Abstract Non-small cell lung cancer (NSCLC) with PIK3CA mutations demonstrates significant challenges in treatment due to enhanced bone metastasis and immune checkpoint resistance. This study investigates the efficacy of tumor-targeting peptide 1-modified cancer stem cell-derived extracellular vesicles (TMTP1-TSRP-EVs) in reshaping the tumor microenvironment and reversing immune checkpoint resistance in NSCLC. By integrating TMTP1-TSRP into EVs, we aim to specifically deliver therapeutic agents to NSCLC cells, focusing on inhibiting the PI3K/Akt/mTOR pathway, a crucial driver of oncogenic activity and immune evasion in PIK3CA-mutated cells. Our comprehensive in vitro and in vivo analyses show that TMTP1-TSRP-EVs significantly inhibit tumor growth, reduce PD-L1 expression, and enhance CD8^+ T cell infiltration, effectively reversing the immune-suppressive microenvironment. Moreover, the in vivo models confirm that our approach not only suppresses bone metastases but also overcomes primary resistance to immune checkpoint inhibitors by modulating the expression of key immunological markers. These findings suggest that targeted delivery of TMTP1-TSRP-EVs could provide a novel therapeutic strategy for treating PIK3CA-mutant NSCLC, offering significant improvements over traditional therapies by directly targeting the molecular pathogenesis of tumor resistance and metastasis. graphic file with name 41419_2025_7685_Figa_HTML.jpg [34]Open in a new tab Molecular Mechanisms Reshaping the TME to Halt PI3K-Mutant Bone Metastasis of NSCLC and Overcoming Primary ICI Resistance. (Created by BioRender). Subject terms: Cancer, Cell biology Introduction Non-small cell lung cancer (NSCLC) is a significant contributor to cancer-related mortality worldwide, with high incidence and mortality rates [[35]1–[36]3]. Recent studies have highlighted the crucial role of the tumor microenvironment (TME) in the initiation, progression, and development of resistance to treatment in tumors [[37]4, [38]5]. The TME comprises various components such as tumor cells, immune cells, stromal cells, and blood vessels. These components interact through intricate signaling networks to regulate tumor growth and dissemination [[39]6, [40]7]. Particularly, alterations in the immune microenvironment significantly impact immune evasion by tumors and resistance to treatment [[41]8–[42]10]. While traditional cancer therapies mainly target tumor cells, emerging research indicates that modulating the TME can significantly enhance treatment efficacy [[43]11]. Consequently, targeting the TME has become a vital direction in cancer research [[44]12–[45]14]. The abnormal activation of the PI3K/Akt/mTOR signaling pathway has been extensively studied in various cancers, particularly in NSCLC [[46]15–[47]17]. Mutations in the PIK3CA gene represent a common mechanism of activation in this pathway, closely linked to tumor cell proliferation, survival, invasion, and metastasis [[48]18–[49]20]. Studies have shown that PIK3CA mutations not only promote tumor growth and metastasis but also correlate closely with immune evasion and resistance to immune checkpoint inhibitors (ICIs) [[50]21]. PIK3CA mutations impact the tumor immune microenvironment through various mechanisms, including facilitating the infiltration of immune-suppressive cells and impairing the function of effector T cells [[51]22]. These findings suggest that targeted therapeutic strategies against PIK3CA mutations hold promise in overcoming the limitations of current treatment methods and enhancing treatment efficacy. Cancer stem cells (CSCs) play a crucial role in tumor initiation, progression, and treatment resistance [[52]23]. CSCs possess the ability to self-renew and differentiate, forming new tumor cells to sustain tumor growth and metastasis [[53]24, [54]25]. Recent research has unveiled that CSCs regulate the TME by secreting extracellular vesicles (EVs) [[55]26–[56]28]. EVs, membrane-bound vesicles with diameters ranging from 30 to 150 nm, contain biological molecules such as proteins, RNA, and miRNA, enabling intercellular information transfer [[57]29–[58]31]. Utilizing CSC-derived EVs as drug carriers enhances drug targeting and therapeutic efficacy, offering novel avenues and approaches for anti-cancer therapy [[59]32–[60]34]. The transformable specific-responsive peptide (TSRP) is a novel peptide capable of specifically recognizing and targeting tumor cells, inhibiting tumor growth and metastasis by regulating key signaling pathways [[61]35]. TSRP can suppress tumor cell proliferation and migration by downregulating the activity of the PI3K/Akt/mTOR signaling pathway. Furthermore, TSRP enhances tumor immune surveillance by increasing the infiltration of CD8^+ T cells and inhibiting the expression of PD-L1, thereby reversing immune checkpoint resistance [[62]35]. These findings provide a strong theoretical basis for the application of TSRP in NSCLC. This study aims to investigate the mechanism of action of Tumor-Targeting Peptide 1-modified cancer stem cell-derived extracellular vesicles encapsulating transformable specific-responsive peptide (TMTP1-TSRP-EVs) in reshaping the TME, preventing bone metastasis of PI3K-mutant NSCLC, and reversing primary immune checkpoint resistance. Initially, we utilized bioinformatics analysis and CRISPR/Cas9 technology to establish PIK3CA mutant and wild-type NSCLC cell lines, identifying the key pathways and genes targeted by TSRP therapy. Subsequently, we isolated CSCs and their secreted EVs through sphere formation assays, followed by the construction of TMTP1-TSRP-EVs using chemical conjugation and electroporation techniques. Finally, both in vitro and in vivo experiments were conducted to validate the reshaping effect of the TME and the impact on bone metastasis of PI3K-mutant NSCLC and immune checkpoint resistance. This study not only aims to elucidate the molecular mechanisms underlying the interaction between PI3K mutations and the immune microenvironment but also provides novel insights and approaches for the targeted therapy of NSCLC, holding significant scientific and clinical implications. By comprehensively understanding the mechanism of action of TSRP and its clinical potential, we aspire to offer more effective treatment strategies for NSCLC patients, ultimately improving their prognosis and survival rates. Materials and methods Acquisition and analysis of NSCLC gene expression and mutation data The single-nucleotide polymorphism (SNP) mutation data of TCGA-Lung Adenocarcinoma (LUAD) and TCGA-Lung Squamous Cell Carcinoma (LUSC) datasets were retrieved from the Cancer Genome Atlas (TCGA) database ([63]https://portal.gdc.cancer.gov/). Specifically, the “Masked Somatic Mutation” data was selected for download, encompassing a total of 1100 cases of NSCLC patients (including TCGA-LUAD and TCGA-LUSC), providing insights into the somatic mutation landscape. Additionally, RNA-seq data (NSCLC-Counts data) from the TCGA-NSCLC dataset, comprising all available NSCLC samples (N = 1016) and normal tissues (N = 108), were obtained. Clinical information essential for the study, such as age, gender, TNM staging, histological grading, and overall survival (OS) status, was also acquired from TCGA. In addition, we obtained somatic gene mutation information for 100 samples of LUSC-KR from the International Cancer Genome Consortium (ICGC) database ([64]https://dcc.icgc.org/) up to November 27, 2019. Only patients with complete clinical data were included, while those with missing information on TNM staging, gender, age, and survival status were excluded. All bioinformatics statistical analyses were carried out using the R software version 4.2.1 and the corresponding software packages. For the analysis and visualization of TCGA mutation annotation format (MAF) files, the “maftools” package in R was utilized. Similarly, the analysis and visualization of mutation data from the IICGC were conducted using the “GenVisR” package in R. Subsequently, the Perl programming language was employed to extract highly mutated genes from TCGA and ICGC databases, and the intersection was obtained using the “venn” tool to identify genes with high mutation frequencies. Following this, based on the mutation severity of the mutated genes, the samples were categorized into wild-type and mutant groups. The relationship between these intersected genes and tumor mutation burden (TMB) was then assessed and visualized using the “ggpubr” package in R. TMB calculation TMB refers to the total number of somatic mutations detected per million bases (mutations per Mb). In this study, we utilized whole-exome sequencing (WES) data of NSCLC from the TCGA database to calculate TMB scores. The Perl language was employed to compute the mutation frequency for each sample, defined as the number of mutations in each sample divided by the exome length (38 million). Analysis of the tumor immune microenvironment in NSCLC samples Using the TIMER database, we inferred the abundance of infiltrating CD4^+ T cells, CD8^+ T cells, B cells, macrophages, neutrophils, and dendritic cells (DC) in NSCLC samples. Cell culture The human NSCLC cell lines A549 (CCL-185) and H1703 (CRL-5889) were both obtained from ATCC (USA), while human CD8^+ T cells (1506) were sourced from LDEBIO (China). A549, H1703, and CD8^+ T cells were cultured in RPMI-1640 medium (A1049101, Gibco, USA) containing 10% FBS (12484028, Gibco, USA) and 1% penicillin/streptomycin (15140148, Gibco, USA). Before experimentation, CD8^+ T cells were stimulated for 24 h with 2 µL of CD3/CD28 activator (11161D, Gibco, USA). When investigating the impact of primary immune checkpoint resistance, Nivolumab (200 μg/mL; T9907, TargetMol, USA) was introduced into the cell culture medium. For functional validation of TSRP, 2 mg/mL of TSRP (LSPPRYPCKLVFFPLGVRGKKWWKK-Dip-K-NH2) purchased from Hangzhou Dangang Biochemical Technology Co., Ltd. (China) was added to the cell culture medium for subsequent experiments. In mechanistic studies, 4 μg/mL of SC79 (HY-18749) obtained from MedChemExpress (USA) was added to the cell culture medium [[65]36]. 293 T cell line was obtained from ATCC (CRL-3216) and cultured in DMEM medium (11965092, Gibco, USA) containing 10% FBS, 10 μg/mL streptomycin, and 100 U/mL penicillin. The cells were maintained in a humidified cell culture incubator (Heracell™ Vios 160i CR CO[2] incubator, 51033770, Thermo Scientific™, Germany) at 37 °C with 5% CO[2]. Passaging was performed when the cell growth reached 80%–90% confluency [[66]37]. Cell co-culture NSCLC cells A549/H1703 were co-cultured in vitro with CD8^+ T cells at a ratio of 1:5. The co-culture was continued for 48 h, followed by the collection of supernatant and NSCLC cells. Flow cytometry cell sorting of NSCLC cells was performed using the EpCAM antibody for further experiments [[67]38–[68]41]. The cell co-culture groups were as follows: (1) DMSO + PBS group: CD8^+ T cells + A549/H1703 cells co-cultured with DMSO + PBS; (2) SC79 + PBS group: CD8^+ T cells + A549/H1703 cells co-cultured with 4 μg/mL SC79 + PBS; (3) SC79 + TSRP group: CD8^+ T cells + A549/H1703 cells co-cultured with 4 μg/mL SC79 + 2 mg/mL TSRP; (4) EVs group: CD8^+ T cells + A549/H1703 cells co-cultured with CSCs-EVs; (5) TMTP1-EVs group: CD8^+ T cells + A549/H1703 cells co-cultured with TMTP1-EVs; (6) TMTP1-EVs group: CD8^+ T cells + A549/H1703 cells co-cultured with TMTP1-EVs; (7) TSRP-EVs group: CD8^+ T cells + A549/H1703 cells co-cultured with TSRP-EVs; (8) TMTP1-TSRP-EVs group: CD8^+ T cells + A549/H1703 cells co-cultured with TMTP1-TSRP-EVs. Prior to the co-culture of CD8^+ T cells and A549/H1703 cells, 20 μg/mL of EVs from each group were added to the A549/H1703 cell culture medium for 24 h of co-incubation. DMSO, SC79, PBS, and TSRP solutions were added to the culture medium during the co-culture of CD8^+ T cells and A549/H1703 cells. Generation of cells overexpressing wild-type PIK3CA and mutant PIK3CA overexpression cells Cells overexpressing PIK3CA were generated using lentiviral transfection technology. The lentiviral vectors used were constructed by Genechem (Shanghai, China). The lentivirus overexpressing PIK3CA was packaged in the pLenti-RFP vector. These vectors were transduced into cells via lentivirus transduction and selected with puromycin (400051, Sigma–Aldrich, USA) to generate stable cells overexpressing PIK3CA. Using the CRISPR/Cas9 editing system, we generated the PIK3CA-E545K (c.1633 G > A) mutation. The mutation vector used was pCIG PIK3CA-E545K (73055, Addgene, USA), while the wild-type vector was pCIG PIK3CA Wildtype (73056, Addgene, USA). Targeted mutagenesis of p.E545K was performed in A549 or H1703 cells using the QuickChange II site-directed mutagenesis kit (200523, Agilent, USA) to produce the PIK3CA p.E545K variant and the PIK3CA wild-type (Fig. [69]S1). Surviving cells were obtained through restrictive dilution cloning, and PIK3CA-E545K heterozygous and homozygous mutant cells were identified via DNA sequencing. Specifically, we obtained A549 wild-type and homozygous mutant cells (A549 WT/^−; A549^−/MUT) and H1703 wild-type and heterozygous mutant cells (H1703 WT/WT; H1703 WT/MUT). Fluorescence in situ hybridization (FISH) FISH analysis was performed on formalin-fixed, paraffin-embedded NSCLC tissues using the Abbott-Vysis HER2/CEP17 dual-color probe (Abbott, USA, 05N56-020). Initially, tissue sections embedded in paraffin and deparaffinized with xylene were rehydrated in different ethanol gradients (100%, 85%, and 70%). Subsequently, the sections were treated with proteinase K solution (200 μg/mL; 25530015, Invitrogen, USA) and pepsin (0.005% in 0.01 M HCl solution; P7012, Sigma–Aldrich, USA) for further processing. The slides were dehydrated in different ethanol gradients (70%, 85%, and 100%), and the probe mixture was added to the slides. Subsequently, coverslips were immediately placed on top, and the edges were sealed with rubber cement. The slides were denatured at 85 °C for 5 min and then incubated overnight at 37 °C. Following hybridization, FISH signals from 20 to 30 cells were counted. The standard definition for PIK3CA amplification was set as FISH signal ≥2.2 compared to the control probe. Fluorescence images were captured using an Olympus BX43 microscope (Olympus, Tokyo, Japan) under FITC and Texas Red wavelengths, and image processing was carried out using Gene Data Manager 7.2.7.33397. Characterization of TSRP TSRP peptide was subjected to characterization experiments by adding 15 μg/mL of MMP-2 (HY-[70]P73296, MedChemExpress, USA) into the culture medium. For the HPLC analysis, a 4.6 × 250 nm Sinochrom ODS-BP column (5 μm) was utilized. The mobile phase consisted of 0.1% trifluoroacetic acid in 100% acetonitrile as solvent A and 0.1% trifluoroacetic acid in 100% water as solvent B. The retention time (RT) was observed at a wavelength of 220 nm with a flow rate of 1.0 mL/min. Turbidity testing involved measuring the absorbance of different solution groups at 500 nm wavelength over a duration of 5 h, with readings taken every 15 s. Ultimately, data was collected using GraphPad software. CCK-8 experiment for cell viability assessment CD8^+ T cells were digested, resuspended, and adjusted to a concentration of 1 × 10^5 cells/mL. Subsequently, 100 μL of cell suspension was seeded into a 96-well plate for standard cultivation. After cell adhesion, drugs were added for treatment, followed by overnight incubation. At 0, 12, 24, and 48 h post-incubation, cell viability was evaluated according to the instructions of the CCK-8 assay kit (C0041, Beyotime, Shanghai). During each assessment, 10 μL of CCK-8 detection solution was added, and the plate was then incubated in a cell culture incubator for 4 h. The absorbance at 450 nm was measured using an enzyme-linked immunosorbent assay (ELISA) reader to calculate cell viability, utilizing the formula: Cell Viability = (ΔA_sample - ΔA_blank) / (ΔA_control - ΔA_blank), where ΔA_sample represents the absorbance difference of the sample, ΔA_blank is the absorbance difference of the blank, and ΔA_control is the absorbance difference of the control group. Transwell experiment for assessing cell migration ability NSCLC cells transfected for 48 h were collected and suspended in a serum-free medium at a concentration of 10^5 cells per well. Subsequently, 200 μL of cell suspension (2 × 10^4 cells/well) was seeded in the upper chamber of Transwell plates, while 800 μL of medium containing 20% FBS was added to the lower chamber. Following a 24-h incubation at 37 °C, the Transwell plates were removed, washed twice with PBS, fixed with formaldehyde for 10 min, and then rinsed three times with distilled water. The cells were stained with 0.1% crystal violet, incubated at room temperature for 30 min, and washed twice with PBS, and the migrated cells were photographed using an inverted light microscope (CKX53, Olympus, Japan). Cell counting and analysis of cancer cell migration ability were performed using ImageJ software. TUNEL staining NSCLC cells from each group were fixed with 4% paraformaldehyde at room temperature for 15 min and then permeabilized with 0.25% Triton X-100 for 20 min. Samples were blocked with 5% bovine serum albumin (BSA, 36101ES25, Yeasen Biotechnology (Shanghai) Co., Ltd., China) and subsequently stained with TUNEL (C1086, Beyotime Biotechnology Co., Ltd, Shanghai, China) reagent. DAPI staining solution (C1002, Beyotime Biotechnology Co., Ltd, Shanghai, China) was applied to counterstain in the dark. Apoptotic cell images were captured under a confocal microscope (LSM 880, Carl Zeiss AG, Germany). TUNEL-positive cells (green fluorescence) indicated apoptotic cells, while DAPI-labeled cell nuclei emitting blue fluorescence represented the total cell count. The apoptotic cell rate was determined by calculating the ratio of apoptotic cells to total cells in five different fields per group, expressed as a percentage: apoptotic cell rate = (number of apoptotic cells/total cell count) × 100%. In vivo animal experimentation In this animal study, female non-obese diabetic/severe combined immunodeficiency (NOD-SCID) mice aged 6–8 weeks (obtained from Huxley-Janda Experimental Animals Co., Ltd, Hunan, China) were utilized to establish a metastatic tumor model by directly injecting 1 × 10^7 pretreated and co-cultured A549/H1703 cells into the left ventricle of the mouse heart. The specific procedure involved anesthetizing the mice, making two transverse incisions towards the axilla to create a “Y” shaped incision, and accessing the thoracic cavity. Gently displacing the lung tissue revealed the heart, where a cell suspension was slowly injected into the left ventricle using a micro syringe or injection needle [[71]42]. To monitor tumor growth in real time, tumor growth was monitored promptly and documented photographically, and mice forming bone metastatic lesions were selected for further experiments. Tumor volume was assessed every other day, and 1 × 10^5 human peripheral blood mononuclear cells (hPBMCs; PCS-800-011, ATCC, USA) were intravenously injected via the tail vein (Fig. [72]S2). On days 13 and 23, flow cytometry was employed to measure the concentration of human CD45 in the mouse blood [[73]42–[74]44]. Bioluminescence imaging data were acquired using the in vivo imaging system (IVIS) Spectrum CT system (PerkinElmer, USA). Micro-CT data were obtained utilizing the vivaCT 80 system (Scanco, Switzerland). For the EdU labeling assay, each mouse was intraperitoneally injected with 100 μg of EdU ([75]C10640, Thermo Fisher, USA) 24 h prior to bone retrieval. The Click Plus EdU 647 imaging kit ([76]C10419, Thermo Fisher, USA) was employed for EdU staining. Tumor cells were stained with GFP antibody, and the ratio of EdU-positive cells to GFP-positive cells was calculated [[77]45]. The mice were randomly divided into 22 groups, each consisting of 6 mice: (1) PBS group; (2) TSRP group; (3) DMSO + PBS group; (4) SC79 + PBS group; (5) SC79 + TSRP group; (6) Blank group; (7) EVs group; (8) TMTP1-EVs group; (9) TSRP + DIR group; (10) TSRP-EVs group; (11) TMTP1-TSRP-EVs group. Each group was injected with A549 cells or H1703 cells for modeling. When the diameter of the metastatic tumor reached 100 mm^3, 80 µL of PBS/TSRP (14 mg/kg) or 100 µL of 100 µg EVs labeled with DIR (40757ES25, YEASEN, China) was intravenously injected into the mice [[78]46, [79]47]. EVs were administered three times per week, while PBS/TSRP was injected every two days for a total of three injections. After injections, whole-body fluorescence imaging was conducted at 2, 4, 6, 8, and 24 h using a near-infrared dual-zone small animal IVIS (Photon). At 24 h post-injection, the liver, spleen, kidneys, heart, lungs, and tumors were dissected, and the fluorescence intensity of each organ or tissue was measured. Mice in the SC79 + PBS group and SC79 + TSRP group received intraperitoneal injections of a 10 mg/kg SC79 solution (an AKT pathway activator), while the DMSO + PBS group received DMSO solution, administered three times per week [[80]48]. The biodistribution of EVs was monitored for 30 days using the Kodak imaging system, and DiR-labeled EVs were analyzed using the Kodak Image System to determine their distribution [[81]49]. In studying the impact of primary immune checkpoint resistance, mice were treated with Nivolumab (10 mg/kg, three times a week) via tail vein injection [[82]49, [83]50]. Intra-tumoral tissue imaging: The collected tumor tissue was subjected to ex vivo fluorescence imaging, fixed in 4% paraformaldehyde for 24 h, placed in a 15% sucrose PBS solution for 24 h until sedimentation, and then transferred to 30% sucrose for another 24 h until sedimentation. Subsequently, the tumor tissue was frozen and sliced into 20 μm sections, followed by staining with 1 mg/mL DAPI for 10 min at room temperature. After washing twice with PBS (pH 7.4), the sections were immediately examined under a laser scanning confocal microscope (LSM 700, Carl Zeiss Microscopy, Germany) [[84]51, [85]52]. Histological staining techniques For Hematoxylin and Eosin (H&E) staining, tissue samples are first fixed and then sectioned. The paraffin is removed by slicing the wax blocks in xylene, followed by dehydration in 100%, 95%, and 70% ethanol, rehydration, and either mounting or rinsing with water. The prepared sections are immersed in Hematoxylin staining solution (H8070, Solarbio, Beijing, China) for 5–10 min at room temperature. Subsequently, the sections are quickly differentiated in 1% hydrochloric acid ethanol solution for 10 s, rinsed with distilled water, dehydrated in 95% ethanol, and placed in Eosin staining solution (G1100, Solarbio, Beijing) for 5–10 min. After standard dehydration, clearing, and mounting, the slides are observed under an optical microscope. Immunohistochemistry (IHC) staining The antibody list is provided in Table [86]S1. Mouse tumor tissue is fixed in 4% paraformaldehyde overnight, followed by embedding in paraffin and sectioning at a thickness of 4 μm. Deparaffinization is accomplished using xylene, and hydration is achieved through a series of ethanol washes (anhydrous ethanol, 95% ethanol, and 75% ethanol for 3 min each). Subsequently, the sections are subjected to antigen retrieval by boiling in 0.01 M citrate buffer for 15–20 min, followed by a 30-min room temperature incubation in 3% H[2]O[2] to inactivate endogenous peroxidases. The sections are then treated with goat serum blocking solution, incubated at room temperature for 20 min, and excess fluid is removed. First, the primary antibodies are added and left to incubate at room temperature for 1 h, followed by washing with PBS. Following the addition of the primary antibody, the samples were incubated at room temperature for 1 h, washed with PBS, and then incubated with secondary IgG anti-rabbit antibody for 20 min at 37 °C. Subsequently, the samples were washed with PBS and incubated with streptavidin-peroxidase (SP) at 37 °C for 30 min, followed by another PBS wash. The DAB substrate (P0202, Beyotime Biotechnology Co., Ltd) was then added for 5–10 min for color development, followed by a 10-min water rinse to stop the reaction. Counterstaining with hematoxylin (C0107, Beyotime Biotechnology Co., Ltd) was performed for 2 min, followed by differentiation in a hydrochloric acid alcohol solution and a 10-min water rinse prior to dehydration in graded alcohols (xylene transparent) and sealing with 2–3 drops of neutral resin. For analysis, under a light microscope, five random high-power microscope fields were selected per slide, with 100 cells counted in each field to calculate the percentage of positive cells. Flow cytometry Flow cytometry was used to detect the levels of macrophages and CD8^+ T cells. Macrophages derived from THP-1 in a co-culture model were collected by centrifugation at 1200 × g for 5 min at 4 °C, followed by resuspension in staining buffer. Single-cell suspensions from macrophages or tumor tissues were incubated in the dark at 4 °C for 30 min, and cells were dissociated using StemPro™ Accutase™ (A1110501, Gibco, USA) cell dissociation reagent. The cell pellets were then centrifuged at 1000 × g/min for 5 min and washed twice with PBS. Subsequently, cells were resuspended in 100 μL of PBS and stained with antibodies from Table [87]S2. After antibody incubation, cells were washed three times with PBS and promptly analyzed using a flow cytometer (Beckman, USA). High-throughput sequencing and analysis of mRNA in tumor tissues of the NSCLC mouse bone metastasis model Tumor tissue samples from PI3K-mutated A549 mice metastasized to the bone in the PBS group and the TSRP group (6 mice each) were randomly selected. Total RNA was isolated from the 12 samples using the total RNA isolation kit (12183555, Invitrogen, USA), and the quantity of total RNA was quantified by measuring the OD value with a UV spectrophotometer (BioSpectrometer basic, Eppendorf, USA). The integrity of these total RNAs was assessed using agarose gel electrophoresis. High-quality total RNA was reverse transcribed into cDNA to construct RNA libraries, which were sequenced using Illumina’s NextSeq 500 platform. The raw image data obtained from sequencing was converted into raw reads through base calling. To ensure the quality of raw reads, cutadapt was used to remove sequencing adapter sequences and filter out low-quality sequences, resulting in “clean reads.” These clean reads were aligned to the human reference genome using Hisat2 software, and then gene expression was quantified using the R software package to generate a gene expression matrix. The “limma” package in R was used to identify differentially expressed genes in the high-throughput sequencing data, with the criteria defined as |log2FC | > 1 & p-value < 0.05. A volcano plot was generated using the ggplot2 package, and a heatmap was created using the pheatmap package. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis was performed using the Xiantao Academic Database ([88]https://www.xiantaozi.com/). The CIBERSORT package was utilized to analyze the proportions of 22 immune cell types in the 12 samples, with a filtering condition of p < 0.05, to obtain the immune cell proportion matrix data for each tumor sample. Bar plots and violin plots were generated using the “vioplot” and “ggpubr” packages in R to illustrate the relationship between gene mutations and tumor invasion of the immune system. Isolation and identification of CSCs The A549 cell line was cultured in a high-glucose DMEM medium (11965126, Gibco, USA) supplemented with 10% FBS. Spheroid-enriched (SE) cell lines derived from the aforementioned A549 cells were cultured under the same conditions. Two key elements of the enrichment method included employing adherent culture conditions and repeatedly selecting cells with anchorage-independent spheroid growth capability. When the monolayer culture reached 90% confluence, floating individual cells or spheroids were collected, resuspended, and immediately replated until subconfluency. After repeating this process eight times, the culture was enriched with spheroid-forming cells that remained suspended under adherent culture conditions, termed as stem cell-like cell lines. Subsequently, CSCs could be obtained by flow cytometry sorting for CD44 and CD133 double-positive cells. Isolation and preparation of EVs When CSC fusion reached 80%–90% confluence, the supernatant was discarded, and cells were washed with 2× PBS. Subsequently, 25 mL serum-free IMDM medium (12440053, Gibco, USA) was added to each culture flask, and cells were further incubated for 48 h at 37 °C, 5% CO[2] in a humidified atmosphere. The cell supernatants were collected into 50 mL centrifuge tubes and centrifuged at 300 × g for 10 min at 4 °C to remove cell debris. The supernatant was then transferred to another 50 mL centrifuge tube, and upon collection, immediate isolation of EVs was conducted. At 4 °C, the supernatant was centrifuged at 2000 × g for 20 min to transfer it to sterile tubes for use in an ultracentrifuge, followed by centrifugation at 16,500 × g for 30 min at 4 °C. The supernatant was transferred to ultracentrifuge tubes and centrifuged at 120,000 × g for at least 70 min at a fixed-angle rotor at 4 °C. The supernatant was thoroughly discarded. Subsequently, 1 mL of 4 °C PBS was added to each ultracentrifuge tube, and the pellet was resuspended using a micropipette. The solutions from the same group were mixed in the ultracentrifuge tube, followed by the addition of 4 °C PBS to exceed three-quarters of the tube volume. Centrifugation at 120,000 × g for 60 min at 4 °C was conducted to remove excess supernatant, and the pellet was resuspended again with sterile PBS, yielding CSC-derived EVs. Using electroporation technology, TSRP was encapsulated into CSCs-EVs. The previously centrifuged CSCs-EVs pellet was resuspended in electroporation buffer containing 1.15 mM potassium phosphate (pH 7.2), 25 mM potassium chloride, and 21% OptiPrep working solution (D1556, Sigma–Aldrich, USA). The suspension of EVs was filtered through a 0.22 μm filter. Purified TSRP protein was then added to the EVs at a weight ratio of 1:5 to form EVs-TSRP complexes via electroporation using a Gene Pulser Xcell (Bio-Rad). Following electroporation, the EVs were centrifuged at 100,000 × g, 4 °C for 2 h, and the pellet was resuspended in cold PBS solution. The TMTP1 peptide (sequence: NVVRQ) was synthesized by Xi’an Huachen Biotechnology Co., Ltd. using Fmoc chemistry synthesis on a solid-phase synthesizer. The peptide was purified using high-performance liquid chromatography, and its sequence and structure were confirmed by mass spectrometry. To enable targeted, specific delivery, the 5′-COOH-modified TMTP1 peptide was covalently linked to the amine groups on the surface of EVs via EDC-NHS coupling, facilitating surface functionalization of the EVs for targeted delivery. To activate the carboxyl group of the TMTP1 peptide, a mixture of denatured (heated at 85–95 °C for 10 min) and re-natured (cooled to room temperature for 15 min) TMTP1 peptide (5 μM) was prepared with EDC (46 mg, 0.3 mmol; 22980, Thermo Scientific™, USA) and NHS (35 mg, 0.3 mmol; 24500, Thermo Scientific™, USA) for one hour. 1 mL of EV suspension (200 mg/mL) was added to DNase/RNase-free water. The stable NHS esters rapidly reacted with the amines present on the surface of EVs when gently stirred and left to incubate overnight at room temperature. To remove the unconjugated TMTP1 peptide, the reaction mixture was filtered through a 100 kDa cutoff Amicon centrifugal filter at 5000 × g for 30 min, followed by two washes with cold DNase/RNase-free water. Gel retardation assay confirmed the stability of the TMTP1-EVs complex, with free TMTP1 peptide as the positive control and wild-type EVs as the negative control. The concentration of unconjugated TMTP1 peptide in the supernatant was quantified using the NanoDrop™ One UV-Visible spectrophotometer (840-317400, Thermo Scientific™, USA) at 260 nm after filtration. Subsequently, the conjugation efficiency of the TMTP1 peptide was calculated using the following formula: Conjugation efficiency of TMTP1 peptide (%) = (initial amount of TMTP1 peptide added - amount of unconjugated TMTP1 peptide in the supernatant) / initial amount of TMTP1 peptide added × 100. The EVs were grouped as follows (Fig. [89]S3): (1) EVs group (EVs derived from wild-type CSCs); (2) TSRP-EVs group (CSCs-derived EVs encapsulating TSRP); (3) TMTP1-EVs group (CSCs-derived EVs successfully conjugated with TMTP1 peptide on their surface); (4) TMTP1-TSRP-EVs group (CSCs-derived EVs encapsulating TSRP and successfully conjugated with TMTP1 peptide on their surface). Characterization of EVs In order to identify the characteristics of the EVs under investigation, EVs were resuspended in RIPA lysis buffer (89901, Thermo Scientific™, Germany), and their specific markers, such as Alix, TSG101, and CD81, were detected using Western blot. The negative control protein marker used was calnexin. Antibodies used in this study included rabbit anti-Alix (ab275377, 1:1000), anti-TSG101 (ab133586, 1:1000), anti-CD81 (ab286173, 1:500), and anti-calnexin (ab92573, 1:20,000), all purchased from the UK-based company Abcam. By utilizing Nanoparticle Tracking Analysis (NTA), the size and concentration of EVs were determined. The EV samples were suspended in PBS and then diluted 500 times with Milli-Q water. Subsequently, the diluted EVs were injected into the sample chamber of the NanoSight LM10 (Malvern) instrument using a sterile syringe, ensuring the absence of air bubbles and filling the chamber completely. The NanoSight LM10 is equipped with a 640 nm laser and a fluoroelastomer O-ring (Viton). The videos were analyzed using NanoSight version 2.3 software (NanoSight Ltd, Amesbury, UK) with a gain set at 6.0 and a threshold of 11 to capture the particle trajectories. The software generated concentration and size distribution profiles of the diluted samples, from which the original EVs’ concentration was calculated based on the dilution factor. After centrifuging the resuspended EVs to form a pellet, the pellet was fixed in a fixative solution (2% paraformaldehyde, 2.5% glutaraldehyde) at 4 °C for 1 h. The fixed pellets underwent three washes with PBS (15 min each), followed by fixation in 1% osmium tetroxide for 1.5 h and another three washes with PBS (15 min each). The samples were dehydrated in a graded series of alcohols, infiltrated with epoxy resin overnight, embedded, and polymerized at 35 °C, 45 °C, and 60 °C for 24 h. After ultra-thin sectioning and lead-uranium staining, observation was conducted using a transmission electron microscope (JEM-1011; JEOL, Tokyo, Japan) at an accelerating voltage of 80 kV, with image capture performed using a side-mounted Camera-Megaview III (Soft Imaging System, Münster, Germany). Each experiment was repeated three times. Immunofluorescence staining to assess the uptake of EVs by NSCLC cells EVs derived from CSCs were seeded into a 24-well plate, and Dil dye (C1036, Beyotime Biotechnology Co., Ltd., Shanghai) was added to 40 μg of EVs to achieve a final concentration of 25 μM. The mixture was then allowed to react for 30 min at room temperature, followed by rapid centrifugation to remove unbound dye. Subsequently, the cells were washed three times with PBS and fixed with 4% paraformaldehyde (AR1068, BOSTER Biological Technology Co., Ltd., Wuhan) for 30 min. Finally, the cell nuclei were stained with DAPI (4′,6-diamidino-2-phenylindole, C1005, Beyotime Biotechnology Co., Ltd., Shanghai) for 30 min, and the cells were imaged at 400× magnification using a BX53 fluorescence microscope equipped with a camera (Olympus). Image analysis was performed using ImageJ Pro Plus 6.0 software. Relative expression level of target genes detected by RT-qPCR Total RNA was extracted from tissues or cells using Trizol reagent (15596026, Invitrogen, USA), and the concentration and purity of the total RNA were assessed at 260/280 nm using NanoDrop LITE (ND-LITE-PR, Thermo Scientific™, Germany). The extracted total RNA was reverse transcribed into cDNA using the PrimeScript RT reagent Kit with gDNA Eraser (RR047Q, TaKaRa, Japan). Subsequently, the expression of various genes was analyzed by RT-qPCR using SYBR Green PCR Master Mix reagents (4364344, Applied Biosystems, USA) and the ABI PRISM 7500 Sequence Detection System (Applied Biosystems). The primers for the genes were synthesized by TaKaRa (Table [90]S3), with GAPDH serving as the reference gene. The relative expression levels of the genes were analyzed using the 2^−ΔΔCt method, where ΔΔCt = (average Ct value of target gene in the experimental group - average Ct value of reference gene in the experimental group) - (average Ct value of target gene in the control group - average Ct value of reference gene in the control group). All RT-qPCR analyses were performed in triplicate. Western blot Initially, cells or tissues were collected and lysed using an enhanced RIPA lysis buffer containing protease inhibitors (P0013B, Beyotime Biotechnology Co., Ltd, Shanghai, China). Subsequently, the protein concentration was determined using a BCA protein quantification assay kit (P0012, Beyotime Biotechnology Co., Ltd, Shanghai, China). The proteins were separated by 10% SDS-PAGE and transferred to a PVDF membrane (FFP39, Beyotime Biotechnology Co., Ltd). The membrane was then blocked with 5% BSA (ST023, Beyotime Biotechnology Co., Ltd) at room temperature for 2 h to prevent nonspecific binding. After blocking, primary antibodies (rabbit anti-human, detailed information in Table [91]S4) were added at a diluted concentration and incubated at room temperature for 1 h. Following primary antibody incubation, the membrane was washed and incubated with an HRP-conjugated goat anti-rabbit secondary antibody (ab6721, 1:2000, Abcam, UK) at room temperature for 1 h. Pierce™ ECL Western blot substrate (32209, Thermo Scientific™, Germany) A and B solutions were mixed in equal amounts in a darkroom; the mixture was then added onto the membrane, and the membrane was exposed in a gel imager. The Western blot images were captured using the Bio-Rad imaging system (BIO-RAD, USA), and the bands of interest were quantified for grayscale using ImageJ analysis software, with GAPDH and β-Tubulin as internal references. Each