Zeiss LSM780).

For PD-L1 silencing studies, BMDCs were seeded in 12-well plates at 8.0 × 10⁵ cells/well. After 24 hours, the medium was replaced with 1 mL fresh medium containing 1-μg/mL IFNγ, and 8 μg/mL of PD-L1 siRNA in the form of NG, LNP, or LNG was added to the medium. After incubation for 48 hours, the cells were collected, washed with FACS buffer, incubated with anti-CD16/32 for 10 minutes at room temperature, and stained with antibodies against CD11c and PD-L1 (Biolegend) at 4°C for 20 minutes in the dark. Cells were washed with FACS buffer and resuspended in 2-µg/mL DAPI before analysis via flow cytometry (BD LSRFortessa SORP). For luciferase silencing studies, Luc-DC2.4 cells were seeded in 48-well plates at 1.0 × 10⁵ cells/well. After overnight incubation, the medium was replaced with 0.5-mL fresh medium containing 8 μg/mL of luciferase siRNA in the form of NG, LNP, or LNG. After incubation for 24 hours, the expression of luciferase in DC2.4 cells was determined using Luciferase Assay System with Reporter Lysis Buffer assay kits (Promega E4030) following the manufacturer’s instructions. The luminescence intensity was measured using a microplate reader (Bio Tek).

For the biodistribution study, 2 × 10⁵ B16F10 cells were intravenously injected into female C57BL/6 mice on day 0. On day 10, the mice were administered with 1,1'-dioctadecyl-3,3,3',3'-tetramethylindotricarbocyanine iodide (DiR)–labeled LNP or LNG intratracheally using a microsprayer aerosolizer. Then, the mice were sacrificed at 1, 24, and 48 hours to harvest major organs for ex vivo imaging using an IVIS optical imaging system. The signal intensity was analyzed using the Living Image software. To measure the uptake of LNG and LNP by different subsets of cells, including DCs (CD11c⁺F4/80⁻), interstitial macrophages (IM; CD11c⁻F4/80⁺), alveolar macrophages (AM; CD11c⁺F4/80⁺), bronchial epithelial cells (CD326⁺CD31⁻CD24⁺), and alveolar epithelial cells (CD326⁺CD31⁻CD24⁻) in the lungs (35, 36), the lung tissues were harvested at 1, 24, 48, and 96 hours after inhalation of DiR-labeled LNP or LNG. The lungs were further cut into small pieces and treated with 1-mg/mL collagenase-type IV and 0.1-mg/mL DNase I in serum-free RPMI for 30 minutes at 37°C with gentle shaking. The cell suspensions were passed through 70-µm strainers and washed with FACS buffer. The cells were incubated with CD16/32 for 10 minutes and then incubated with CD11b-PE-Cy5 (BioLegend, M1/70), CD11c-FITC (BioLegend, N418), F4/80-BV605 (BioLegend, BM8), CD326 antigen-presenting cell (APC; BioLegend, G8.8), CD31-FITC (BioLegend, 390), and CD24-PE (BioLegend, 30-F1) on ice before flow cytometry (BD LSRFortessa SORP). In some experiments, the lung tissues were preserved and prepared into 10-μm cryosections, which were co-stained with anti-mouse CD11c-PE (1:200; BioLegend, N418) and anti-luciferase (1:500; Sigma-Aldrich, L0159), followed by staining with the FITC–anti-rabbit secondary antibody (1:800; Jackson Immuno) and observation using a confocal laser scanning microscope (Leica SP8 STED).

The lungs collected from mice were homogenized in PBS and then centrifuged at 900 g for 20 minutes. The supernatants were transferred to fresh microtubes, and the concentration of IFNβ was measured with ELISA kits (Invivogen) following the manufacturer’s instructions.

All animal experiments were in accordance with and approved by the Institutional Animal Care and Use Committee at Tsinghua University. For the B16F10 lung metastasis model, female C57BL/6 mice of age 6 to 8 weeks (Beijing Vital River Laboratory Animal Technology Co., Ltd.) were intravenously injected with 2 × 10⁵ B16F10 cells (35). After confirming visible multifocal lung lesions on the surface of both lungs by euthanizing randomly selected animals on day 7, the mice were randomly divided into five different groups and intratracheally administrated with PBS, free pIC, NG-pIC, LNP-pIC, and LNG-pIC (containing 0.5-mg/kg pIC) every 5 days (37). In some experiments, female TLR3^−/− C57BL/6 mice (Shanghai Model Organisms Center, Inc.) and female MAVS^−/− C57BL/6 mice (Jackson Laboratory) aged 6 to 8 weeks were immunized as described above. The mice were sacrificed for enumeration of metastatic lung foci on day 17. For tumor-infiltrating cytotoxic lymphocytes analysis, the lung tissues were harvested and cut into small pieces, treated with 1-mg/mL collagenase-type IV and 0.1-mg/mL DNase I in serum-free RPMI for 30 minutes at 37°C with gentle shaking. The cell suspensions were passed through a 70-µm strainer and washed with FACS buffer. The cells were incubated with CD16/32 for 10 minutes and then incubated with anti-CD3 (BioLegend, 30-F11), anti-CD8 (BioLegend, 53.6.7), anti-CD4 (BioLegend, GK1.5), anti-NK1.1 (BioLegend, PK136), anti-CD80 (BioLegend, 16-10A1), and anti-CD86 (BioLegend, M5/114.15.2) on ice before flow cytometry (BD LSRFortessa SORP).

For the 4T1-Luc lung metastasis model, BALB/c mice (6–8 weeks, female, Vital River) were intravenously injected with 2 × 10⁵ 4T1-Luc cells on day 0. After confirming the establishment of 4T1-Luc lung metastases on day 7 by measuring the luciferase signal with IVIS imaging, the mice were randomly divided into different groups and intratracheally administrated with PBS, LNG-pIC (containing 0.5-mg/kg pIC), LNG-siRNA (containing 0.5-mg/kg siRNA), and LNG-pIC/siRNA (containing 0.5-mg/kg pIC and 0.5 mg/kg siRNA) every 5 days. In some experiments, animals receiving LNG-pIC (containing 0.5-mg/kg pIC) were intraperitoneally injected with αPD-L1 (BioXCell, B7-H1, 2.5-mg/kg) every 5 days (11). The tumor growth was monitored by measuring the bioluminescence signal of tumor cells using an IVIS imaging system.

Untreated or LNG-pIC–treated animals (with B16F10 metastases) were euthanized on day 18 to harvest the spleens and blood (38). Spleens were prepared into single-cell suspensions by passing through 70-µm filters. Red blood cells in the spleens and the blood were lysed by the acetone cyanohydrin ketal lysis buffer for 5 minutes at room temperature. Splenocytes and blood cells from the same group were pooled together, followed by processing using the MojoSort Mouse CD8 T Cell Isolation Kit (Biolegend) to isolate CD8⁺ T cells. Two million CD8⁺ T cells per mouse were intravenously injected into C57BL/6 mice with B16F10 metastases (the metastases were established 7 days before receiving the adoptive T-cell therapy by intravenously injecting 2 × 10⁵ B16F10 cells). On day 17 postestablishment of the metastases, the mice were sacrificed for enumeration of metastatic lung foci.

C57BL/6 mice (6–8 weeks, female, Vital River) were injected intravenously with 2 × 10⁵ B16F10 cells on day 0 to establish the lung metastasis model. On days 7 and 12, mice were treated with LNG-pIC/siRNA (containing 0.5-mg/kg pIC and 0.5 mg/kg siRNA) as described above. Lymphocyte depletion was achieved using antibodies against CD4 (BioXCell, GK1.5), CD8 (BioXCell, 2.43), and NK1.1 (BioXCell, PK136). Antibodies (200 µg/dose) were administered intraperitoneally on days −2, 0, 2, 6, 13, and 20 (39).

Statistical analysis was performed using Prism 8.0 (GraphPad Software). Statistical significance was determined by one-way or two-way ANOVA. No data were excluded. P values less than 0.05 were considered statistically significant. All values are presented as mean ± SEM. ^∗, P < 0.05; ^∗∗, P < 0.01; ^∗∗∗, P < 0.001; ^∗∗∗∗P < 0.0001.

Summary and representative data generated in this study are present in the article and the Supplementary Data files. All raw data can be obtained from the corresponding author upon request.

We first developed a scalable protocol for the preparation of LNG. Briefly, we used a microfluidic device to encapsulate hyaluronic acid-bisphosphonate (HA-BP; Supplementary Fig. S1) and pIC within a lipid shell and then used Ca²⁺ to crosslink HA-BP and pIC (Supplementary Fig. S2A). To prepare the NG control, HA-BP, and pIC were crosslinked by calcium phosphate in a water-in-oil (W/O) emulsion, followed by extensive washing with ethanol to obtain NG (Supplementary Fig. S2B). Dynamic light scattering showed that LNG was 127.6 ± 1.3 nm in diameter, slightly bigger than NG. LNP without HA-BP NG was 130.7 ± 1.5 nm in diameter, suggesting the presence of HA-BP NG did not significantly change the size (Supplementary Table S1). Cryo-electron microscopy (Cryo-EM) confirmed that LNG, LNP, and NG were homogeneous and spherical (Fig. 1A). NG exhibited a zeta potential of around −20 mV. In contrast, LNG had a zeta potential of around −5 mV, close to the neutral charge of LNP (Fig. 1B). Notably, both LNG and LNP exhibited good stability without any significant change in size or polydispersity index after incubation with BALF for 2 and 24 hours. In contrast, NG had a significant increase in size and polydispersity index under the same condition (Fig. 1C and D). These results indicate that the microfluidic device enabled the facile preparation of LNG with good stability. Moreover, both LNG and LNP could efficiently encapsulate pIC, with an encapsulation efficiency of over 80% (Fig. 1E; Supplementary Table S1). Compared with NG, which only partially protected pIC from degradation in the presence of BALF, LNG and LNP offered more complete protection in the presence of BALF (Fig. 1F).

LNG and LNP were taken up by DCs similarly, resulting in colocalization with lysosomes (Supplementary Fig. S3A and S3B). Enhancing the endosomal membrane permeability is an important strategy to promote cytosolic delivery efficiency (40–42). This motivated us to investigate whether LNG can increase endosomal membrane permeability. We took advantage of a fluorescent dye AO, which has red fluorescence within acidic endosomes and green fluorescence in neutral nuclei. The decrease in red fluorescence of AO has been widely used as a readout to evaluate the increase of endosomal membrane permeability (43, 44). Confocal microscopy indicated that NG- and LNP-treated BMDCs exhibited a moderate decrease in the red fluorescence of AO within endosomes compared with untreated BMDC. LNG-treated BMDC showed even less red fluorescence than NG and LNP (Fig. 2A; Supplementary Fig. S4A–C), indicating that NG within the lipid shell contributed to increased endosomal membrane permeability. The flow cytometry analysis showed a similar pattern, with LNG showing a much weaker red fluorescence compared with NG (P < 0.001) and LNP (P < 0.05; Fig. 2B). Notably, LNG did not compromise the viability of BMDC (Supplementary Fig. S5), indicating the excellent safety profile of LNG. To directly analyze the cytosolic delivery efficiency of LNG, we used PD-L1 siRNA as a model molecule because the action site of siRNA is in the cytosol. In parallel with the effect on the endosomal membrane permeability, LNG exhibited 200% and 77% less PD-L1 than NG (P < 0.0001) and LNP (P < 0.0001), respectively (Fig. 2C), indicating LNG was more effective than NG and LNP in delivering siRNA into the cytosol. As luciferase-expressing cells have been widely used to quantify the cytosolic delivery efficiency of siRNA (45, 46), we also treated luciferase-expressing DCs with different formulations. As shown in Fig. 2D, LNG-siRNA induced 3.7-fold lower luciferase expression than LNP-siRNA and 7.3-fold lower luciferase expression than NG-siRNA. Coupled with the endosomal membrane permeability increase after LNG treatment, these findings indicate that LNG is a robust platform for efficient cytosolic delivery.

To learn the retention behavior of LNG in the lungs, DiR-labeled LNG or DiR-labeled LNP was administered to mice with B16F10 lung metastases through intratracheal inhalation. LNG was evenly distributed throughout the whole lungs as early as 1 hour, with the DiR fluorescence signal slowly decreasing over time until at least 48 hours after administration (Fig. 3A). Notably, no significant signal was detected in other normal organs, indicating most LNG was retained in the lungs, which was beneficial for maximizing the activity while avoiding the systemic side effect. DiR-labeled LNP exhibited a distribution profile similar to that of LNG (Fig. 3B). Metastases-bearing lung tissues obtained 24 and 48 hours postinhalation were subjected to immunofluorescent staining. The images clearly showed that DiR-labeled NPs (purple) co-localized predominantly with CD11c⁺ APCs (red). Co-staining of B16F10 tumor cells with anti-luciferase (green) indicated that the LNP or LNG was distributed well into individual metastases and captured by intratumoral APCs (Fig. 3C). We further quantified the uptake of LNG by different subsets of cells within the lungs. In particular, LNG was efficiently taken up by innate immune cells such as DCs (DCs, CD11c⁺F4/80⁻), IMs (CD11c⁻F4/80⁺), and AMs (CD11c⁺F4/80⁺). One hour after inhalation, 11.5% DC, 17% IMs, and 30% AMs were DiR+, which increased over 24 hours for DCs and IMs and increased over 48 hours for AMs, followed by a decrease at 96 hours. Moreover, 1 hour after inhalation, 6.3% bronchial epithelial cells (CD326⁺CD31⁻CD24⁺) and 5.7% alveolar epithelial cells (CD326⁺CD31⁻CD24⁻) were DiR+, which increased over 48 hours, followed by a decrease at 96 hours. DiR-labeled LNP followed a similar pattern in terms of uptake by different subsets of cells (Fig. 3D). Interestingly, we found LNG had increased retention in the lungs compared with NG (Supplementary Fig. S6), indicating the lipid shell is beneficial for facilitating the retention in the lungs. Together, these results suggest that inhalation allows for efficient delivery of LNG to APCs in the lungs. The NG within the lipid shell did not significantly change the delivery profile at the tissue level and cellular level compared with LNP without the NG core. These features, coupled with the efficient cytosolic delivery of LNG, make LNG a more promising platform than LNP for cancer immunotherapy.

Having seen a beneficial delivery profile of LNG following pulmonary delivery, we sought to learn their impact on the immune responses. Naïve C57BL/6 mice were administered with different formulations containing pIC through intratracheal inhalation (47). After 24 hours, we harvested lungs to measure the concentrations of IFNβ. pIC or NG-pIC failed to improve IFNβ secretion in the lungs significantly (Fig. 4A). LNP-pIC induced 6.9-fold more IFNβ than pIC (P < 0.0001). Remarkably, LNG-pIC induced 13.7-fold more IFNβ than pIC (P < 0.0001) and two-fold more IFNβ than LNP-pIC (P < 0.0001). We next established the B16F10 lung metastasis model by intravenously injecting B16F10 cells into C57BL/6 mice and treated these animals with different formulations containing pIC on days 7, 12, and 17 through intratracheal inhalation. We harvested lungs to analyze the activation of different immune cells on day 18. LNG-pIC induced 14.3-fold and 1.4-fold higher levels of CD80⁺CD86⁺ DCs than pIC (P < 0.0001) and LNP-pIC (P < 0.05), respectively, while pIC or NG-pIC failed to significantly upregulate CD80⁺CD86⁺ on DCs compared with the untreated control (Fig. 4B and C; Supplementary Fig. S7). Moreover, LNG-pIC induced more CD8⁺ T than pIC and LNP-pIC, while pIC or NG-pIC failed to significantly improve the CD8⁺ T-cell levels (Fig. 4D and E). LNG-pIC also induced 2.3-fold and 2.1-fold more IFNγ⁺ CD8⁺ T cells than NG-pIC and LNP-pIC, respectively, while pIC alone failed to increase IFNγ⁺ CD8⁺ T cells compared with the untreated control (Fig. 4F and G). pIC or NG-pIC did not significantly boost NK activation, but LNP-pIC and LNG-pIC both enhanced the NK activation compared with the untreated control (P < 0.05; Fig. 4H). Together, these results indicate that LNG-pIC was overall more effective than other pIC formulations in activating immunity. In line with the enhanced immune activation, LNG-pIC was significantly more effective than LNP-pIC (P < 0.05), NG-pIC (P < 0.001), and pIC (P < 0.0001) in suppressing the lung metastases of B16F10 tumors (Fig. 4I and J). We next investigated the histopathological changes of lung slices with the hematoxylin and eosin staining. As shown in Supplementary Fig. S8A, LNG-pIC exhibited less tumor burdens compared with other groups, further demonstrating the excellent therapeutic efficiency of LNG-pIC. LNG-pIC exhibited longer survival compared with LNP-pIC and NG-pIC, while pIC alone failed to improve the survival compared with the untreated control (Supplementary Fig. S8B). To demonstrate the role of T-cell immunity, we performed adoptive transfer of T cells to untreated mice with B16F10 metastases. As shown in Supplementary Fig. S9, the adoptive transfer of T cells from animals receiving LNG-pIC to untreated mice with B16F10 metastases led to less metastases compared with the untreated B16F10 metastases or adoptive transfer of T cells from animals receiving PBS to untreated B16F10 metastases. Altogether, these results indicate that T cells activated by LNG-pIC play important roles in mediating the therapeutic effect. Notably, treatment with LNG-pIC or other formulations did not cause significant changes to the body weight or serum ALT, AST, ALP, and BUN levels (Supplementary Fig. S10A–E). LNG-pIC did not cause any toxicity to the major organs (Supplementary Fig. S10F) either, indicating the excellent safety profile for pulmonary delivery of this formulation.

We next investigated critical factors that may affect the activity of LNG-pIC. As the tail length of phospholipid anchoring polyethylene glycol (PEG) can affect the dissociation rate of lipid-PEG from LNG (48), we compared the immune activation of LNG containing DMG-PEG or 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy (PEG); DSPE-PEG]. Similar levels of IFNβ were detected for DMG-PEG or DSPE-PEG modified LNG (Supplementary Fig. S11), indicating the immune activation of LNG was not sensitive to the tail length of phospholipid anchoring PEG. LNP-pIC formulations exhibited a similar pattern, although the absolute IFNβ levels were lower than that of LNG-pIC (Supplementary Fig. S11). Notably, blank LNG without pIC failed to induce a detectable level of IFNβ or any meaningful therapeutic effect compared with the untreated control, indicating pIC was critical in mediating the immune activation and therapeutic effect (Fig. 5A–C). Moreover, knocking out MAVS (the signaling protein downstream of cytosolic MDA5/RIG-I) abrogated the secretion of IFNβ and therapeutic effect, while knocking out TLR3 did not have a significant impact on IFNβ secretion or the therapeutic efficacy (Fig. 5D–F). Together, these results suggest that enhanced cytosolic sensing of pIC achieved by LNG was critical in mediating the effect of LNG-pIC.

As LNG-pIC enhanced CD8⁺ T-cell levels in the lung tumor microenvironment, this motivated us to use PD-L1 siRNA with LNG-pIC. LNG-siRNA or LNG-pIC alone potently suppressed the lung metastases of B16F10 tumors. Remarkably, LNG-pIC/siRNA substantially suppressed the lung metastases of B16F10 tumors, indicating a clear synergy between pIC and PD-L1 siRNA (Fig. 6A and B). Hematoxylin and eosin staining confirmed that LNG-pIC/siRNA more efficiently decreased tumor burdens in the lungs compared with LNG-siRNA or LNG-pIC (Fig. 6C). Immunofluorescence staining indicated that LNG-pIC/siRNA induced higher levels of CD8⁺ T cells in the lungs compared with LNG-siRNA or LNG-pIC (Fig. 6D). We next investigated PD-L1 expression levels on B16F10 cells and other immune cells. As shown in Supplementary Fig. S12A–C, LNG-pIC/siRNA induced lower PD-L1 expression on tumor cells, DC, and macrophages compared with LNG-pIC and LNG-siRNA. Moreover, LNG-pIC/siRNA induced lower PD-1 expression CD8⁺ and CD4⁺ T cells compared with LNG-pIC and LNG-siRNA (Supplementary Fig. S12D and S12E). We next sought to understand how different subsets of immune cells have contributed to the potent therapeutic effect of LNG-pIC/siRNA. During the therapeutic study, we used different antibodies to deplete CD4⁺, CD8⁺, or NK cells (39). Interestingly, depleting any of these cells strongly compromised the therapeutic efficacy of LNG-pIC/siRNA, with lung metastases similar to or even more severe than the untreated animals (Fig. 6E and F). These results indicate that multiple subsets of immune cells indeed contribute to the therapeutic effect of LNG-pIC/siRNA.

To learn whether LNG-pIC/siRNA can suppress the lung metastases of other tumor cells, we established the breast cancer lung metastasis model by intravenously injecting 4T1-Luc cells into BALB/c mice. LNG-pIC alone exhibited a stronger therapeutic effect than LNG-siRNA alone, which only slightly suppressed lung metastases. Remarkably, LNG-pIC/siRNA induced a more potent therapeutic effect than LNG-pIC and LNG-siRNA (Fig. 7A–C). The median survival of LNG-pIC/siRNA was 40 days, which was longer than 31 days for LNG-pIC (P < 0.01) and 21 days for LNG-siRNA (P < 0.01; Fig. 7D). We have performed additional experiments to compare LNG-pIC/siRNA and LNG-pIC combined with systemic αPD-L1 mAb therapy. As shown in Supplementary Fig. S13, LNG-pIC/siRNA was as good as LNG-pIC combined with systemic αPD-L1 mAb. These results indicate that siRNA might offer an alternative strategy for enhancing immunotherapy. Moreover, as siRNA and pIC can be easily packed within the same formulation, the spatiotemporal distribution of siRNA can be easily tuned to meet the therapeutic need while minimizing the systemic side effect.

In this study, we have developed an LNG platform for enhanced pulmonary delivery of pIC to suppress lung metastases. Our results indicate that LNG promoted cytosolic delivery by enhancing endosomal membrane permeability. Moreover, following pulmonary delivery, LNG was retained in the lungs for at least 48 hours and taken up by multiple populations of APCs. These features allowed for efficient secretion of type I interferons in the lungs, which ultimately induced potent T-cell immunity that synergized with PD-L1 siRNA to suppress the lung metastases of multiple tumor models.

Lung cancer, including lung metastases of melanoma and breast cancer, is a common cause of patient death (1, 49). While ICB have revolutionized the field of cancer therapy, the response rate remains low, which is partially due to the lack of antitumor T-cell immunity (7, 8). Pulmonary delivery of immunostimulatory agents represents a promising strategy to remodel the lung tumor microenvironment to induce antitumor immunity (10, 11). In particular, pIC is a widely used vaccine adjuvant that can act on TLR3 and MDA5/RIG-I to activate APCs (50). Recent studies have shown that cytosolic MDA5/RIG-I is more effective than endosomal TLR3 in inducing type I interferon secretion, which is critical for promoting adaptive antitumor immunity (16–18). Therefore, efficient delivery of pIC to the cytosol of APCs in the lung tumor microenvironment holds great promise to induce antitumor immunity. However, pIC suffers from degradation and poor cytosolic delivery to activate MDA5/RIG-I (19). Our results indicate that the virus-inspired LNG consisting of an NG core coated with a lipid shell efficiently protected pIC from degradation in the presence of BALF. LNG was also more effective than the state-of-the-art LNP in promoting the cytosolic delivery of pIC. We demonstrated that the enhanced cytosolic delivery was ascribed to anionic polymer–based NG within LNG, which significantly improved the endosomal membrane permeability compared with LNP lacking the NG. Mechanistically, anionic polymers can be protonated within acidic endosomes (51). We speculate that the consumption of H⁺ can promote the influx of anionic Cl⁻ and increase the osmotic pressure within endosomes, resulting in enhanced endosomal permeability. Our strategy thus complements the mechanism utilized by the LNP for cytosolic delivery and may have broad applications for other molecules acting in the cytosol.

Inhalable formulations are regarded as one of the most promising strategies to locally target the lungs (35). While inhalation allows both LNG and LNP to directly reach deep lungs and bypasses the use of lung-targeting moieties, the presence of other barriers such as the endosomal membrane can prevent therapeutic molecules from reaching the target in the cytosol. In line with this, although both LNG and LNP exhibited similar lung retentions, LNG had substantially more efficient cytosolic delivery than the state-of-art LNP, resulting in more efficient delivery of pIC to the cytosolic targets, ultimately improving the antitumor immunity and therapeutic effects. Moreover, LNG is compatible with engineering using different targeting moieties in the future and holds great promise to treat diseases outside the lungs following intravenous injection (52).

To understand the impact of LNG-pIC on ICB therapy, we combined PD-L1 siRNA and pIC in the same LNG platform. Remarkably, the combination therapy exhibited a strikingly strong therapeutic effect. Depleting CD8⁺, CD4⁺, or NK cells strongly compromised the therapeutic effect of LNG-pIC/siRNA. These results are consistent with previous studies showing that multiple immune cell subsets can contribute to the therapeutic effect of immunostimulatory molecules.

In conclusion, we have developed an LNG for efficiently delivering immunostimulatory molecules to remodel the lung tumor microenvironment. Pulmonary delivery of LNG-pIC results in extended lung retention and efficient uptake by APCs. LNG-pIC also promotes the cytosolic sensing of pIC by enhancing the endosomal membrane permeability, leading to higher levels of type I interferons in the lungs and stronger antitumor immunity than the state-of-the-art LNP formulations. Remarkably, combining LNG-pIC with PD-L1 siRNA further enhances the potency of LNG-pIC and even regresses established lung metastases for multiple tumor models. Due to the ease of manufacturing and good biocompatibility of our formulations, LNG holds great promise for treating different lung tumors and may inspire the development of more innovative strategies for cancer immunotherapy in the future.

J. Li reports grants from the National Natural Science Foundation of China, National Key Research and Development Program of China, Tsinghua University Initiative Scientific Research Program, Tsinghua-Peking Center for Life Sciences, and Tsinghua University Spring Breeze Fund during the conduct of the study; additionally, J. Li has a patent for preparation and application of inhalable liponanogels pending. L. Luo reports grants from the National Natural Science Foundation of China, National Key Research and Development Program of China, Tsinghua University Initiative Scientific Research Program, Tsinghua-Peking Center for Life Sciences, and Tsinghua University Spring Breeze Fund during the conduct of the study; in addition, L. Luo has a patent for preparation and application of inhalable liponanogels pending. J. He reports grants from the National Natural Science Foundation of China, the National Key Research and Development Program of China, Tsinghua University Initiative Scientific Research Program, Tsinghua-Peking Center for Life Sciences, and Tsinghua University Spring Breeze Fund during the conduct of the study. J. Yu reports grants from National Natural Science Foundation of China, National Key Research and Development Program of China, Tsinghua University Initiative Scientific Research Program, Tsinghua-Peking Center for Life Sciences, and Tsinghua University Spring Breeze Fund during the conduct of the study; in addition, J. Yu has a patent for preparation and application of liponanogels pending. X. Li reports grants from National Natural Science Foundation of China, National Key Research and Development Program of China, Tsinghua University Initiative Scientific Research Program, Tsinghua-Peking Center for Life Sciences, and Tsinghua University Spring Breeze Fund during the conduct of the study; in addition, X. Li has a patent for preparation and application of liponanogels pending. X. Shen reports grants from National Natural Science, National Key Research and Development Program of China, Tsinghua University Initiative Scientific Research Program, Tsinghua-Peking Center for Life Sciences, and Tsinghua University Spring Breeze Fund during the conduct of the study. J. Zhang reports grants from National Natural Science Foundation of China, National Key Research and Development Program of China, Tsinghua University Initiative Scientific Research Program, Tsinghua-Peking Center for Life Sciences, and Tsinghua University Spring Breeze Fund during the conduct of the study. J.M. Karp reports personal fees and other support from Corner Therapeutics and Edge Immune, and personal fees from Eterna Therapeutics during the conduct of the study; for grants, personal fees, and other support from multiple sources outside the submitted work, see https://www.karplab.net/team/jeff-karp; in addition, J.M. Karp has multiple patents in cancer therapy and cancer immunotherapy pending, issued, licensed, and with royalties paid. Companies that have licensed IP generated by J.M. Karp may benefit financially if the IP is further validated. The interests of J.M. Karp were reviewed and are subject to a management plan overseen by his institutions in accordance with its conflict of interest policies. R. Kuai reports grants from National Natural Science Foundation of China, National Key Research and Development Program of China, Tsinghua University Initiative Scientific Research Program, Tsinghua-Peking Center for Life Sciences, and Tsinghua University Spring Breeze Fund during the conduct of the study and other support from Corner Therapeutics outside the submitted work; in addition, R. Kuai has a patent for preparation and application of inhalable liponanogels pending. No disclosures were reported by the other authors.

J. Li: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. L. Luo: Formal analysis, validation, investigation, visualization, methodology, writing–review and editing. J. He: Formal analysis, validation, investigation, visualization, methodology, writing–review and editing. J. Yu: Validation, investigation, visualization, methodology. X. Li: Formal analysis, validation, investigation. X. Shen: Formal analysis, validation, investigation. J. Zhang: Formal analysis, visualization, writing–review and editing. S. Li: Formal analysis, visualization, writing–original draft. J.M. Karp: Formal analysis, writing–review and editing. R. Kuai: Conceptualization, supervision, funding acquisition, validation, investigation, writing–original draft, writing–review and editing.

The work was supported in part by grants from the National Natural Science Foundation of China (82173751 to R. Kuai; 32241031 and 32171195 to S. Li), National Key Research and Development Program of China 2023YFC3403100 to R. Kuai, Tsinghua University Initiative Scientific Research Program (2023Z11DSZ001 and 2022Z11QYJ036 to R. Kuai and 2023Z11DSZ001 to S. Li), start-up packages from Tsinghua University to R. Kuai and S. Li, support from Tsinghua-Peking Center for Life Sciences to R. Kuai and S. Li, support from the Key Laboratory of Innovative Drug Research and Evaluation to R. Kuai, and Tsinghua University Spring Breeze Fund to S. Li (2021Z99CFZ004).