Intrapleural nano-immunotherapy promotes innate and adaptive immune responses to enhance anti-PD-L1 therapy for malignant pleural effusion

Malignant pleural effusion (MPE) is indicative of terminal malignancy with a uniformly fatal prognosis. Often, two distinct compartments of tumour microenvironment, the effusion and disseminated pleural tumours, co-exist in the pleural cavity, presenting a major challenge for therapeutic interventions and drug delivery. Clinical evidence suggests that MPE comprises abundant tumour-associated myeloid cells with the tumour-promoting phenotype, impairing antitumour immunity. Here we developed a liposomal nanoparticle loaded with cyclic dinucleotide (LNP-CDN) for targeted activation of stimulators of interferon genes signalling in macrophages and dendritic cells and showed that, on intrapleural administration, they induce drastic changes in the transcriptional landscape in MPE, mitigating the immune cold MPE in both effusion and pleural tumours. Moreover, combination immunotherapy with blockade of programmed death ligand 1 potently reduced MPE volume and inhibited tumour growth not only in the pleural cavity but also in the lung parenchyma, conferring significantly prolonged survival of MPE-bearing mice. Furthermore, the LNP-CDN-induced immunological effects were also observed with clinical MPE samples, suggesting the potential of intrapleural LNP-CDN for clinical MPE immunotherapy. Malignant pleural effusion (MPE) is the terminal stage of cancer and the current standard of care for MPE is largely palliative. Here the authors design a liposomal nanoparticle loaded with cyclic dinucleotide for targeted activation of STING signalling in macrophages and dendritic cells and show that, on intrapleural administration, the nanoparticle effectively mitigates the immune cold MPE and significantly augments the checkpoint blockade immunotherapy in a mouse MPE model and clinical patients’ samples.

M alignant pleural effusion (MPE) secondary to metastatic cancer represents an enormous challenge in clinical patient management [1][2][3] . The appearance of MPE is an ominous prognostic sign for patients with cancer; the average survival of patients with MPE is 4-9 months 1-3 . Moreover, accumulation of pleural effusion commonly causes dyspnoea that severely compromises quality of life. The current standard of care treatment for MPE includes catheter drainage or chemical/surgical pleurodesis but is largely palliative 4,5 .
MPE is a build-up of extra fluid in the space between the lungs and chest wall, comprising tumour cells and various types of immune cells. Accompanying MPE, disseminated and unresectable tumour foci (carcinomatosis) often develop on the pleural surface. Clinical immunopathological studies suggest that the tumour microenvironment (TME) of MPE is profoundly immunosuppressive with abundant tumour-promoting myeloid immune cells and high levels of immunosuppressive cytokines, which negatively affects antitumour immunity 2,6-9 . Moreover, variable levels of programmed death ligand 1 (PD-L1) expression have been detected on tumour cells in clinical MPE [10][11][12] , and recent clinical trials with anti-PD-1/PD-L1 in patients with MPE have observed some meaningful antitumour activity [10][11][12][13][14] . Previous attempts to boost the anti-MPE immune response involved intrapleural administration of immunostimulants such as bacterial antigens, pro-inflammatory cytokines, oncolytic virus or adenoviral cytokine genes [15][16][17][18] . While variable degrees of efficacy have been reported, many of them were found either to overcome insufficiently the immunosuppressive TME or cause adverse effects 2,19 . Thus, developing a new intrapleural strategy that effectively converts the immune cold into proinflammatory MPE is imperative to improve MPE immunotherapy.
The stimulator of interferon genes (STING) pathway has recently been identified to play an important role in antitumour immunity 20,21 . As a potent STING agonist, the cyclic dinucleotide (CDN), 2'3'-cGAMP, functions in the cytosol to ligate STING and activate the STING pathway and type I interferon (IFN) production 22 . While recent preclinical studies involving intratumoral injection of CDN have demonstrated its ability to enhance antitumour immunity in solid tumours [23][24][25][26] , the potential of intrapleural CDN has not been explored. However, there are some concerns with the use of free CDN in MPE. Built from labile phosphodiester bonds, the CDN is susceptible to degradation by ecto-nucleotide pyrophosphatase/ phosphodiesterase (ENPP1) [27][28][29] . Soluble ENPP1 exists in human and mouse serum and higher levels of ENPP1 have been reported in malignant effusion 27,29,30 . Moreover, recent studies have shown stage for response to anti-PD-L1 immunotherapy against MPE in mouse models. Furthermore, MPE samples of patients with non-small cell lung cancer (NSCLC) were obtained to demonstrate that the immunological effects induced by LNP-CDN in mice could be reproduced in humans.

Intrapleurally injected LNP-CDN targets phagocytes in MPE.
We synthesized a series of LNPs with variable surface compositions of different phospholipids with/without polyethylene glycol (PEG), to load CDN complexed with CaP ( Supplementary Fig. 1). Among them, the PS-coated LNPs with/without DSPE-PEG2000 (10%; LNP-1 and LNP-2) exhibited excellent in vitro phagocyte targeting specificity and PS-mediated cell uptake ( Supplementary  Fig. 1). The PEGylated LNP-2 was stable over time with little CDN release at pH 7.4, but destabilized rapidly to release CDN at an acidic pH ( Supplementary Fig. 1). In vivo pharmacokinetic data revealed that LNP-2 achieved better MPE retention and less extrathoracic distribution than the non-PEGylated LNP-1 ( Supplementary Fig. 2). Therefore, LNP-2-based LNP-CDN was chosen to use in this study. LNP-CDN has an average diameter of ~120 nm and a negative surface charge of −15 mV (Fig. 1a). A mouse MPE model of Lewis lung cancer (LLC) developed both the fluid in the pleural cavity and multifocal tumours on the pleural surface, clearly seen by in vivo magnetic resonance imaging (MRI) and at autopsy (Fig. 1a). An in vivo imaging system (IVIS) at different times post-intrapleural injection of 1,1'-dioctadecyl-3,3,3','3'-tetramethylindotricarbocyanine iodide (DiR)-labelled LNP clearly showed that the majority of signals remained in the chest region and were sustained for at least 48 h (Fig.  1b,c). Ex vivo IVIS detected DiR-LNP signals in pleural tumours, tumour-draining lymph nodes (TDLNs), lung and liver (Fig. 1d), which were quantitated by high-performance liquid chromatography (HPLC) (Supplementary Fig. 2). There was minimal DiR-LNP in the blood and other major organs. Immunocytochemistry of the cells collected from MPE after intrapleural DiR-LNP revealed predominant uptake of LNP-CDN by CD11c + monocytes/macrophages and dendritic cells (DCs; Fig. 1e). Immunohistochemistry of pleural tumour tissues depicted well-dispersed LNPs that also colocalized well with CD11c + monocytes/macrophages and DCs (Fig. 1f). Consistently, our flow data showed that DiR-LNPs were captured primarily by macrophages, CD103 + DCs and CD11b + DCs, while minimal uptake was observed in tumour cells, T cells or natural killer (NK) cells in MPE, pleural tumours and TDLNs ( Fig.  1g-i). CD103 + DCs are considered as the most competent APCs to cross-prime CD8 + T cells [36][37][38] . Together, these data indicate that intrapleural LNP-CDN enables phagocyte-targeted delivery of CDN in both MPE and pleural tumours in the pleural cavity. Notably, LNP-CDN itself or carried by APCs can migrate from MPE to TDLNs, which probably potentiates APC-mediated cross-priming of cytotoxic T cells in TDLNs.

Intrapleural LNP-CDN reprograms myeloid immune cells.
To elucidate the immune landscape in MPE and its response to intrapleural LNP-CDN, we applied scRNA-seq without bias to characterize cell type-specific transcriptional profiles after intrapleural PBS, LNP-CDN, anti-PD-L1 antibody (Ab) or LNP-CDN + anti-PD-L1 Ab. About 4,000 single cells from the pooled MPE of three LLC MPE-bearing mice per condition were subjected to scRNA-seq. As depicted in t-distributed stochastic neighbour embedding (t-SNE) projection, unsupervised clustering singled out 20 distinct cell clusters, comprising tumour cells and various immune cells of monocytes/macrophages, neutrophils, DCs, NKs, CD4 + T cells and CD8 + T and B cells (Fig. 2a,b). Clearly, MPE contained a large number of myeloid cells including macrophages (Cd68 and Adgre1; ~55%) and neutrophils (Ly6g; ~30%). In-depth scRNA-seq clustering of the macrophage population yielded six subclusters (Fig. 2c). Mac_3, Mac_4 and Mac_6 were exclusively induced after LNP-CDN alone or in combination with anti-PD-L1 Ab (Fig. 2c-e), which showed significantly upregulated M1-associated genes ( Supplementary Fig.  3). By contrast, Mac_1 and Mac_2 that exhibited high expression of M2-associated genes were primarily found in the PBS control or anti-PD-L1 alone group, while Mac_5 was a mixture of cells from each treatment (Fig. 2c-e and Supplementary Fig. 3).
To investigate the transcriptome dynamics during macrophage repolarization, we applied the Monocle2 method for trajectory analysis. Monocle determines and orders single cells along a trajectory of two alternative cellular fates, namely the M2 and M1 fates ( Fig. 2f and Supplementary Fig. 4). Mac_1 and Mac_2 were located close to the M2 fate, while Mac_3 to Mac_6 were closer to the M1 fate (Fig. 2f). LNP-CDN alone or in combination clearly induced macrophage repolarization and gradual transition from the M2 to the M1 fate (Fig. 2g). Moreover, the trajectory heatmap revealed sequential gene expression changes (Fig. 2h). Clearly, gradually upregulated or downregulated expression of genes was shown during the switch from M2 to M1 ( Fig. 2i and Supplementary Fig. 4). Furthermore, we examined the enriched functions associated with these transitional genes by extracting functional gene ontology (GO) terms and biological pathways (Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways). We identified gene enrichment associated with antigen processing and presentation and the biological process related to inflammatory response during M2 to M1 repolarization (Fig. 2j).
To validate the scRNA-seq immunophenotyping, we conducted flow cytometry and found that LNP-CDN indeed skewed the M2 (F4/80 + arginase 1 + ) to M1-like macrophages (F4/80 + iNOS + ) in both MPE and pleural tumours (Fig. 2k,l). On the contrary, intrapleural free CDN had no immunological effects, which probably resulted from rapid degradation of CDN by ENPP1, which was detected at a high concentration of 7,200 pg ml −1 in MPE ( Supplementary Fig. 5). Intrapleural LNP-CDN plus anti-PD-L1 Ab was found to further enhance further the ratio of M1/M2 (Fig. 2k,l). An enzyme-linked immunosorbent assay (ELISA) also detected increased levels of pro-inflammatory cytokines and chemokines such as type I IFNs, interferon-γ (IFN-γ), interleukin (IL)-2, IL-12 and the IL-15-IL-15R complex in MPE after intrapleural LNP-CDN (Fig. 2m). Together, these data clearly indicate that the TME in MPE is profoundly immunosuppressive with enrichment of M2/N2-like

LNP-CDN promotes cytotoxic effector CD8 + T cells and NK cells.
In addition to its effects on myeloid cells, intrapleural LNP-CDN led to marked expansion of MPE-infiltrating CD8 + T cells ( Fig. 3a and Supplementary Figs. 6 and 7). scRNA-seq depicted five distinct subclusters of CD8 + T cells (Fig. 3b). CD8_3 and CD8_4 were predominantly induced after intrapleural LNP-CDN and to a lesser degree by anti-PD-L1 Ab (Fig. 3b,c). While both the subclusters exhibited upregulated expression of effector molecules (Ifng, Prf1, Gzmb and Klrg1) and costimulatory receptors (Cd28 and Icos), consistent with an activating effector phenotype, CD8_4 was found to have higher expression of cell cycle transcripts (Cdk1, Mki67 and Pcna) (Fig.  3d,e). CD8T_1, also largely expanded by LNP-CDN, presented a naive-like phenotype (Tcf7 + , Ccr7 + , Pdcd1 − ) but also expressed high levels of stem cell antigen 1 (Sca-1 or Ly6a), the haematopoietic stem cell engraftment genes Hoxa1 and Lpxn and memory markers such as Il2rb and Cxcr3 (Fig. 3d,e), probably associated with the stem-like memory CD8 + T phenotype [39][40][41] . Conversely, most T cells under PBS treatment were located in CD8_2 and CD8_5 (Fig. 3b,c), of which CD8_2 showed high expression of Ccr7, Il7r and Tcf7 but low expression of Cd44, probably indicative of naive-like T cells, while CD8_5 expressed high levels of Tox, Pdcd1, Lag3 and Ctla4 (Fig. 3d,e and Supplementary Fig. 8), commonly linked to T cell exhaustion 42,43 . We also applied the Monocle2 to CD8 + T cells and yielded trifurcating trajectories (Supplementary Fig. 9). Concurring with the t-SNE characterization, CD8T_1 and CD8T_2 were enriched in the memory/naive branch, CD8T_3 and CD8T_4 were largely located in the effector branch, while CD8T_5 was found primarily in the exhausted branch ( Supplementary Fig. 9). In response to LNP-CDN or LNP-CDN + anti-PD-L1, CD8 + T cells were found to accumulate on the effector trajectory with some on the memory/ naive trajectory ( Supplementary Fig 9). Similarly, bulk RNA-seq analysis showed that the combination immunotherapy enriched genes in multiple signalling pathways relevant to T cell proliferation and activation, differentiation and migration and T cell-mediated immunity and cytotoxicity (Supplementary Figs. 10 and 11).
Notably, LNP-CDN alone or combined with anti-PD-L1 Ab significantly increased the number of polyfunctional CD8 + T cells with both IFN-γ + and granzyme B (Gzmb) + in MPE and pleural tumours ( Fig. 3f,g and Supplementary Fig. 7). Utilizing the LLC-ovalbumin (OVA) MPE-bearing mice, our data showed that there was a significant increase in the uptake of exogenous OVA-fluorescein isothiocyanate (FITC) by CD103 + DCs ( Supplementary Fig. 12) and presentation of the SIINFEKL-MHC class I complex on CD103 + DCs in MPE, pleural tumours and TDLNs (Fig. 3h,i) post-intrapleural LNP-CDN. LNP-CDN also led to more than a tenfold increase in SIINFEKL tetramer + CD8 + T cells in MPE (Fig. 3j,k); a marked increase in SIINFEKL tetramer + CD8 + T cells was also seen in pleural tumours, TDLNs and spleen (Supplementary Figs. 7 and 13). In different cohorts of the MPE mice, sequential analysis of MPE from the same individuals showed that a single intrapleural dose of LNP-CDN or LNP-CDN + anti-PD-L1 was able to promote the pro-inflammatory TME and activate CD8 + T cells and their cytotoxic functions in MPE (Supplementary Figs. 14 and 15). Taken together, these data demonstrate that intrapleural LNP-CDN promotes APC sensing of tumour antigen (TA) and cross-priming of tumour-specific CD8 + T cells and expands the populations of both polyfunctional effector CD8 + T cells and stem-like memory CD8 + T cells in MPE.
Flow cytometry revealed that intrapleural LNP-CDN with/without anti-PD-L1 Ab promoted polyfunctional production of IFN-γ and Gzmb (Fig. 4f,g), and expression of the activating Fcγ receptor (FcγR), CD16, in NK cells (Fig. 4h,i and Supplementary Fig.  17). Moreover, an in vitro tumour cell killing assay showed that NK cells isolated from the MPE pretreated with LNP-CDN killed significantly more LLC cells compared to the PBS control (Fig. 4j). Adding anti-PD-L1 Ab led to even more deaths of LLC cells ( Fig.  4j and Supplementary Fig. 18). Besides its function to disrupt the PD-1-PD-L1 axis, the IgG2b isotype of anti-PD-L1, 10 F.9G2, is known to interact with mouse FcγRs to induce antibody-dependent cellular cytotoxicity (ADCC) against PD-L1 + tumour cells 46 . Indeed, anti-PD-L1 F(ab') 2 was found to have lower cytotoxicity in the above cytolysis study. Together, these data demonstrate that intrapleural LNP-CDN revamps the cytotoxic activity of NK cells, which is further enhanced by combining with anti-PD-L1 Ab through its blocking of PD-1/PD-L1 interaction and the ADCC effect.
Intrapleural LNP-CDN enhances anti-PD-L1 immunotherapy. It is known that PD-L1 is inducible by IFNs to serve as a counter-regulatory mechanism [47][48][49] . LNP-CDN was found to increase the expression of PD-L1 and MHC class I on tumour cells at the transcript and protein levels ( Supplementary Fig. 19). Thus, combining anti-PD-L1 Ab to counteract overexpressed PD-L1 may be a rational strategy to augment anticancer immunity. Anti-PD-L1 Ab is commonly administered systemically. To explore an alternative administration, we compared intra- Articles NaturE NaNOtECHNOLOgy pleural versus intraperitoneal (i.p.) injection of anti-PD-L1 Ab in this study. Pharmacokinetic data showed that the intrapleural approach achieved a tenfold and threefold higher pleu-ral concentration of anti-PD-L1 Ab at 3 h and 24 h, respectively ( Supplementary Fig. 18). Intrapleural anti-PD-L1 also led to significantly higher Ab concentrations in pleural tumours and TDLNs PBS Lag3 Tox  Fig. 18). Thus, we included intrapleural anti-PD-L1 Ab in our treatment study but with less than one-third of the i.p. dose (100 µg). Intrapleural LNP-CDN or anti-PD-L1 (30 µg) monotherapy delayed tumour growth and MPE progression; the combination treatment resulted in a further decrease in MPE volume and pleural tumour burden and significantly prolonged survival in wild-type but not STING −/− mice bearing LLC stably transfected with firefly luciferase (LLC-Luc) MPE (Fig. 5a-f and Supplementary Fig. 18). There was no significant difference in survival benefit between intrapleural and systemic anti-PD-L1 Ab (Fig. 5c). Similar antitumour immune responses and therapeutic efficacy of the combination immunotherapy were also observed in the CMT-167-Luc MPE model (Fig.  5g-j and Supplementary Figs. 20 and 21). Immunohistochemistry of pleural tumours revealed significantly more apoptotic cells induced by the combination treatment (Fig. 5k,l). Intriguingly, in contrast to the enlarged and distorted angiogenic vessels observed in pleural tumours with PBS control, LNP-CDN alone or in combination resulted in 'normalized' blood vessels (Fig. 5m-p). Through bulk RNA-seq, the combination treatment was found to upregulate the expression of vascular normalization/stabilization genes and anti-angiogenic genes, while downregulating pro-angiogenic genes ( Supplementary Figs. 10 and 11).
Because MPE is often associated with NSCLC, we created a mouse model with concomitant lung cancer and MPE and treated mice with intrapleural combination immunotherapy. Assessment of the treatment effects by IVIS and ex vivo examinations of lung tumour foci showed that combination immunotherapy significantly inhibited lung tumour growth ( Supplementary Fig. 22), indicating that antitumour immune responses are not confined to the pleural cavity but extended to the lung parenchyma. Moreover, long-term surviving MPE mice resisted secondary LLC tumour but not B6 melanoma challenge, suggesting that combination treatment triggers tumour-specific antitumour memory (Supplementary Fig. 23).
To determine if the observed antitumour immune responses were derived from the integrative efforts of both innate and adaptive immune cells, we conducted depletion studies and analysed the effects of loss of individual immune cells. Depletion of macrophages/DCs in MPE by LNP-clodronate (LNP-Clod) completely abrogated the treatment-induced cytotoxicity observed in both in vivo apoptosis assay and Kaplan-Meier survival study ( Supplementary Fig. 18). Sequential depletion studies revealed that depleting CD8 + T cells or NK cells starting 1 day before treatment or depleting CD8 + T cells 7 days after treatment largely abolished the antitumour effects of combination immunotherapy ( Supplementary  Fig. 18). However, loss of NK cells at the latter time had little impact Hcst  IL10  TGFB  CCL17  CCL18  ARG1  CD209A  IRF4  LYVE1  CD86  NOS2  IL12  TNF  IL23  CXCL10  CCL2  IL2  IFN1A  IFN1B  IL6  IL8  IL1B  IRF3  IL18  on animal survival. These data suggested the critical role of NK cells during the early phase of antitumour immunity. Together, these data indicate that the observed anticancer immunity is indeed macrophage/DC-initiated/mediated and collaborative efforts of the innate and adaptive effector lymphocytes are required to execute the antitumour activity. Intrapleural LNP-CDN and anti-PD-L1 Ab were well tolerated by the MPE mice. Safety studies showed no abnormality in blood liver enzyme levels or morphological changes in major organs from the short-term treated or the long-term surviving mice ( Supplementary  Fig. 24), supporting the notion that intrapleural LNP-CDN alone or in combination with anti-PD-L1 Ab is safe.
Immunological effects of LNP-CDN are confirmed in human MPE. To assess the translational potential of intrapleural LNP-CDN, we obtained clinical MPE samples (n = 5 patients) during thoracentesis of patients with NSCLC with confirmed MPE cytology (Fig.  6a). Flow cytometry of MPE identified tumour cells (~30%) and diverse immune cells (~60%) with a wide range of variation between individual patients (Fig. 6b). Among the immune cells, CD3 + T cells, monocytes/macrophages and neutrophils were more prevalent. Despite the high percentage of T lymphocytes, the fraction of CD8 + T cells was much lower than that of CD4 + T cells and a considerable number of CD4 + regulatory T cells were present (Fig. 6b). Monocytes/macrophages isolated from each patient's MPE exhibited homogeneously upregulated expression of M2-associated genes (Fig. 6c). ELISA detected low levels of IFNs and other pro-inflammatory cytokines/chemokines in MPE ( Supplementary  Fig. 25). These data are in good agreement with previous clinical observations, suggesting the immune cold MPE [50][51][52][53] .
To study the utility of LNP-CDN in human MPE, we first evaluated its targeting specificity by incubating DiR-LNP with fresh MPE or isolated macrophages from MPE ( Supplementary Fig. 26 and 27). Flow cytometry and immunofluorescence microscopy revealed the predominant uptake of DiR-LNP by monocytes/macrophages and DCs (Fig. 6d,e). We next investigated the ex vivo effects of LNP-CDN on various immune cells in MPE. Incubation of LNP-CDN with isolated macrophages induced increased expression of M1-associated genes in all five patient samples (Fig. 6c,f). We also isolated NK cells from individual samples and treated NK cells with the supernatant from the above LNP-CDN-treated macrophages ( Supplementary  Fig. 27). Flow cytometry detected a marked increase in expression of the activating receptor, NKG2D and production of intracellular IFN-γ (Fig. 6g,h). Moreover, incubation of LNP-CDN with fresh MPE led to a drastic increase in IFN-γ + CD8 + T cells in all five samples (Fig. 6i). Furthermore, the LNP-CDN-activated NK cells demonstrated enhanced cytolysis of autologous tumour cells (Fig. 6j). As variable levels of PD-L1 were expressed on pleural tumour cells in all samples (Fig. 6k), combining avelumab, the clinically approved IgG1 anti-PD-L1 monoclonal Ab, further increased NK cell cytotoxicity (Fig. 6j). Together, these data support the potential use of intrapleural LNP-CDN in human MPE.

Conclusions
Consistent with previous reports, our data showed that MPE in mice and humans was profoundly immunosuppressive with abundant tumour-promoting myeloid cells. Despite being a potent STING agonist, 2'3'-cGAMP CDN has a short biological half-life in MPE because MPE contains a high level of ENPP1 that degrades CDN. We developed the CDN-loaded liposomal nanoparticle LNP-CDN, which protected CDN from enzymatic degradation in MPE and exhibited its specific targeting of macrophages and DCs to activate STING signalling. Through scRNA-seq, intrapleural LNP-CDN was found to induce drastic changes in the transcriptional landscape in MPE by reprogramming myeloid cells, activating DCs for cross-presentation of TA, promoting polyfunctional NK cells and CD8 + T cells and expanding stem-like memory CD8 + T cells. We further showed that LNP-CDN-induced remodelling of the MPE TME set a stage for response to anti-PD-L1 ICB.
In the context of clinical practice, LNP-CDN can be administered serially via indwelling pleural catheters. Clinically, the presence of MPE often precludes surgical intervention and many patients with MPE are not fit for chemotherapy due to their extremely poor condition. Thus, successful management of MPE may renew opportunities for combining with other treatment options to maximize therapeutic efficacy.

online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41565-021-01032-w.

Preparation and characterization of LNP-CDN. The liposome LNP-CDN
with CaP core complexed with CDN (LNP) was prepared in two steps using the water-in-oil reverse microemulsion method 54 (Rhod-b) was added to the second lipid mixture with a molar ratio of 1%; alternatively, DiR was used for labelling the NPs by adding DiR directly to the second lipid mixture at a molar ratio of 5%. Loading of CDN without the use of CaP in liposomes (lipid mixture, 20 mM, brain-PS:DSPC:cholesterol:DSPE-PEG2000 = 5:4:1:1) was also prepared using the above thin-film hydration method.
The size, size distribution and zeta potential of LNP-CDN in aqueous solution were measured with a Malvern Zetasizer Nano ZS90 and analysed by Zetasizer (7.12). Transmission electron microscopy (TEM) measurements were performed on an FEI Tecnai Bio Twin transmission electron microscope.
The release of CDN from LNPs was assessed by the dialysis of LNP-CDN solution against release medium at different pHs (pH 7.4, 6.5 and 5.0) at various time points by Agilent 1100 HPLC. The LNPs were incubated in PBS (pH 7.4, 6.5 and 5.0; 0.01 M) with 10% fetal bovine serum (FBS) (v/v) at 37 °C for 12 h to study particle stability and pH-responsive ability. For the NSCLC MPE model, 6-8-week-old C57BL/6 mice (female:male at 1:1; Charles River Laboratories) were injected intrapleurally with LLC-Luc (2 × 10 5 ) or CMT167-Luc (2 × 10 5 ) lung cancer cells. STING knockout mice (B6(Cg)-Sting1tm1.2Camb/J, 6-8 weeks old, female:male at 1:1; The Jackson Laboratory) were also used for the MPE model. Development of MPE was monitored by bioluminescence imaging (BLI) or MRI. For the subcutaneous tumour model, 1 × 10 6 of LLC cells (left flank) and 5 × 10 5 B16 melanoma cells (right flank) were injected. Tumour volumes were measured longitudinally by a caliper up to 3 weeks. The maximum tumour burden in mice allowed by the IACUC of Wake Forest University School of Medicine is less than 20 mm in any one direction (WF IACUC SOP#19). We strictly followed the guidelines in this study.
In vivo bio-distribution of DiR-LNP. Mice with an established LLC-Luc MPE model were intrapleurally injected with LNPs labelled with DiR and longitudinally monitored at 1, 3, 24 and 48 h post-injection using the IVIS Lumina system. Randomly selected animals were killed at 24 and 48 h (n = 3 per time point) and major organs were dissected and imaged ex vivo by IVIS. MPE cells were collected and costained with anti-mouse CD11c-FITC (1:100 dilution; BioLegend, clone N418) and anti-luciferase (1:800 dilution; Sigma-Aldrich, catalogue no. L0159) followed by cy3-anti-rabbit secondary antibody (1:1,000 dilution; Jackson ImmunoResearch) and observed using a fluorescence microscope. DiR signals were recorded and merged with the CD11c image and the luciferase-stained image of the same field. Isolated pleural tumours were also preserved and sectioned for immunofluorescence microscopy as described above. DiR + cells in MPE, pleural tumours and DLNs were also assessed by flow cytometry. To quantify the tissue concentrations of LNPs, LLC-Luc mice with MPE were killed at 1, 2, 8, 24, 48 and 72 h after intrapleural injection of Rhod-b-labelled LNPs (n = 3 per time point). Major organs and blood were collected for HPLC analyses. Genes used for cell ordering were determined in an unsupervised way by their differential expression across cells. Dimensionality reduction and trajectory construction was performed on the selected genes with default methods and parameters, through the DDRTree method and orderCells function. Cells in less differentiation type, for example, M2 phenotype, informed us of the start point and initial state of the pseudotime. The visualization function plot_cell_trajectory was used to plot the minimum spanning tree on cells. The functions plot_genes_in_pseudotime and plot_genes_branched_ heatmap were used to visualize gene expression changes along the cell trajectory. Flow cytometry. LLC or CMT MPE mice were randomly grouped and treated on day 11 with intrapleural PBS, free CDN (1 μg), LNP-CDN (1 µg), anti-PD-L1 Ab (30 µg) or LNP-CDN plus anti-PD-L1 Ab. Mice were killed on day 13 and pleural fluid was gently aspirated using a 1 ml syringe through the diaphragm and its volume was measured with a 1 ml pipette. Solid tumours on the pleural surface and mediastinal lymph nodes were also collected and processed for analysis.
Flow cytometry was performed on a BD Canto II flow cytometer; data were collected with the FACSDiva software v.6.1.3 and analysed using the FlowJo software v.10.1 (FlowJo LLC). A list of antibodies used is summarized in Supplementary Table 1. All antibodies for mouse flow cytometry were used at a 1:100 dilution. Doublets and debris of dead cells were excluded before various gating strategies were applied ( Supplementary Figs. 28 and 29). Gates and quadrants were set based on isotype control staining and the mean fluorescence intensity (MFI) values were calculated by subtracting the MFI of the isotype control antibodies.

Corresponding author(s): Dawen Zhao
Last updated by author(s): Sep 22, 2021 Reporting Summary Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.
nature portfolio | reporting summary

March 2021
Data Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy The single cell RNA-seq data are available in the NCBI Gene expression Omnibus database (GEO) with accession number GSE164487. The bulk mRNA-seq data are available in the NCBI Gene expression Omnibus database (GEO) with accession number GSE179783. The authors declare that other data supporting the findings of this study are available within this article and its Supplementary Information, source data, and all additional data are available from the corresponding author on reasonable request.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Authentication
None of the cells were authenticated in the lab.

Mycoplasma contamination
The cells lines were tested negative for mycoplasma every two week.
Commonly misidentified lines (See ICLAC register) No commonly misidentified cell lines were used.

Wild animals
The study did not involve wild animals.
Field-collected samples The study did not involve samples collected from the field.

Ethics oversight
Institutional Animal Care and Use Committee at the Wake Forest University School of Medicine Note that full information on the approval of the study protocol must also be provided in the manuscript.

Human research participants
Policy information about studies involving human research participants

Population characteristics
Malignant pleural effusion samples (100 ml/patient) from 5 patients diagnosed with non-small cell lung cancer (NSCLC) were collected during thoracentesis and distributed by Tumor Tissue and Pathology Shared Resource (TTPSR) of the Wake Forest Baptist Medical Center Comprehensive Cancer Center (WFBMC-CCC). Acquisition of de-identified MPE samples from TTPSR for research use was in accordance with Institutional Review Board of Wake Forest University (IRB protocol # IRB00040151). All patients provided written informed consent.

Recruitment
Eligible patients were invited to participate by one of the investigators or a designee of one of the investigators. If they are willing to participate, the investigator or their designee reviewed the consent form and obtained informed consent. All patients with non-small cell lung cancer (NSCLC) with planned thoracentesis were consented to MPE collection protocol. Patients were approached in clinic by study personnel without relation to age, race, gender, disease stage, or prior therapies. Patients with MPE collected had no significant biases appeared to be present.

Ethics oversight
Institutional Review Board of Wake Forest University (IRB protocol # IRB00040151) Note that full information on the approval of the study protocol must also be provided in the manuscript.

Flow Cytometry
Plots Confirm that: The axis labels state the marker and fluorochrome used (e.g. CD4-FITC).
The axis scales are clearly visible. Include numbers along axes only for bottom left plot of group (a 'group' is an analysis of identical markers).
All plots are contour plots with outliers or pseudocolor plots.
A numerical value for number of cells or percentage (with statistics) is provided.