LSD1 inhibition sensitizes anti-PD1 blockade immunotherapy by inhibiting the long-range attack of tumor-derived extracellular vesicles

Abstract

Programmed cell death protein-1/programmed cell death ligand-1 (PD1/PD-L1) blockade has shown promise in cancer therapy but remains limited by low response rates. Recent efforts have explored strategies to enhance immunotherapy efficacy. Histone lysine-specific demethylase 1 (LSD1) inhibition can enhance tumor immune responses by downregulating PD-L1 expression. Furthermore, PD-L1 in tumor cell-derived extracellular vesicles (EVs) contributes to the immunosuppressive tumor microenvironment (TME) and promotes immune evasion. Here, we found that LSD1 inhibition can mediate the rearrangement of PD-L1 on tumor cell surfaces, reduce the secretion of EVs and PD-L1 levels in the TME, and ultimately block the long-range immunosuppression caused by tumor cell-released EVs. Therefore, we developed a TME-targeted synergistic therapy system with a dual mechanism in which anti-PD1 therapy blocks immune checkpoints, and forsythiaside A (FA) acts as an LSD1 inhibitor to regulate EVs secretion. Additionally, CD4⁺ T cells can directly kill tumor cells by inducing G1/S cell cycle arrest. Ultimately, this system activates the tumor immune response within the TME, effectively inhibiting the growth of non-small cell lung cancer tumors. Our work highlights the signaling role of EVs and the capacity of CD4⁺ T cells to arrest the cell cycle, offering a new approach to enhance response to anti-PD1/PD-L1 therapy.

Graphical abstract

This study highlights the critical role of LSD1 in tumor immunotherapy, underscores the function of EVs as key signaling hubs bridging tumors and T cells, and reveals the direct contribution of CD4⁺ T cells to tumor clearance.

A TME-targeted synergistic therapy system (FPNs) can reduce the secretion of extracellular vesicles (EVs), and ultimately inhibit the long-range attack of EVs, therefore sensitizing anti-PD1 Blockade Immunotherapy. (By Figdraw).

1. Introduction

Lung cancer, a highly heterogeneous disease, is the leading cause of cancer morbidity and mortality worldwide, with non-small cell lung cancer (NSCLC) accounting for approximately 85% of lung cancer cases^1,2. Among the commonly used clinical treatments, immune checkpoint blockade therapy involving anti-programmed cell death 1 (PD1) and anti-programmed cell death ligand 1 (PD-L1) antibodies is the best choice for first-line or follow-up treatment of NSCLC patients to reduce immune tolerance3, 4, 5, 6. For example, durvalumab⁷, pembrolizumab⁸, nivolumab⁹, and atezolizumab¹⁰ have been successfully used in the clinical treatment of NSCLC, and these drugs exert their therapeutic effects by blocking PD-L1 on tumor cells or PD1 on T cells to restore tumor immunity^11,12. However, only 15%–20% of patients respond to PD1/PD-L1 blockade due to acquired13, 14, 15, 16 and innate immune resistance^17,18 associated with PD-L1 expression. Studies have shown that PD-L1 expression in cancer cells and immune cells in the tumor microenvironment (TME) is adaptively upregulated in response to treatment¹⁹. Moreover, interferon-γ (IFN-γ) can induce the expression of PD-L1 in the TME and impair the immune function of T cells²⁰. Only PD-L1-positive or tumor-infiltrating lymphocyte (TIL)-positive tumors respond to PD1/PD-L1 blockade²¹. Therefore, recent studies have focused on directly regulating the expression of PD-L1 in tumor cells22, 23, 24, 25 or directly blocking the interactions between PD-L1 on the tumor cell surface and PD1 on the surface of T cells26, 27, 28 to increase the effectiveness of immune checkpoint therapy. However, the TME is a complex ecosystem, and its immunosuppressive nature facilitates rapid and uncontrollable tumor growth and formation²⁹. Therefore, interactions and communication among other cellular or noncellular components in the TME may also affect the efficacy of PD1/PD-L1 immune checkpoint therapy.

Extracellular vesicles (EVs) are membrane-surrounded vesicles that contain and release bioactive materials and play a role in cell-cell communication³⁰. EVs are secreted from cells to the TME³¹; contain large amounts of bioactive proteins, lipids, metabolites, RNA, and DNA obtained from parental cells32, 33, 34; and play pleiotropic roles in key processes in tumor progression, including the development of the TME³⁵, angiogenesis³⁶, immune evasion³⁷, tumor cell invasion and metastasis³⁸, and multidrug resistance³⁹. Previous studies have shown that the presence of PD-L1 in EVs can reduce T cell activity, promote tumor immune evasion, and lead to a decrease in the efficacy of anti-PD1 therapy40, 41, 42. However, the inhibition of tumor cell-derived EVs secretion can inhibit tumor growth in an immune-dependent manner^43,44. Therefore, reducing PD-L1 expression in EVs may increase the efficacy of PD1/PD-L1 blockade. Lysine-specific demethylase 1 (LSD1) is a histone demethylase that specifically removes mono- and di-methylated lysines of histone H3 lysine K4 (H3K4) with the CoREST complex and the androgen receptor, resulting in transcriptional modulation45, 46, 47. LSD1 is dysregulated in a variety of tumors and is strongly associated with malignant progression in these tumors. Studies have shown that LSD1 levels, which are strongly increased in gastric cancer tumors, are directly correlated with PD-L1 levels^48,49. LSD1 inhibition can elicit tumor cell-intrinsic innate immune activation, increase T cell infiltration, and increase tumor sensitivity to anti-PD1 therapy50, 51, 52. Therefore, LSD1 inhibition may alleviate the immunosuppressive TME and increase the efficacy of anti-PD1/PD-L1 therapy by modulating EVs-associated PD-L1 secretion from tumor cells.

In the past decade, the discovery of LSD1 inhibitors has resulted in significant clinical progress, with 10 compounds entering clinical research, including 8 irreversible inhibitors based on six trans-2-phenylcyclopropylamine (TCP), such as TCP, ORY-1001, GSK-2879552, INCB059872, IMG7289, ORY-2001, TAK-418 and LH-180253, 54, 55, and two reversible inhibitors, CC-90011 and SP-2577⁵⁶. The irreversible inhibitors are usually structural modifications of TCP, which inactivates LSD1 through a single electron transfer mechanism and then results in four different TCP-FAD adducts through cyclopropyl ring cleavage via different pathways⁵⁷. These TCP-FAD adducts occupy the FAD binding pocket and irreversibly inhibit LSD1 activity^58,59. However, owing to the severe toxicity caused by covalent binding with FAD, some clinical trials have been suspended. To reduce toxicity, in recent years, research has focused on reversible LSD1 inhibitors, but only two compounds have entered human clinical trials. Currently, there is a lack of research on novel scaffolds and highly efficient compounds that can directly bind to LSD1 to inhibit its activity.

Here, we investigated whether NSCLC cell-derived EVs regulate immune escape and the effectiveness of anti-PD1 therapy by regulating PD-L1 signaling communication by using LSD1. CD4⁺ T cells can directly kill tumor cells by arresting the cell cycle. Therefore, a TME-targeted synergistic therapy system, PEOz-FA-anti-PD1 (FPNs), with dual mechanisms for treating NSCLC was designed (Scheme 1). In this system, anti-PD1 was loaded into liposomes to achieve immune checkpoint therapy. Forsythiaside A (FA), a natural product extracted from Forsythia suspensa (Thunb.) Vahl, was loaded to inhibit LSD1 activity. Forsythiae Fructus (“Lianqiao” in traditional Chinese medicine) is an herb that has been widely utilized in the clinic for thousands of years with the efficacy of clearing heat, detoxifying, and reducing swelling, and is often used to treat inflammation, fever, and infectious diseases^60,61. After LSD1 demethylation was inhibited, the secretion of EVs by tumor cells and tumor cell expression of PD-L1 were downregulated via an EVs-mediated cell surface PD-L1 rearrangement mechanism. Moreover, this system promoted CD4⁺ T cell activation, which ultimately inhibited long-range attack by tumor cell-derived EVs. This synergistic therapeutic approach not only utilizes anti-PD1 blockade immune checkpoints but also encapsulates structurally labile FA in a phospholipid bilayer to maintain drug stability. In vitro and in vivo experiments demonstrate that FPNs have strong antitumor activity and low systemic toxicity. This synergistic therapy system has dual functions in immunotherapy, and CD4⁺ T cells gain increased capacity to kill tumor cells by arresting the cell cycle at the G1/S phase, which may provide an alternative strategy for sensitizing tumors to PD1/PD-L1 immune checkpoint therapy.

Scheme 1.

The tumor microenvironment (TME)-targeted synergistic therapy system constructed by anti-PD1 and LSD1 inhibitor forsythiaside A (FA) exerts anti-tumor ability. Anti-PD1 is an immune checkpoint inhibitor; FA mediates the PD-L1 rearrangement mechanism on the surface of tumor cells by inhibiting the demethylation activity of LSD1, regulating the recovery of recycling endosomes and the efflux function of multivesicular bodies (MVBs), reducing the secretion of extracellular vesicles (EVs) and PD-L1 levels in the TME, and ultimately inhibiting the long-range attack of EVs. Meanwhile, CD4⁺ T cells have been found to directly exert tumor-killing function by inhibiting the cell G1/S cycle.

2. Materials and methods

2.1. Reagents

Phosphatidylcholine (SPC) and cholesterol (CHO) were purchased from Aladdin Reagent (Shanghai, China). DSPE-PEOz2000 (R-PL2056-2k) was purchased from Ruixi Bio-Technology (Xi’an, China). Forsythiaside A (DL004901) was purchased from Desite Bio-Technology (Chengdu, China). InVivoMAb anti-mouse PD-1 (CD279) (BE0146) was obtained from Bio X Cell (Lebanon, NH, USA). The antibodies used in this study are shown in Supporting Information Table S1. Human lung cancer cell lines A549, H460, H1299, H226, and H1975, as well as mouse lung cancer cell lines LLC and LA795, were purchased from Wuhan Pricella Biotechnology Company (Wuhan, China). All cells were identified by short tandem repeats (STRs). Human CD4⁺ T cells, human PBMC, and mouse CD4⁺ T cells were purchased from Beijing Huizhi Heyuan Company (Beijing, China). All cells were spiked with 10% (v/v) fetal bovine serum (FSP500, ExCell, Suzhou, China) in Roswell Park Memorial Institute 1640 (RPMI-1640) (8117072, Gibco, Thermo Fisher Scientific, Waltham, MA, USA) or Dulbecco’s modified Eagle’s medium (DMEM) (6124134, Gibco, Thermo Fisher Scientific, Waltham, MA, USA) as required by the instructions. All cells were cultured in an incubator at 37 °C, 5% CO₂.

Paraffin sections derived from paired lung cancer and adjacent normal tissues from lung cancer patients were collected from the Department of Oncology, Xin Hua Hospital Affiliated with Shanghai Jiao Tong University School of Medicine (Shanghai, China). The collection and use of these clinical samples were approved by the ethical committee of Xin Hua Hospital Affiliated with Shanghai Jiao Tong University School of Medicine (Approval No. XHEC-D-2025-117). Donors had provided broad written consent permitting.

2.2. CRISPR/Cas9-mediated LSD1 knockout

a) Plasmid Preparation. The plasmid vector was linearized using the restriction enzyme BsmbI of NEB to generate cohesive ends. Subsequently, the target sgRNA sequences were ligated into the linearized lentiCRISPR v2 vector using T4 DNA ligase. Following ligation, monoclonal colonies were isolated and validated through sequencing analysis. Positive clones were inoculated into Luria-Bertani (LB) medium supplemented with 50 mg/L ampicillin and incubated at 37 °C with agitation at 200 rpm for approximately 18 h (Minquan Instrument Co., Ltd., MQT-60R, Shanghai, China). Bacterial cells were harvested, and plasmids were extracted utilizing an endotoxin-free plasmid mini-prep kit (Tiangen Biotechnology, DP118, Beijing, China). b) Lentivirus Packaging and Quantification. 293FT cells were cultured and transfected according to the manufacturer’s guidelines provided by Lipo293™ transfection reagent (Beyotime Biotechnology, C0521, Shanghai, China). Plasmid DNA was combined with the transfection reagent and introduced into cells. The culture medium was replaced, and subsequently, the supernatant containing lentiviral particles was harvested and filtered through a 0.45 μm sterile membrane. Lentiviral particles were concentrated and purified using PEG Lentivirus Purification Reagent (Inovogen Technology, P1201, Beijing, China). The viral precipitates were resuspended in PBS, aliquoted at 50 μL per tube, and stored at −80 °C until further use. c) Cell Culture and Lentiviral Infection. Cells were seeded at a density of 1 × 10⁵ cells per well into 24-well culture plates and subsequently infected with lentivirus at the appropriate multiplicity of infection (MOI). After a 24 h incubation period, the medium was replaced. Two days after antibiotic selection, genomic DNA was extracted, amplified by PCR (Roche Diagnostics, LightCycler® 480, Indianapolis, IN, USA), and sequenced to confirm genomic modifications. For monoclonal isolation, cells were diluted to a concentration of 1—1.5 cells per 100 μL, seeded into 96-well plates, and cultured for approximately one week. Monoclonal colonies were identified under the microscope, marked, and expanded following digestion and passaging. d) Validation of Gene Knockout. Cell suspensions were centrifuged at 1500 rpm for 5 min (Thermo Fisher Scientific, Sorvall ST 40R, Waltham, MA, USA), and the supernatant was removed. Cell pellets were lysed, and genomic DNA was extracted. PCR amplification was performed using primers specific for sequencing purposes. PCR products were analyzed by gel electrophoresis, and target bands were excised, purified, and subjected to sequencing analysis. Sequencing results were compared with the reference genome to confirm successful gene knockout. The sequences of sgLSD1 used to construct the LSD1 KO cell line are shown in Supporting Information Table S2. Knockout efficiency was further validated by Western blot analysis (detailed antibody information is listed in Table S1).

2.3. Preparation and purification of recombinant proteins

The LSD1 cDNA fragment (157–852AA) was cloned into the pET-28b plasmid, added to BL21(DE3) E. coli, and then screened and amplified by kanamycin. The recombinant protein was expressed by adding 0.25 mmol/L isopropyl β-d-thiogalactoside (IPTG) and incubating on a shaker for 12–16 h (16 °C, 160 rpm). (Minquan Instrument Co., Ltd.). The precipitate was collected by centrifugation; the bacteria were lysed by ultrasonic crusher; the protein solution after lysis was centrifuged for 50 min (4 °C, 8000 rpm) (Xiangyi Instrument Co., Ltd., CH160R, Changzhou, China); the precipitate was discarded, and the supernatant was passed through a 0.22 μm filter membrane. Finally, the recombinant LSD1 protein was purified by Ni-NTA resin affinity chromatography (C610030, Sangon Biotech, Shanghai, China), and the crude protein samples were concentrated and desalted by repeated low-temperature centrifugation using a 30 kDa ultrafiltration tube.

2.4. LSD1 enzyme inhibition assay

The LSD1 enzyme inhibition assay consisted of 62.5 μL of HEPES buffer (50 mmol/L pH 7.5), small molecule compound or an equal volume of DMSO, 10 μg of recombinant LSD1 protein, 25 μmol/L H3K4me2 peptide (HY-P2256, MedChemExpress LLC, Shanghai, China), 5 U/mL horseradish peroxidase (A002559, Sangon Biotech, Shanghai, China), and 10 μmol/L Amplex red (ST010, Beyotime Biotechnology, Shanghai, China). Incubated at 37 °C for 30 min. A blank control without substrate was set up to eliminate false positive results, and the fluorescence intensity of the sample was measured using a fluorescence microplate reader (Molecular Devices, LLC, San Jose, CA, USA) at an excitation/emission wavelength (λ_ex/λ_em) of 530 nm/590 nm to assess the inhibition of the molecule.

2.5. Cellular thermal shift assay (CETSA)

The cell lysate or protein solution was divided into eight equal parts. Compounds or DMSO (as the control) at equal concentration and volume were added to each aliquot, followed by incubation at 37 °C for 1 h. Samples were then heated for 5 min at 37, 42, 47, or 52 °C. After heating, 5 × SDS sample buffer was added, and the mixtures were vortexed thoroughly and boiled for 5 min. The protein levels were determined by western blotting.

2.6. Drug affinity responsive target stability (DARTS)

Cell lysates or proteins were divided into quartile aliquots and incubated with compounds or equal volumes of DMSO for 1 h at 37 °C and 300 rpm (Hangzhou Allsheng Instruments Co., Ltd., MB100-2A, Hangzhou, China), respectively. Pronase E (HY-114158, MedChemExpress, LLC, Monmouth Junction, NJ, USA) was diluted in 1 × TNC buffer and added to the above mixtures. Adjust the ratio of Pronase E to the total protein of the sample to the appropriate concentration. In the control group, an equal volume of 1 × TNC buffer was added, incubated for 30 min (37 °C, 300 rpm) (Hangzhou Allsheng Instruments Co., Ltd.), and protease inhibitors were added to terminate digestion. Following sample preparation, the protein levels were determined by Western blotting.

2.7. Surface plasmon resonance (SPR) assay

A Biacore T200 (GE Healthcare, Chicago, IL, USA) instrument was used to measure binding affinity at 25 °C. LSD1 was purchased from OriGene (Wuxi OriGene Biotechnology Co., Ltd., TP330555, Wuxi, China). Briefly, FA was diluted to the specified concentration with a running buffer (10 μmol/L HEPES plus 0.1% DMSO). To determine the optimal pH for LSD1 immobilization, pH scouting was performed using 10 mmol/L acetate buffer at pH 4.0, 4.5, 5.0, and 5.5, following pre-enrichment of the LSD1 protein. With the use of the standard amine coupling method, –8000 response units (RU) of LSD1 were immobilized on a CM5 chip. Finally, an LSD1-immobilized flow cell was injected with FA at concentrations of 0.39, 0.78, 1.56, 3.125, 6.25, 12.5, 25, 50, and 100 μmol/L at a flow rate of 30 μL/min for 90 s and allowed to dissociate for 90 s with running buffer serving as the blank control. The Biacore T200 evaluation software was used to calculate the steady-state KD value.

2.8. Reversible experiment assay

The experimental group was supplemented with a final concentration of 20-fold IC₅₀ and LSD1 protein, and the blank group and 100% activity control group were added with the same volume of DMSO and LSD1 protein, and incubated for 30 min (37 °C, 300 rpm, Hangzhou Allsheng Instruments Co., Ltd.) as the detection system before dilution. Samples were drawn from the pre-diluted samples after incubation and added to new sample wells as a diluted detection system and incubated again. The inhibition rate was measured and calculated for the pre-dilution and post-dilution systems.

2.9. Competitive experiment assay

The experiments were grouped according to the IC₅₀ of FA. The concentration of FAD was varied while H3K4me2 was fixed, or vice versa, and the mixtures were incubated at 37 °C with shaking at 300 rpm for 30 min (Hangzhou Allsheng Instruments Co., Ltd.). After the fluorescence microplate reader (Molecular Devices, LLC) was read, the data were processed using GraphPad Prism, and the data were analyzed to obtain K_m and V_max and plotted for analysis.

2.10. Mass spectrometry identification of protein compound modification sites

The gel was first subjected to trypsinization and peptide extraction procedures. The peptide samples were then performed on a nano-HPLC chromatography system (NanoElute, Bruker Daltonics, Billerica, MA, USA) connected to a hybrid trapped ion mobility spectrometry quadrupole time-of-flight mass spectrometer (TIMS-TOF Pro2, Bruker Daltonics, Billerica, MA, USA) via a CaptiveSpray nano-electrospray ion source. Peptide samples were loaded onto the analytical column (75 μm i.d. × 25 cm) for gradient elution. For MS analysis, the accumulation and ramp time were set as 100 ms each. Survey full-scan MS spectra (m/z 100–1700) were obtained in positive electrospray mode. The ion mobility was scanned from 0.7 to 1.3 Vs/cm². The overall acquisition cycle of 1.16 s comprised one full TIMS-MS scan and 10 parallel accumulation-serial fragmentation (PASEF) MS/MS scans. During PASEF MS/MS scanning, the collision energy was ramped linearly as a function of the mobility from 59 eV at 1/K₀ = 1.6 Vs/cm² to 20 eV at 1/K₀ = 0.6 Vs/cm². MS files were processed using Fragpipe (v21.1) for protein/peptide identification.

2.11. Preparation of LSD1^C360A recombinant protein

The primer sequences of the LSD1^C360A mutation were shown in Supporting Information Table S3. Then, the LSD1^C360A cDNA fragment was spliced into pET-28b plasmid. The LSD1^C360A mutation was amplified using BL21 (DE3) strain and finally obtained by Ni-NTA resin affinity chromatography.

2.12. Chemistry

All reagents were purchased from Sigma–Aldrich, Energy Chemical, Aladdin, or Macklin. ¹H and ¹³C NMR spectra were recorded on Bruker instruments at 500 or 600 MHz at ambient temperature with methanol-d₄ as solvent and tetramethylsilane (TMS) as the internal standard. Chemical shifts (δ) were reported in parts per million (ppm) and coupling constants (J) in Hertz (Hz), respectively. High-resolution electrospray ionization (HRESI) mass spectra were acquired on an Agilent 6520B Q-TOF mass spectrometer (Agilent Technologies, Santa Clara, CA, USA). High-performance liquid chromatography (HPLC) was performed on the Agilent 1260 HPLC system using an Agilent ZORBAX SB-C₁₈ (4.6 mm × 250 mm, 5 μm) column, with methanol and H₂O as mobile phase; flow rate, 1 mL/min; and UV absorbance monitored at 254 and 330 nm.

2.13. Formulation and physicochemical characterization of FPNs

Specifically, the thin film hydration method was used to prepare FPNs. Briefly, an equilibrium mixture of lecithin, cholesterol, FA, and DSPE-PEOz at a weight ratio of 16:1:0.7:1 was dissolved in methanol solution, and the mixture solution was rotary evaporated at 37 °C to form a lipid film on the bottle wall. The dried lipid film was rehydrated with PBS (pH 7.40) containing 0.3 mg of anti-PD1 while shaking at room temperature. The suspension was sonicated for an additional 20 min at 4 °C, and subsequently, the unentrapped compounds were removed by centrifugation at 12,000 rpm for 1 h (Eppendorf AG, 5418 R, Hamburg, Germany). After 12 squeezes through a 0.22 μm truncated polycarbonate membrane (Avanti Polar Lipids, LLC, 610000, Alabaster, AL, USA), the liposomes were collected. The FPNs were obtained and stored at 4 °C.

The transmission electron microscopy (TEM, Hitachi High-Tech Corporation, HT7700, Tokyo, Japan) was used to characterize the morphology of FPNs. A nanoparticle tracking analyzer (Particle Metrix GmbH, ZetaView, Inning am Ammersee, Germany) was used to determine particle size, and a Zetasizer Nano ZS9 (Malvern Panalytical, Malvern, UK) was used to determine ζ-potential.

In order to calculate the encapsulation efficiency (EE) of the drug, the actual encapsulated drug was obtained by adding FPNs to methanol and disrupting the emulsion through ultrasonication. The content of FA was determined by ultraviolet spectrophotometry. The EE was calculated as shown in Eq. (1):

EE (%) = actual encapsulated drug amount in FPNs/drug amount × 100 (1)

At the same time, Triton X-100 was used to break down the phospholipid bilayer, and the content of anti-PD1 was quantitatively determined by BCA method. The specific steps were as follows: FPNs and liposomes (without anti-PD1) were prepared by thin film hydration method, separately. Triton X-100 was added, followed by incubation at room temperature for 20 min. The protein content was then determined using the BCA assay. The function of liposomes is to eliminate the influence of phospholipids and cholesterol on BCA quantification, as shown in Eq. (2):

EE (%) = (m_{post-demulsification proteins}−m_{unbroken protein}−m_{empty liposomes before and after demulsification})/m_{input protein amount} × 100 (2)

2.14. Drug release and target study

The FA and anti-PD1 release profile from FPNs was measured by the dialysis method. Two milliliters of FPNs were soaked in 20 mL of PBS containing 0.1% Tween 80. The samples under PBS at pH 6.30 and pH 7.40 were collected at the indicated time points and analyzed with a microplate reader at wavelengths of 330 nm and 562 nm, and the release rate of the drug was estimated.

LLC lung cancer cells were cultured in DMEM supplemented with 10% FBS in a humidified incubator containing 5% CO₂. PD1 protein was used to replace anti-PD1. LLC cells were seeded in 96-well plates at a density of 8 × 10³ cells per well, cultured for 24 h, incubated with FPNs (PD1 protein) for 1, 12, and 24 h, respectively. The PD1 protein was labeled with CoraLite 594 fluorescent secondary antibody; the PD-L1 protein in cells was labeled with CoraLite 488 fluorescent secondary antibody; and the nuclei were labeled with DAPI. Then, the fluorescence intensity at different times was statistically analyzed and co-localized using an ImageXpress Micro Confocal High-Content Imaging System (HCS, Molecular Devices, LLC, San Jose, CA, USA). LLC cells were seeded in 6-well plates at a density of 2 × 10⁵ cells per well, cultured for 24 h, and incubated with FPNs (PD1 protein) for 1, 6, 12, and 24 h, respectively. After incubation, cells were collected and dyed using FITC-labeled PD1 flow cytometry antibody (E-AB-F1131C, Elabscience, Wuhan, China) for flow cytometry detection by CytoFLEX S Flow Cytometer (Beckman Coulter, Inc., Brea, CA, USA).

2.15. HPLC–MS/MS detection of FA content in tumor site

After drug treatment, tumor-bearing mice were euthanized by cervical dislocation at 4, 6, 8, 10, 12, 16, 20, 24, 30, 36, and 40 h. The tumor site was removed, weighed, and 40 mg of the tumor was homogenized in a centrifuge tube containing a methanol solution. The homogenates were centrifuged at 12,000 rpm for 20 min at 4 °C (Eppendorf AG). The resulting supernatants were evaporated to dryness under nitrogen and redissolved in methanol for subsequent HPLC–MS/MS analysis.

A liquid chromatography-triple quadrupole mass spectrometer (LC-TQ) system consisting of 1260 Infinity III and G6475A TQ Mass Spectrometer (Agilent, Agilent Technologies, Santa Clara, CA, USA) was used to detect the concentration of FA. The chromatographic separation was performed using a ZORBAX SB-C₁₈ column (4.6 mm × 150 mm, 5 μm, Agilent, Agilent Technologies, Santa Clara, CA, USA) with a column oven maintained at 35 °C. The mobile phase consisted of water containing 0.1% formic acid (A) and methanol (B). The elution procedure was 0–20 min, 80% A–40% A, 20–22 min, 40% A–0% A. The flow rate was set to 0.6 mL/min. The electrospray ionization (ESI) source was operated in positive mode. Mass spectrum parameters were optimized as follows: Gas Temperature (°C), 300, Gas Flow (L/min), 13, Capillary Voltage (V), 4000, Sheath Gas Temperature (°C), 250, Sheath Gas Flow (L/min), 11. Quantification was performed using Selective Ion Detection Mode (SIM) with a molecular weight of 647.3 m/z. Agilent Mass-Hunter Quantitative Analysis 12.0 was used to process the obtained data.

2.16. Cell cytotoxicity assay

CCK-8 was selected for cytotoxicity assay. 5000 cells were seeded into each well of a 96-well microplate and incubated overnight in an incubator (5% CO₂) at 37 °C. Various concentrations of control FA and FPNs were administered to the cells, respectively. The concentrations of each group were at equivalent FA concentrations of 0–150 μmol/L. CD4⁺ T cells with a 1:4 effector-to-target (E:T) ratio are also added to related experiments. Each well contained 10 μL of CCK-8 reagent, which was added after 24 h of treatment, and the cells were then incubated at 37 °C for 2 h. Lastly, a microplate reader was used to measure the samples at 450 nm.

2.17. Tumor models

All animal experiments were conducted in compliance with the rules of the Animal Ethics Committee of China Pharmaceutical University (Approval No. 2024-05-014), and all experimental procedures were executed according to the Standards for Experimental Animal Care of the National Institutes of Health. LLC cells, LLC KO cells and LA795 cells (5 × 10⁶/100 μL PBS per mouse) were subcutaneously injected into healthy female C57BL/6J, BALB/c, or BALB/c nude mice (5 weeks old) (GemPharmatech Co., Ltd., Nanjing, China) via the right lower leg. The growth of tumor in mice was monitored every other day, upon reaching a tumor size of approximately 100 mm³. The animals were randomly assigned to different groups, and administration was conducted every other day. Tumor volume was calculated according to Eq. (3):

V = 0.5 × Maximal diameter × Perpendicular diameter² (3)

When the tumor volume of any mouse reached more than 1000 mm³, the mice were euthanized and follow-up experiments were carried out.

2.18. In vivo biodistribution

To evaluate the tumor-targeting ability of FPNs, we prepared liposomes with Cy5.5 fluorescent dye as a substitute for FA and anti-PD1. After intravenously injecting free Cy5.5 and Cy5.5-FPNs (equivalent Cy5.5 concentration of 3 mg/kg) into LLC tumor-bearing mice via the tail vein, a Tanon ABL X6 system (Tanon, Shanghai, China) was used to image mice (λ_ex/λ_em = 680 nm/704 nm) following anesthesia at various periods (2, 4, 6, 8, 12, and 24 h). After 24 h, the mice were euthanized at the end of imaging, and their tissues (containing the heart, liver, spleen, lung, kidney, and tumor) were isolated and scanned. Lastly, Living Image 4.0 Software was used for all image analyses.

2.19. Safety evaluation

The hemolytic activity of FPNs was evaluated using a hemolysis assay. FPNs (6.25, 12.5, 25, and 50 μmol/L) were incubated with a 2% suspension of mouse erythrocytes at 37 °C for 1 h, while the negative and positive controls used were PBS and ultrapure water-treated erythrocyte suspensions, respectively. After being centrifuged for 15 min at 3000 rpm (Eppendorf AG), the supernatant was collected, and an enzyme marker was applied to measure the absorption at 541 nm. The hemolysis ratio (HR) was calculated as shown in Eq. (4):

HR (%) = (OD_sample−OD_{negative control})/(OD_positive−OD_{negative control}) × 100 (4)

HR of less than 5% was regarded as nontoxic. After consecutive administration of FPNs, the blood and major organs of the mice were harvested for histological analysis and blood biochemical analysis.

2.20. Extraction and characterization of EVs

EVs were extracted according to the kit (41205ES20, Yeasen Biotechnology, Shanghai, China). When the confluence of the cells was about 60%–70%, the medium was replaced with serum-free medium (41210ES76, Yeasen Biotechnology, Shanghai, China). After continuing to culture to 90% confluency, the cell culture supernatant was collected for EVs extraction. The culture medium was centrifuged at 3000 × g for 10 min at 4 °C (Thermo Fisher Scientific), and the supernatant was collected. An equal volume of EVs isolation reagent was added, vortexed briefly, and the mixture was left to stand at 4 °C. Subsequently, the samples were centrifuged at 10,000 × g for 60 min at 4 °C to pellet the EVs (Thermo Fisher Scientific). The resulting pellet was washed with PBS and stored at −80 °C for long-term preservation. The morphology and particle size of EVs were determined by HT7700 TEM (Hitachi High-Tech Corporation) and ZetaView nanoparticle tracking analysis system (Particle Metrix GmbH), respectively.

2.21. Immunofluorescence staining

Cells were seeded in 96-well plates at a density of 8000 per well and cultured at 37 °C for 24 h for subsequent drug or co-culture treatments. After these, cells were fixed with 4% paraformaldehyde. Subsequently, permeabilization was performed with 0.3% Triton X-100, followed by blocking with 10% BSA in PBS. Then, coverslips were incubated with primary antibody diluted in 10% BSA at 4 °C overnight. After washing three times with PBS, the samples were incubated with fluorescent dye-conjugated secondary antibody for 1h. Furthermore, cells were washed with PBS and stained with DAPI for 15 min. The immunofluorescence images were captured using HCS system (Molecular Devices, LLC).

2.22. Quantitative real-time PCR (qRT-PCR)

Total RNA was isolated using the RNA-easy Isolation Reagent (R701-01, Vazyme Biotech Co., Ltd., Nanjing, China), and RNA concentration was measured by NanoDrop 2000 micro-ultraviolet spectrophotometer (Thermo Fisher Scientific, Inc., 1011U, Waltham, MA, USA). 1 μg RNA was used for reverse transcription to synthesize cDNA with HiScript III RT SuperMix for qPCR (R323-01, Vazyme Biotech Co., Ltd., Nanjing, China), and cDNAs were amplified with ChamQ Universal SYBR qPCR Master Mix (Q711-02, Vazyme Biotech Co., Ltd., Nanjing, China). The relative expression levels of PD-L1 mRNA were normalized to those of the internal control GAPDH using the 2^−ΔΔCt cycle threshold method. The primer sequences of these genes are presented in Supporting Information Table S4.

2.23. Western blot analysis

Cells were thoroughly lysed using RIPA lysis buffer supplemented with 1 mmol/L PMSF and centrifuged to obtain total protein. The concentration of the lysates was determined using the BCA Protein Quantification Kit (Thermo Fisher Scientific, 23227, Waltham, MA, USA). Protein was diluted with 5 × SDS buffer, subjected to heating for 10 min at 100 °C, loaded onto 10% SDS-PAGE, transferred onto PVDF membranes (Bio-Rad Laboratories, Inc., 1620177, Hercules, CA, USA), and subsequently blocked with 5% skim milk in TBST. Then the membranes were probed with corresponding primary antibodies (LSD1, H3K4me1, H3K4me2, PD-L1, TSG101, RAB11, CD9, ALIX, CyclinD1, CDK4, CDK6, GAPDH, and β-actin). After washing, the membranes were incubated with horseradish peroxidase (HRP) conjugated secondary antibody for 2 h at room temperature. Chemiluminescence detection was performed by a very ultrasensitive (femtogram) ECL luminescent substrate kit (U10012, UU-Bio Technology Co., Ltd., Suzhou, China). Detection of immunoblots was captured using a Chemi DOCTM XRS⁺ system (Bio-Rad Laboratories, Inc., Hercules, CA, USA). GAPDH or β-actin was used as a loading control.

2.24. Cell cycle assay

A cell cycle assay after LSD1 knockdown and drug or CD4⁺ T cells treatment was measured by flow cytometry using propidium iodide (PI) staining. Briefly, cells were incubated for 24 h in the respective media. Thereafter, cells were collected and fixed in 70% pre-chilled ethanol at 4 °C overnight and then stained with PI/RNase Staining Buffer (Becton, Dickinson and Company, 550822, Franklin Lakes, NJ, USA) and analyzed by FACS Celesta flow cytometry (Becton, Dickinson and Company, Franklin Lakes, NJ, USA).

2.25. Analysis of immune cells in tumors using flow cytometry

At the endpoint, mice were sacrificed, and tumors were collected and cut into pieces (3–4 mm³). These tumor pieces were then shaken and digested in a medium (pH 7.40) containing 1% fetal bovine serum, 1 mg/mL collagenase IV (C8160, Solarbio Science & Technology Co., Ltd., Beijing, China) and 1 mg/mL Dnase I (D8071, Solarbio Science & Technology Co., Ltd., Beijing, China) for 1 h. Next, the tissue suspension was filtered through a cell strainer to obtain single-cell suspensions. Then these cells were stained with fluorescence-labeled antibodies for analysis of immune cells with CytoFLEX S Flow Cytometer (Beckman Coulter).

2.26. Chemotaxis assay

For CD4⁺ T chemotaxis assay in vivo, LLC lung cancer cell-bearing mice were pre-dosed with FPNs, anti-PD1, or Vehicle, and then intravenously injected with 4 × 10⁴ CD4⁺T cells labeled with DiI. The mice were sacrificed, and the tumors were collected on Day 2 after CD4⁺ T cells injection. An immunofluorescent assay was performed to analyze the tumor-infiltrated DiI-CD4⁺T lymphocytes.

For the in vitro lymphocyte migration assay, LLC cells (2 × 10⁵ cells per well) were seeded into the lower chambers of Transwell inserts and cultured at 37 °C with 5% CO₂ for 24 h. The cells were then treated with FPNs, anti-PD1, or PBS. CD4⁺ T cells (3 × 10⁴ cells) were added to the upper chambers to assess their migration. Migration was evaluated by enumerating the number of migrated cells in the bottom chamber using an inverted fluorescence microscope (Nikon Corporation, Ts2R, Tokyo, Japan) after 24 h.

2.27. Statistical analysis

The data were presented as the mean ± standard deviation (SD). Unpaired Student’s t-test was used to analyze data from two groups for any significant differences, and one-way ANOVA was used for multiple comparisons when the data involved three or more groups. All statistical analyses were performed using SPSS Statistics 25 software (IBM Corporation, Armonk, NY, USA) and GraphPad Prism 8.0 software (GraphPad Software, Inc., Boston, MA, USA). A P-value of less than 0.05 was considered statistically significant (∗P < 0.05; ∗∗P < 0.01; ∗∗∗P < 0.001; n.s, not significant).

3. Results and discussion

3.1. Expression levels of LSD1 in patients with non-small cell lung cancer (NSCLC)

To assess the modulation of tumor immunity by LSD1 in lung cancer, the relationships among LSD1, cancer, and immune cell characteristics were first analyzed using TIMER2.0 (http://timer.cistrome.org), which is a platform that provides robust estimates of immune infiltration levels⁶². The results revealed that LSD1 levels were negatively correlated with the occurrence of many cancers (Fig. 1A) and highly correlated with the proportion of tumor-infiltrating immune T cells (Fig. 1B). Then, using the GEPIA 2 platform (http://gepia2.cancer-pku.cn), the impact of LSD1 on the overall survival of NSCLC patients was analyzed. GEPIA 2 is a valuable and highly cited resource for gene expression analysis based on tumor and normal samples from the TCGA and the GTEx databases⁶³. The results indicated that low expression of LSD1 is associated with good overall survival (OS) in patients with NSCLC and is associated with better survival rates (Fig. 1C). Finally, 80 clinical lung cancer samples were subjected to immunohistochemistry (IHC) staining (Fig. 1D–F). The results showed that LSD1 was overexpressed in lung cancer tissues, and the expression of LSD1 was negatively correlated with CD3 expression in tumor tissues compared with adjacent tissues, which highlighted that LSD1 may be an immunosuppressive factor that can be targeted in tumor immunotherapy.

Figure 1.

Expression levels of LSD1 (KDM1A) in patients with NSCLC. (A) Differential expression analysis of LSD1 protein between various tumor tissues and normal tissues using TIMER2.0 (Red: tumor, blue: normal). (B) Relationship between T cell infiltration level and lung cancer, analyzed by TIMER 2.0. (C) The prognostic significance of LSD1 in LUAD and LUSC, analyzed by GEPIA 2. (D) Representative images of LSD1 expression in normal lung tissues and lung cancer tissues stained by IHC. Scale bar = 50 μm. (E) Quantification analysis in D, F. (F) Correlation of LSD1 and CD3 proteins in IHC-stained lung cancer tissues. Scale bar = 50 μm. Data are presented as mean ± SD (n = 80). ∗P < 0.05, ∗∗P < 0.01 and ∗∗∗P < 0.001.

3.2. FA, a natural LSD1 inhibitor, can directly target LSD1 demethylase activity

In our preliminary experiments, LSD1 inhibitors were screened, and FA (Fig. 2A) was found to effectively inhibit LSD1⁶⁴. To validate the molecular interaction between FA and LSD1, we performed surface plasmon resonance (SPR) analysis and showed that FA bound to LSD1 with high affinity (K_D = 5.20 μmol/L) (Fig. 2B). A drug affinity responsive target stability (DARTS) assay was then performed. As expected, the protease-mediated degradation of LSD1 was inhibited in the presence of FA (Fig. 2C, Supporting Information Fig. S1A), suggesting that FA can interact directly with LSD1. A cellular heat transfer assay (CETSA) was subsequently used to demonstrate that FA could stabilize LSD1 at relatively high temperatures (Fig. 2C, Fig. S1B). LSD1, as a histone demethylase, can catalyze the demethylation of the histone lysines H3K4 and H3K9, and the degree of LSD1 inhibition can be inferred by measuring the methylation of these substrates. The LSD1 quantitative enzyme activity assay revealed that FA significantly inhibited the histone demethylase activity of LSD1, with an IC₅₀ of 8.74 μmol/L (Fig. 2D), and western blotting (WB) revealed that FA significantly increased the accumulation of H3K4me1 and H3K4me2, which reached to 232.74% and 206.47%, respectively. ORY1001 was used as a positive control (Fig. 2E, Fig. S1C–S1E). These data strongly suggested that FA is a specific inhibitor of LSD1. The structure-activity relationship of FA with LSD1 was further explored. The results of the dilution assay revealed that LSD1 activity failed to recover after 40-fold dilution of the LSD1/compound complex, indicating that FA may irreversibly interact with LSD1 (Fig. 2F). Additionally, the mode of inhibition for FA was determined by visualizing the kinetic data in Lineweaver-Burk plots and fitting the data points using the Morison equation (Supporting Information Fig. S2). The results showed that FA is a noncompetitive inhibitor that does not affect the binding of LSD1 to H3K4me2 and FAD substrates.

Figure 2.

Mechanism and structure-activity relationship of FA inhibition of LSD1. (A) Chemical structure of FA. (B) Kinetic analysis of FA against LSD1. (C) The DARTS showed the stability of FA on LSD1 when the ratio of pronase to protein was 1:250. The CETSA determined the thermal stabilization of the LSD1 interaction with FA at a series of temperatures from 37 to 52 °C. (D) LSD1 quantitative enzyme activity assay results. (E) FA can inhibit LSD1 demethylation activity. (F) FA irreversibly inhibited the demethylation activity of LSD1. ORY1001 was the positive control. (G) Molecular docking results (PDB:6W4K). (H) Peptide fragments' MS/MS spectra of recombinant LSD1 incubated without or with FA. (I) Mutated C³⁶⁰ residue of LSD1 into A. (J) LSD1^C360A quantitative enzyme activity assay results. (K) The DARTS and CETSA of LSD1^C360A under the same conditions as the above experiments. Data are presented as mean ± SD (n = 3). ∗P < 0.05, ∗∗P < 0.01 and ∗∗∗P < 0.001.

More importantly, on the basis of the α,β-unsaturated ketone moiety of FA, molecular docking simulations were conducted to explore the structure-activity relationship between FA and LSD1 (Fig. 2G). The 6W4K crystal structure with a free FAD, an H3K4 mimetic peptide as the substrate, and CoREST as a corepressor was selected as the docking template from the PDB. As expected, Cys³⁶⁰ formed a covalent bond with the α,β-unsaturated ketone moiety of FA. Next, recombinant LSD1 protein, incubated with or without FA, was detected by LC–MS/MS analysis. This analysis revealed a mass shift of 478.15 Da for the Cys³⁶⁰-containing peptide C³⁶⁰PL in the presence of FA, which perfectly matched the molecular weights of the fragment ions in the FA mass spectrometry results. It is speculated that this was the result of one molecule of FA hydrolyzing one rhamnose. As a control, the corresponding C³⁶⁰-free peptides in the DMSO group showed no mass shift. Thus, the C³⁶⁰ residue of LSD1 can be covalently modified by FA (Fig. 2H, Supporting Information Fig. S3). To further prove the binding of FA to the C³⁶⁰ residue of LSD1, we mutated the C³⁶⁰ residue of LSD1 to alanine (A) (Fig. 2I), performed a transfection experiment to obtain the mutated protein, and performed affinity detection. The results of the LSD1 enzyme activity assay (Fig. 2J) and DARTS and CESTA assays (Fig. 2K, Supporting Information Fig. S4) revealed that the LSD1^C360A lost its ability to bind to FA. These results strongly demonstrated that FA functions as a natural inhibitor of LSD1 by directly inhibiting its demethylase activity.

3.3. The construction of a synergistic therapy system can increase the stability and targeted effects of FA

However, there was a noticeable finding that FA was not stable in PBS and may undergo chemical structure transformation (Supporting Information Fig. S5). The results of high-performance liquid chromatography (HPLC) revealed that the new compound FAC was produced at 6.81 min in a PBS environment (Supporting Information Fig. S6). FAC has the same molecular formula as FA does⁶⁵, C₂₉H₃₆O₅, as determined by the (⁺)–HRESIMS ion peak at m/z 647.1953 [M + Na]⁺ (calculated for C₂₉H₃₆NaO₅, 647.1946) (Supporting Information Figs. S7 and S8) and the ¹³C nuclear magnetic resonance (NMR) data. Analysis of the NMR data revealed that FAC and FA were isomers with different 3,4-dihydroxyphenylethyl moiety-glucose binding locations (Supporting Information Figs. S9–S15). The heteronuclear multiple bond correlations (HMBC) between H–3′ to C–9‴ and H–8‴ to C–9‴ suggested that the 3,4–dihydroxyphenylethyl moiety bound to the glucose at C–3′ through an ester bond (Supporting Information Figs. S16–S18).

To address the structural instability and nontargeting effects of FA in the biological environment, combined with PD1/PD-L1 immune checkpoint therapy, a targeted, long-term circulation, tumor microenvironment (TME)-sensitive synergistic therapy system, DSPE-PEOz-FA-anti-PD1 (FPNs), was established and characterized. The FPNs were prepared via the film dispersion method⁶⁶. In general, DSPE-poly(2-ethyl-2-oxazoline) (DSPE-PEOz), lecithin, and cholesterol composed the liposome bilayers of FPNs. Anti-PD1 was encapsulated in liposome vesicles, and the unstable drug FA was encapsulated in the liposomal bilayer. Ultraviolet–visible (UV–vis) spectroscopy revealed that FA exhibited two absorption peaks at 280 nm and 330 nm, both of which were also observed in the FPNs, whereas liposomes in the control group had no UV absorption signal (Fig. 3A), indicating that FA was encapsulated successfully in the FPNs. To maximize drug delivery to the target, the encapsulation efficiencies (EEs) of FA and anti-PD1 were evaluated separately using a microplate reader, with FA in the linear range of 0–1 mg/mL at 330 nm and anti-PD1 in the linear range of 0–2000 μg/mL at 562 nm (Supporting Information Fig. S19). When the weight ratio of FA to anti-PD1 was 7:3, the optimum EE values were obtained (85.77% for FA and 81.44% for anti-PD1) (Fig. 3B).

Figure 3.

Characterization of FPNs. (A) UV–Vis spectra of different formulations. (B) Encapsulation efficiency of different weight ratios of FA and anti-PD1. (C) The hydrodynamic diameter of FPNs was measured by NTA. (D) TEM images of FPNs and blank liposomes (negative control). Scale bar = 100 nm. (E) The size distribution of FPNs was determined by TEM imaging. (F) ζ-potentials of FPNs and control liposomes. (G) Particle size variation of FPNs in water, PBS, DMEM, and 10% FBS DMEM after 3 or 7 days of incubation. (H) FA release curve from FPNs in vitro. Data are presented as mean ± SD (n = 3).

Moreover, an appropriate size and good particle size distribution are essential for the in vivo delivery of drug-loaded liposomes⁶⁷. The hydrated particle size of the FPNs was measured via nanoparticle tracking analysis (NTA), and was found to be 120.04 ± 3.15 nm (Fig. 3C), which was conducive to enhancing the passive permeability and retention effect (EPR) on tumors. The morphology of the FPNs and blank liposomes was evaluated using transmission electron microscopy (TEM), which revealed that the FPNs maintained an approximately spherical morphology and uniform distribution (Fig. 3D). The sizes of the FPNs and blank liposomes were measured with Nano Measurement software and were 64.91 ± 13.70 nm and 33.08 ± 6.23 nm, respectively (Fig. 3E). Moreover, the zeta potentials of the blank liposomes and FPNs were −57.40 ± 1.90 mV and −47.93 ± 2.15 mV, respectively (Fig. 3F). The colloidal stability of the FPNs was verified via NTA. There was no significant difference in particle size after treatment with deionized water, PBS, DMEM, or 10% FBS DMEM at 4 °C for Days 1, 3, and 7 (Fig. 3G, Supporting Information Fig. S20). These findings suggested that the structural integrity of FPNs can be maintained for a long time without premature drug leakage. Owing to the large number of tertiary amide structures in its molecules, PEOz can bind to H⁺ in an acidic environment, and N ionization can induce a positive charge in the chain, increase the charge density, change the chain from hydrophilic to hydrophobic, weaken its stability, and finally release the drug⁶⁸. Therefore, we studied the release behavior of FPNs in a physiological environment (pH 7.40) and an acidic environment mimicking the TME (pH 6.30). The results showed that both FA and anti-PD1 were continuously released in a weakly acidic environment, and their levels tended to stabilize after 12 h. After 24 h, the FA encapsulated by FPNs accumulated and released nearly 95.90% of the payload under weakly acidic conditions, which was 1.46 times that released in the physiological environment (Fig. 3H), while anti-PD1 accumulated and released nearly 72.08% of the payload, which was 3.00 times that released in the physiological environment (Supporting Information Fig. S21).

3.4. Targeting and in vivo biodistribution of FPNs in the TME

Next, we prepared P-FPNs in which the anti-PD1 was replaced with the PD1 protein to visualize the release process in LA795 cells. PD1 was labeled with fluorescent dyes after incubation, and changes in fluorescence were detected by the HCS system and flow cytometry. The HCS results showed that the PD1 protein bound to intracellular PD-L1, and the red fluorescence intensity gradually increased with increasing incubation time (Fig. 4A). The fluorescence intensity at 24 h was 3.05 times greater than that at 1 h (Fig. 4B). Similar results were obtained via flow cytometry (Fig. 4C). These results indicated that the FPNs can release drugs in the TME under the action of DSPE-PEOz to exert therapeutic effects. In summary, various experiments have verified that FPNs are platforms with high drug loading capacity, high stability, and a pH-triggered release mechanism.

Figure 4.

Targeting and in vivo biodistribution of FPNs in TME. (A) HCS fluorescence images of LLC cells incubated with FPNs for 1, 12, and 24 h. Scale bar = 50 μm. (B) Quantification of FPNs release experiments in A, C. (C) Flow cytometry analysis of LLC cells incubated with FPNs for 1, 6, 12, and 24 h. (D) In vivo fluorescence images of LLC tumor-bearing nude mice after intravenous injection of Cy5.5-FPNs or Cy5.5. Tumor regions were marked by the circles. (E) Quantitative plot of the average fluorescence intensity of tumors divided by regions of interest. (F) Fluorescence images of visceral organs and tumors harvested at 24 h post-injection. (G) Average fluorescence intensity of tumors and visceral organs. Data are presented as mean ± SD (n = 3). ∗P < 0.05, ∗∗P < 0.01 and ∗∗∗P < 0.001.

Through the EPR effect, nanoparticles in the range of 100–200 nm or polymers of macromolecular compounds larger than 40 kDa can gradually accumulate in the tumor vascular area and escape removal by the liver and spleen, passively targeting tumors. Through this effect, nanoparticles can successfully deliver anticancer compounds to solid tumor tissues or environments^69,70. Next, we prepared Cy5.5-FPNs with Cy5.5 instead of FA or anti-PD1 to evaluate the tumor-targeting ability of the FPNs. The biodistribution of the FPNs was assessed using an in vivo imaging system (IVIS). The fluorescence signal of Cy5.5-FPNs at 24 h was 2.42 times greater than that at 2 h, and the accumulation of free Cy5.5 at the tumor site was much lower than that of Cy5.5-FPNs over time (Fig. 4D and E). 24 h after injection, the mice were euthanized, and major organs and tumors were collected for ex vivo imaging. The fluorescence intensity of Cy5.5 in tumors in the FPNs group was 64.03-fold greater than that in the free Cy5.5 group (Fig. 4F and G). These results indicated that the FPNs could effectively accumulate at the tumor site. In addition, both groups showed strong fluorescence signals in kidneys tissue, which may be since larger nanoparticles in the size range of 20–100 nm can use fenestrations in capillary endothelial cells (approximately 70–80 nm) to access the mesangium. Nanoparticles in the range of 75 ± 25 nm have been reported to target mesangial cells within the kidney glomerular compartment71, 72, 73. The average size of FPNs (64.91 ± 13.70 nm) falls within this range, and their structural integrity likely facilitates passive retention in the renal microvasculature. Meanwhile, in order to better demonstrate the accumulation and release of FPNs in vivo, HPLC–MS/MS was used to detect the FA content in mouse tumor sites at different time points (Supporting Information Fig. S22). The results showed that after intravenous injection of FPNs into the tail vein of tumor-bearing mice, FA rapidly accumulated at the tumor site within 0–12 h, and the concentration of FA increased slowly within 12–24 h, maintaining a relatively stable level within 36 h. This indicated that FPNs enriched FA at the tumor site through the EPR effect and the active targeting mechanism of DSPE-PEOz, achieving the sustained release characteristics of the FPNs treatment system and consistent with in vivo imaging results.

3.5. Biosafety analysis and in vivo antitumor activity evaluation of FPNs

First, an in vivo biosafety evaluation of the FPNs was conducted. There were no significant differences in the organ coefficients of the heart, liver, spleen, lung, or kidney between the FPNs group and the PBS group (Supporting Information Fig. S23), and no significant tissue damage or other side effects were detected via hematoxylin and eosin (HE) staining (Supporting Information Fig. S24). The levels of the kidney function markers BUN and CRE and the liver function markers GOT and GPT were within the normal ranges in the FPNs group and were not substantially different from those in the PBS group, demonstrating that FPNs did not induce obvious hepatic or kidney damage (Supporting Information Fig. S25A). Furthermore, a hemolysis assay was performed; under ultrapure water conditions, erythrocytes were completely ruptured, whereas no significant hemolysis was observed when the cells were exposed to PBS or various concentrations of FPNs (Fig. S25B), and the calculated hemolysis rate was less than 5% in all groups (Fig. S25C). Together, these results suggested that the FPNs did not induce severe systemic toxicity and possessed suitable biosafety.

Therefore, cell line-derived allograft (CDA) models using LLC and LA795 mouse lung cancer cell lines were established to further verify the effect of FPNs in the TME (Fig. 5A). Taking the results of the LLC model as an example, routine monitoring revealed significant inhibition of tumor growth in the FPNs group compared with the other treatment groups and no significant change in body weight (Fig. 5B–D, Supporting Information Fig. S26A and S26B). After 2 weeks of treatment, the mice in the FPNs group were euthanized, and the tumor weight was 7.44 times lower than that in the PBS group (Fig. 5E, Fig. S26C). Next, the tumors were stained with HE (Fig. 5G, Fig. S26D). Each group exhibited various degrees of necrosis in images of the tumor tissue sections. In the FPNs group, tumor cells were diffusely distributed in the lamina with a high nucleus-plasma ratio, unclear boundaries were observed, and heteromorphic nuclei were dispersed throughout the lamina. The apoptosis and necrosis of some tumor cells were obvious, with necrosis accounting for ∼50% of the total area. In addition, the expression of Ki67, a cell proliferation marker, was significantly lower in the FPNs group, which was 0.39 times higher than that in the PBS group, and the TUNEL staining results also showed the same trend. The apoptosis rates in the FPNs group were 47.70, 34.93, 3.74, 2.66, and 1.87 times greater than those in the liposome, FA, anti-PD1, and FA + anti-PD1 groups, respectively (Fig. 5F and G, Fig. S26D and S26E). The results indicated that the FPNs showed a strong therapeutic effect on tumors in vivo and were significantly more effective than the anti-PD1 agent or FA alone or in combination. Moreover, an anti-H3K4me2 antibody was used to detect the accumulation of the LSD1 substrate in tumors (Fig. 5H, Fig. S26F), and compared with the control, FA inhibited the demethylation activity of LSD1 in vivo. Consistent results were also found in the LA795 model, which together demonstrated that FPNs enable the precise and maximal delivery of compounds to better inhibit tumor growth.

Figure 5.

In vivo antitumor activity evaluation. (A) Depiction of the therapy. (B) Changes of mice body weight in LLC mice (n = 5). (C) Photographs of LLC and LA795 tumor tissues excised on Day 17 after different therapies (n = 5). (D) Changes of tumor volume in LLC mice under different conditions (n = 5). (E) Weights of tumors harvested from LLC mice (n = 5). (F) Statistical analysis data of TUNEL staining and Ki67 staining in LLC tumor. (G) H&E staining images (Scale bar = 100 μm), TUNEL staining images (Scale bar = 50 μm), and Ki67 staining images (Scale bar = 50 μm) of posttreatment LLC tumor sections. (H) Pictures of IHC of H3K4me2 in LLC and LA795 tumor tissues following treatments. Scale bar = 100 μm. Data are presented as mean ± SD (n = 3), unless otherwise noted. ∗P < 0.05, ∗∗P < 0.01 and ∗∗∗P < 0.001.

3.6. FPNs inhibit tumor growth by reshaping the tumor immune microenvironment and promoting the T cell immune response

Currently, PD1/PD-L1 immunotherapy, in which anti-PD1 can inhibit tumor cell immune escape, thereby killing tumors, is widely studied⁷⁴. Therefore, gating strategy-based flow cytometry was used to detect infiltrating immune cells in tumors. The results showed that FPNs treatment affected the proportion of Treg cells in the tumor tissue of LLC tumor-bearing mice, which contributed to the formation of tumor immune suppressive tumor microenvironment (TME) and promoted tumor growth in lung cancer. The data showed that the proportion of Treg cells in the PBS group was 26%, which was 3.11 times (8.35%) higher than in the FPNs group and 2.25 times (11.57%) higher than in the FA + anti-PD1 group (Fig. 6A, Supporting Information Fig. S27A). In addition, after treatment with FPNs and FA + anti-PD1, a significant increase in dendritic cells (DCs) activation was observed, with the proportion of activated DCs being 10.17 times and 7.78 times higher than the PBS group, respectively (Fig. 6B, Fig. S27B). Tumor-associated macrophages (TAMs) are the most prominent subpopulation of infiltrating cells in tumor tissue, playing a critical role in the TME. Specifically, the relative proportion of M1-type TAMs decreases, while the increase of M2-type TAMs is associated with tumor occurrence and development⁷⁵. Flow cytometry analysis showed that FPNs treatment significantly regulated the TAM subpopulation. Compared with the PBS group, the M1/M2 ratio increased to 0.40 in the presence of FPNs and 0.04 in the presence of FA + anti-PD1 (Fig. 6C, Fig. S27C). At the same time, it was worth noting that the proportion of CD4⁺T cells infiltration in the treatment group was significantly increased, as compared with those in the control group, the number of CD8^–CD4⁺T cells in the FPN and FA+anti-PD1 groups increased from 34.20% to 49.60% and 46.10%, respectively, in LLC tumors(Fig. 6D, Fig. S27D), indicating that drug treatment significantly increased the proportion of infiltrating immune cells. Moreover, IHC staining for the CD4 protein in tumor tissues yielded consistent results (Fig. 6E, Supporting Information Fig. S28). Furthermore, we used an in vitro Transwell migration assay and an in vivo T cell chemotaxis assay to verify the effect of the FPNs on immune cell infiltration. The in vivo results (Fig. 6F and H) revealed that intravenously injected CD4⁺ T cells aggregated in the tumor tissue under the action of drugs, and the fluorescence intensity in the FPNs group was 1.95 times higher than that in the anti-PD1 group. The in vitro results (Fig. 6G and H) also revealed that 3.14-fold more CD4⁺T cells migrated to the chamber housing the tumor cells in the FPNs group than in the anti-PD1 group. Next, the levels of the T cell killing factors IFN-γ, TNF-α, and IL-2 in tumor tissues were measured by ELISA (Fig. 6I, Supporting Information Fig. S29). Taking LLC tumor samples as an example, the FPNs group presented significantly elevated levels of IFN-γ, TNF-α, and IL-2, which were 1.25, 1.23, and 1.24 times greater than those in the FA + anti-PD1 group, respectively. In summary, FPNs with tumor targeting ability can exert excellent antitumor effects in vivo while increasing the infiltration of tumor immune cells.

Figure 6.

FPNs affect tumor growth by participating in the immune response. (A–D) After treatment with FPNs, FA + anti-PD1, anti-PD1, FA, Liposome, and PBS, flow cytometry was performed to detect the infiltration levels of Treg cells (CD25⁺Foxp3⁺), activated DC cells (CD11C⁺MHC II⁺), M1 TAMs (CD86⁺)/M2 TAMs (CD206⁺) ratio and CD4⁺ T cells (CD4⁺CD8^-) in LLC and LA795 tumor tissues. (E) Pictures of immunohistochemistry of CD4 in LLC and LA795 tumor tissues following treatments. Scale bar = 100 μm. (F) CD4⁺ T trafficking and biodistribution of DiI-labeled CD4⁺ T cells in LLC tumor with PBS, anti-PD1, or FPNs treatment administered intravenously. Scale bar = 100 μm. (G) In vitro transwell migration assay. (Image display used ImageJ to transform the bright field view into black-and-white (B&W) high-contrast mode). Scale bar = 100 μm. (H) Quantitative results for F and G. (I) Changes of IFN-γ, TNF-α, and IL-2 levels in tumor tissues of LLC mice after different drug treatments. Data are presented as mean ± SD (n = 3). ∗P < 0.05, ∗∗P < 0.01 and ∗∗∗P < 0.001.

3.7. FPNs induce the killing ability of CD4⁺T cells through cell cycle G1/S phase arrest

CD4⁺T cells play an auxiliary and inhibitory role in inducing and maintaining CD8⁺ CTL responses, and there is increasing evidence that CD4⁺ T cells play a more direct role in antitumor immunity; however, the mechanism by which CD4⁺ T cells directly mediate tumor clearance is not defined⁷⁶. Therefore, we established a coculture system of cancer cells and CD4⁺ T cells to verify the role of CD4⁺ T cells in the drug treatment of FPNs. First, CCK-8 experiments were used to prove that FPNs alone had no cytotoxicity. However, in the presence of CD4⁺ T cells, LLC cells died, and the IC₅₀ of the FPNs was 27.62 μmol/L (FA equivalent concentration) (Fig. 7A), indicating that CD4⁺T cells may have a killing effect on cancer cells. The HCS system was used to detect the killing effect of T cells. Compared with those in the control group, T cells in the FPNs group killed up to 49.91% and 64.15% of A549 and LLC cells, respectively (Fig. 7B, Supporting Information Fig. S30A). These results suggest that CD4⁺ T cells may have the ability to kill tumor cells directly and that FPNs can significantly enhance the killing effect. Anticancer drugs or cells often work by inhibiting tumor cell cycle progression or promoting apoptosis^77,78; therefore, we conducted further experiments to investigate the effects of CD4⁺ T cells on tumor cell cycle progression. Flow cytometry revealed that drug treatment significantly increased the proportion of cells in the G1 phase, and the percentage of A549 and LA795 cells increased from 36.50% to 37.80% to 64.60% and 68.10%, respectively, after FPNs infiltration (Fig. 7C, Fig. S30B). As expected, WB revealed a significant decrease in the expression of the G1/S phase marker proteins CDK6, CDK4, and Cyclin D1 in the FPNs group (Fig. 7D, Fig. S30C). To better explain that CD4⁺T cells, rather than FPNs, exhibit cycle inhibition on tumor cells. A549 and LA795 cells were exposed to FPNs environment for flow cytometry cycle detection, and the data showed that FPNs did not affect the cell cycle (Supporting Information Fig. S31). The above data suggested that CD4⁺ T cells can directly kill tumor cells by inducing cell cycle arrest in the G1/S phase.

Figure 7.

CD4⁺ T cells can directly kill tumor cells through G1/S phase arrest. (A) CCK-8 analysis of cells treated with different concentrations of FPNs in the presence and absence of CD4⁺ T cells. (B) Cell survival of A549 and LLC cells under different conditions after treatment with anti-CD3/CD28-activated CD4⁺ T cells. Scale bar = 200 μm. (C) Activated CD4⁺ T cells induced cell cycle arrest at G1/S phase under different conditions determined by flow cytometry. (D) Expression of the marker proteins of G1/S phase arrest, CDK4, CDK6, and cyclinD1, in A549 and LLC cells under different conditions after treatment with anti-CD3/CD28 activated CD4⁺ T cells. Data are presented as mean ± SD (n = 3). ∗P < 0.05, ∗∗P < 0.01 and ∗∗∗P < 0.001.

3.8. Inhibition of LSD1 suppresses tumor growth by promoting the killing ability of CD4⁺ T cells

Surprisingly, we found that FA alone also enhanced the killing effect of CD4⁺ T cells, suggesting that LSD1 may promote tumor growth by inhibiting T cell responses. Therefore, we constructed LSD1 knockout A549 and LLC cell lines (Fig. 8A). LLC and corresponding LSD1-KO cells were inoculated subcutaneously into BALB/c and BALB/c nude mice, and after 3 weeks, the tumors were excised and weighed. The results showed that in BALB/c nude mice, the tumor weights and volumes in the LSD1-KO group were comparable to those in the LLC group (Fig. 8B and C). In contrast, in BALB/c mice, tumors were significantly inhibited after LSD knockout, with a mean tumor weight that was 44.24% of that in the LLC group. The body weights of the mice in each group remained the same (Fig. 8D, Supporting Information Fig. S32). Flow cytometry revealed a 50.37% increase in the proportion of infiltrating CD4⁺T cells in LSD1-KO tumors compared with that in normal tumors (Fig. 8E and F). In addition, the results of coculture of tumor cells with activated CD4⁺ T cells revealed that CD4⁺ T cells could directly kill cancer cells and that LSD1 deletion significantly enhanced the killing effect; the killing capacity of CD4⁺ T cells against A549 and LLC cells was 68.17% and 54.23%, respectively, whereas that of their corresponding LSD1-KO cells was as high as 92.08% and 80.83%, respectively (Fig. 8G, Supporting Information Fig. S33A). However, LSD1 knockout slowed cell growth to some extent (Fig. S33B). Taken together, these findings suggested that LSD1 was a suppressor of T cell response in lung cancer, and that LSD1 knockout may significantly inhibit the growth of lung cancer cells by increasing the CD4⁺ T cells' killing efficiency.

Figure 8.

Inhibition of LSD1 hinders tumor growth by participating in CD4⁺ T cell responses. (A) Expression of LSD1 in LLC and A549 cells with or without LSD1 knocked out. (B) Changes of tumor volume in BALB/c and BALB/c nude mice bearing LLC cells, whether LSD1 was abrogated or not (n = 5). (C) Images of tumors of BALB/c and BALB/c nude mice bearing LLC cells (n = 5). (D) Tumor weight of BALB/c mice and BALB/c nude mice, whether LSD1 was abrogated or not (n = 5). (E, F) Percentage of CD4⁺ T cells in CD3⁺ infiltration cells isolated from tumors of BALB/c mice bearing LLC cells in the presence or absence of LSD1. (G) Cell survival of LLC and A549 cells in the presence or absence of LSD1, treated with anti-CD3/CD28-activated T cells. Scale bar = 200 μm. (H) Activated CD4⁺ T cells induced cell cycle arrest in the G1/S phase, and knockout of LSD1 had a facilitating effect. (I) Flow cytometry quantification of cycles in A549 cells and LLC cells. (J) Expression of the marker proteins of G1/S phase arrest in LLC and A549 cells under anti-CD3/CD28 activated CD4⁺ T cells and with or without LSD1. Data are presented as mean ± SD (n = 3), unless otherwise noted. ∗P < 0.05, ∗∗P < 0.01 and ∗∗∗P < 0.001.

The LSD1-KO cells were subsequently co-incubated with CD4⁺ T cells for cell cycle analysis. Flow cytometry revealed that CD4⁺ T cells directly arrested cells at the G1/S phase, and the proportions of A549 and LLC cells in the G1 phase increased from 60.20% to 22.70% to 73.60% and 25.60%, respectively. After LSD1 knockout, the proportion of T cells in the G1 phase increased from 51.10% to 25.50% to 80.80% and 42.60% (Fig. 8H and I), respectively, indicating that LSD1 knockout significantly enhanced the cytotoxicity of T cells. WB also revealed that, compared with those in the A549 and LLC groups, the expression of CDK6, CDK4, and Cyclin D1 was lower after CD4⁺ T cells infiltration (Fig. 8J, Supporting Information Fig. S34), and LSD1 knockout significantly arrested tumor cells at the G1/S phase. The above results further revealed that CD4⁺ T cells can directly kill tumor cells by inducing cell cycle arrest in the G1/S phase, and this cell cycle arrest is hindered by LSD1.

3.9. Inhibition of LSD1 regulates PD-L1 expression through the EVs-mediated cell surface PD-L1 rearrangement mechanism

Inhibition of LSD1 can stimulate antitumor immunity, and IHC staining was performed to explore the regulatory effect of FPNs drug treatment on PD-L1. The results revealed that the addition of FA and FPNs significantly inhibited the expression of PD-L1 (Fig. 9A, Supporting Information Fig. S35A), suggesting that FA may affect PD-L1 expression by inhibiting LSD1 activity. To further explore the regulatory effect of LSD1 on PD-L1 in lung cancer, total PD-L1 and membrane-associated PD-L1 levels in lung cancer cells were quantified, revealing significantly varied levels of PD-L1 in different lung cancer cell lines (Fig. S35B). Furthermore, the relative levels of total and membrane-bound PD-L1 varied significantly among the cell lines. PD-L1 expression was significantly downregulated when LSD1 was knocked out using sgRNA (Fig. 9B, Fig. S35C). Because PD-L1 is a transmembrane protein that has an immunosuppressive function in extracellular regions^79,80, the level of membrane PD-L1 in LSD1-knockout cells was also evaluated. Unexpectedly, the expression level of membrane PD-L1 did not change significantly (Fig. 9C). The use of the LSD1 inhibitor ORY1001 also produced the same results (Supporting Information Fig. S36A–S36C). We hypothesized that the inhibition of LSD1 activity may influence the distribution of PD-L1. Previous studies reported that PD-L1 can be degraded through the proteasomal and lysosomal pathways^81,82. Therefore, A549 cells were treated with the protein synthesis inhibitor cycloheximide (CHX) and the lysosomal inhibitor chloroquine (CQ), and the data showed that LSD1 deletion did not reduce the degradation of PD-L1 by proteasomes or lysosomes (Fig. 9D, Fig. S36D). Therefore, we speculated that the endosomal transport pathway was responsible for PD-L1 synthesis and degradation.

Figure 9.

Inhibition of LSD1 activity maintained the expression of membrane PD-L1. (A) Pictures of IHC of PD-L1 in LLC and LA795 tumor tissues following treatments. Scale bar = 100 μm. (B) Expression of PD-L1 in A549 and LLC cells in the presence or absence of LSD1. (C) Expression of membrane PD-L1 in LLC and A549 with or without LSD1 knocked out. (D) Expression of PD-L1 in A549 cells with or without LSD1 in the presence of 20 μmol/L CQ. (E) Images of PD-L1, RAB11, and TSG101 in A549 cells and LLC cells in the presence of LSD1 or not. Scale bar = 20 μm. (F, G) Expression of RAB11 and TSG101 in A549 and LLC cells in the presence of LSD1 or not. (H) Morphology of EVs in A549 cells in the presence or absence of LSD1. Scale bar = 2 μm. Data are presented as mean ± SD (n = 3). ∗P < 0.05, ∗∗P < 0.01 and ∗∗∗P < 0.001.

To verify this hypothesis, we used fluorescent probes to label PD-L1, the recycling endosome protein RAB11, and the multivesicular bodies (MVBs) protein TSG101 in A549, LLC, and corresponding LSD1-KO cell lines. The results showed that PD-L1 colocalized with RAB11 and TSG101 in the cytoplasm (Fig. 9E, Supporting Information Fig. S37A). WB revealed that LSD1 deficiency significantly increased the expression of RAB11 and decreased the expression of TSG101 (Fig. 9F and G, Fig. S37B). Scanning electron microscopy (SEM) results also revealed that the number of extracellular vesicles (EVs) decreased when LSD1 was knocked out (Fig. 9H, Supporting Information Fig. S38). To further validate these findings in vivo and strengthen the mechanistic link between LSD1 inhibition and EVs secretion, we conducted an additional experiment in tumor-bearing mice. Mice were treated with FPNs or PBS, and EVs were subsequently extracted from excised tumor tissues. TEM confirmed the presence of intact EVs in both groups, displaying characteristic round-shaped vesicles with a size range of 50–150 nm and clearly visible bilayer membrane structures (Supporting Information Fig. S39A). The NTA detection of hydrated particle size results showed that the particle size of EVs in the PBS group and FPNs group were 138.4 nm and 154.9 nm, respectively (Fig. S39B). Notably, TEM also revealed an overall reduction in EVs density in the FPNs-treated tumors. Consistently, NTA demonstrated that the concentration of tumor-derived EVs in the FPNs group was significantly lower than that in the PBS group (Fig. S39C), providing direct quantitative evidence that LSD1 inhibition effectively suppresses EVs secretion in vivo. IHC staining of tissues derived from the in vivo experiments similarly demonstrated that tumors expressed less TSG101 after treatment with LSD1 inhibitors (Supporting Information Fig. S40), supporting the notion of reduced EVs biogenesis and release. Collectively, these results indicated that LSD1 inhibition or deletion reduces EVs secretion, promotes recycling endosome activity, and enhances the transport of intracellular PD-L1 back to the membrane, thereby maintaining membrane-associated PD-L1 levels and modulating tumor-immune interactions.

3.10. Inhibition of LSD1 reduces the secretion of EVs and EVs-PD-L1

EVs-PD-L1 also have immunosuppressive functions, and the secretion of these EVs may be regulated by LSD1^42,83. Therefore, EVs from A549 and LSD1-KO A549 cells were extracted using a kit for TEM and NTA analysis. A-EVs and KO-EVs exhibited double-layer membrane structures and saucer-like morphologies, with hydrated particle sizes ranging from 144.47 to 150.91 nm, respectively (Fig. 10A, Supporting Information Fig. S41A). The EVs secretion inhibitor GW4869 was subsequently used to confirm that EVs-PD-L1 were derived from cancer cells. When A549 cells were exposed to GW4869, a significant accumulation of membrane PD-L1 and total PD-L1 was observed, whereas the levels of EVs and EVs-PD-L1 were reduced (Fig. 10B–D). The levels of EVs and EVs-PD-L1 released from LSD1-KO cells were also significantly lower than those from A549 cells (Fig. 10E). Moreover, the concentration of A549-KO EVs decreased by 50% relative to A549 EVs (Fig. 10F). Additionally, GW4869 induced the accumulation of membrane PD-L1 in LSD1-knockout A549 cells (Fig. 10G, Fig. S41B). Taken together, these results indicated that PD-L1 is present in cell-derived EVs and that LSD1 abrogation decreases EVs secretion and the accumulation of EVs with PD-L1.

Figure 10.

Inhibition of LSD1 reduces the expression of EVs and EVs-PD-L1. (A) TEM images of purified EVs from A549 cells in the presence or absence of LSD1 (A-EVs and KO-EVs, for short). Scale bar = 100 nm. (B) Expression of total PD-L1 in A549 cells treated with GW4869 and quantitative results. (C) Expression of membrane PD-L1 in A549 cells treated with GW4869 and quantitative results. (D) Expression of PD-L1, ALIX, TSG101, and CD9 in EVs, from the same number of cells in the presence of 10 μmol/L GW4869 or not. (E) Expression of PD-L1 in A-EVs and KO-EVs. (F) Concentration of A-EVs and KO-EVs from the same number of cells. (G) Expression of membrane PD-L1 in A549 or A549 KO cells treated with or without 10 μmol/L GW4869. (H) Representative HCS images of Dil-stained A-EVs and KO-EVs with wells coated with recombinant PD1. Scale bar = 200 μm. (I) HCS images of the binding of T cells to A-EVs and KO-EVs. Scale bar = 20 μm. Data are presented as mean ± SD (n = 3). ∗P < 0.05, ∗∗P < 0.01 and ∗∗∗P < 0.001.

To further verify the role of EVs-PD-L1 in the TME, a PD1/PD-L1 binding assay was performed to detect the ability of these EVs-PD-L1 to bind to PD1. Dil-labeled A-EVs bound to recombinant PD1-coated wells, and this binding was significantly abrogated in LSD1-KO cells (Fig. 10H, Fig. S41C). In line with these findings, the results of the HCS imaging analysis also revealed that Dil-labeled A-EVs and FITC-labeled CD4⁺T cells were colocalized, and the fluorescence intensity of A-EVs was 2.06 times greater than that of KO-EVs (Fig. 10I, Fig. S41D and S41E), indicating that EVs could directly bind to T cells, whereas LSD1-KO attenuated this interaction.

3.11. LSD1 remotely mediates immune escape by regulating EVs secretion

To further investigate whether EVs-PD-L1 have a regulatory immunosuppressive effect on tumor development, we co-cultured EVs with T cells and performed flow cytometry analysis. The results revealed that the addition of an anti-CD3/CD28 antibody significantly increased the expression of the T cell activation marker CD69 and that the addition of A-EVs downregulated the expression of CD69, whereas KO-EVs or PD-L1 blockade with the anti-PD-L1 antibody restored the expression of CD69 in T cells (Fig. 11A, Supporting Information Fig. S42A). These results suggested that EVs-PD-L1 can also exert immunosuppressive effects, whereas LSD1-KO can attenuate these effects. To further confirm that EVs can regulate the cytotoxicity of T cells, activated CD4⁺ T cells were cocultured with A549 cells and treated with A-EVs or KO-EVs. The results (Fig. 11B, Fig. S42B) revealed that activated CD4⁺ T cells could kill tumor cells with a cell survival rate of 14.20%, that the presence of EVs could reduce the effect and maintain tumor cell survival, and that A-EVs could increase cell survival to 62.79%, whereas KO-EVs had no significant effect on T cell killing due to the low expression of PD-L1. Moreover, the secretion of IL-2, IFN-γ, and TNF-α can also be inhibited by A-EVs and increased by KO-EVs (Fig. 11C). These findings consistently demonstrated that EVs could directly inhibit TCR-mediated T cell activation and killing.

Figure 11.

LSD1 remotely mediates immune escape by regulating the secretion of EVs. (A) Expression of CD69 in CD4⁺ T cells incubated with EVs. NT, no treatment. (B) Cell survival of A549 cells co-incubated with activated CD4⁺ T cells in the presence of EVs. Scale bar = 200 μm. (C) ELISA analysis of TNF-α, IFN-γ, and IL-2 concentrations in CD4⁺ T cells in the presence of EVs. (D) HCS images of A549 cells after being treated with EVs. Scale bar = 20 μm. (E) Expression of membrane PD-L1 in LLC cells treated with EVs. (F) PD-L1 expression in A549 cells following incubation with A-EVs. (G, H) HCS images of PD-L1 in A549 cells and LLC cells after being incubated with A-EVs, L-EVs and KO-EVs. Scale bar = 50 μm. Data are presented as mean ± SD (n = 3). ∗P < 0.05, ∗∗P < 0.01 and ∗∗∗P < 0.001.

Moreover, cancer cell-derived EVs are key players in cell-to-cell communication in the TME^84,85. Therefore, we further explored whether EVs can deliver PD-L1 to cancer cells and increase their immune escape ability. A Dil probe was used to label EVs-PD-L1 and incubated with A549 cells, and the results revealed that EVs-PD-L1 can be transported into target cancer cells (Fig. 11D, Supporting Information Fig. S43A and S43B). As expected, A-EVs can induce the accumulation of membrane PD-L1 in target cells, whereas PD-L1 accumulation was significantly reduced after treatment with KO-EVs, this result was also confirmed in LLC cells (Fig. 11E, Fig. S43C and S43D). A similar effect was observed for total PD-L1 levels (Fig. 11F, Supporting Information Fig. S44A). The fluorescence intensity of PD-L1 expressed by A549 or LLC cells was significantly increased by the addition of A-EVs or L-EVs, indicating that EVs could increase the content of PD-L1 in cells, whereas the EVs extracted after LSD1 knockout had no significant effect on recipient cells (Fig. 11G and H, Fig. S44B). T cell killing analysis also revealed that A549 cells treated with A-EVs tended to escape T cell killing, whereas KO-EVs had no significant effect (Supporting Information Fig. S45). These results further suggested that the fusion of EVs with cells can not only increase the expression of PD-L1 but also increase their ability to undergo immune escape, whereas LSD1 deletion can attenuate this effect.

4. Conclusions

In summary, we developed a TME-targeted synergistic therapy system, FPNs, which combine immune checkpoint blockade therapy with LSD1 inhibitors. Drug delivery in the form of liposomes ensures the maximum utilization of drugs and the stability of small-molecule compounds and can be used to construct synergistic multinetwork immunotherapies for tumors. After the FPNs entered the TME, anti-PD1 and FA were released under the action of the pH-sensitive compound PEOz. Anti-PD1 bound to PD1 on the surface of CD4⁺ T cells and blocked its binding to PD-L1 on the surface of tumor cells to reduce the occurrence of immune escape. Concurrently, FA was absorbed by tumors and inhibited the demethylase activity of overexpressed LSD1 in tumor cells, which ultimately reduced the secretion of EVs by tumor cells and the expression of PD-L1 in tumor cells in the TME with the help of the EVs-mediated cell surface PD-L1 rearrangement mechanism, inhibiting the long-range attack of EVs, and increasing the immune killing activity of T cells. In addition, we verified that CD4⁺ T cells play an indispensable role in the TME by arresting tumor cells in the G1/S phase to kill them directly. Our findings highlight the critical role of LSD1 in tumor immunotherapy, underscore the function of EVs as key signaling hubs bridging tumors and T cells, and reveal the direct contribution of CD4⁺ T cells to tumor clearance. This study provides a promising comprehensive strategy for the development of multinetwork synergistic tumor immunotherapy.

Author contributions

Yu Zhao: Conceptualization, Data curation, Formal analysis, Methodology, Software, Visualization, Validation, Writing-original draft, Writing-review & editing; Danyang Zhang: Conceptualization, Data curation, Methodology, Software; Lin Meng: Data curation, Methodology, Software; Siming Shan: Formal analysis, Software; Chao Zhang: Conceptualization, Resources, Supervision, Writing-review & editing; Zhenzhong Deng: Data curation, Methodology; Chao Han: Conceptualization, Investigation, Project administration, Resources, Supervision, Writing-review & editing; Lingyi Kong: Funding acquisition, Project administration, Resources, Supervision, Writing-review & editing.

Conflicts of interest

The authors have no conflicts of interest to declare.

Appendix A. Supporting information

The following is the Supporting Information to this article: