Collection of Tumor samples

Specimens from lung cancer surgeries, ranging in size from 1 to 2.5 cm3, were swiftly delivered to the research facility following excision. Comprehensive patient clinical data can be found in Supplementary Table S1.

Organoid cultureFibroblast culture

The fresh tumor sample was washed with cold PBS twice and then minced the tissue into 1–2 mm3 pieces with scissors. Collected minced tissue into a 15 ml centrifuge tube, added 10 ml collagenase I (3 mg/ml) and a ROCK inhibitor Y-27,632 dihydrochloride (10µM), and then incubated and digested the tissues in a 37 °C cell incubator for 2–3 hours. Filtered the digested tissue using a 70 µM cell strainer to isolate single-cell suspensions from undigested large clusters. Transferred the remaining undigested clusters to a fresh 10 cm cell culture dish containing 10 ml of DMEM, enriched with 10% FBS (Gibco-Invitrogen) and 50 U/mL penicillin/streptomycin (C0222, Beyotime). Moved the cell culture dish into a cell incubator with a 5% CO2 concentration. After a 5–10 days culture, fibroblasts could be found adhered to the cell culture dish and then refreshed the culture medium. Fibroblasts were passed at a ratio ranging from one to three to one to four every 3–5 days. Then we could collect the fibroblasts with TE buffer for lung cancer cells-fibroblasts co-culture.

Organoid culture with fibroblast/MSCs in Mi-gel

Mi-gel (G23001, Morga) and Mi-gel buffer B (E23001, Morga) were used as the organoids culture matrix material. The Mi-gel was put in a 37 °C incubator for melt, otherwise, the Mi-gel buffer B was put on ice. For the study, the suspension of cells and fibroblasts was acquired from the methods outlined within the “Fibroblast culture”. Then the cells were divided into three different groups, lung cancer cells group, lung cancer cells-fibroblasts co-culture group (1:1), and lung cancer cells-MSCs co-culture group (1:1). The cell mixture underwent a PBS wash and was subjected to centrifugation at 1500 rpm for 5 min to eliminate the PBS. This process was repeated twice, followed by an additional centrifugation step aimed at completely removing any remaining liquid. Resuspend cells with melted Mi-gel and Mi-gel buffer B were added, and the ratio of Mi-gel and Mi-gel buffer B was 20:1. The suspension was fully mixed with a pipette, and then dropped the mixture (~ 20 µl per drop) into 6-well cell culture plate, incubated the culture dish in 37 °C cell incubator with 5% CO2. Once solidified, added 2 ml of lung cancer organoid medium per well, and the medium was refreshed every 3 days. Lung cancer organoid medium was prepared as described in our previously published study [33].

For organoids passaging, we harvested the Mi-gel drops with a cell scalper and smashed the Mi-gel drops into small pieces with 1 ml pipette tips, then collected the Mi-gel and cell pellet into a 15 ml tube and centrifuged at 1500 rpm for 5 min. Washed the Mi-gel and cell pellet with PBS and then centrifuged at 1500 rpm for 5 min. Removed the PBS and digested the Mi-gel and cell pellet with organoids digestion solution (D23001, Morga) at 37 °C for 5–10 min. Organoids were then centrifuged at 1000 rpm for 5 min. Washed the organoids with culture medium once, and centrifuged again. LCOs were grown for 2–3 weeks and then were passed at a ratio ranging from one to two to one to three.

Ultra-low adsorption culture

The suspension of lung cells and fibroblasts was obtained through the methods described in the “Organoid culture with fibroblast/MSCs in Mi-gel”. 1 million lung cancer cells were harvested for the LCCs group and 0.5 million lung cancer cells and 0.5 million fibroblasts were harvested and fully mixed with a 1 ml pipette. Washed the cells with PBS and followed a 1500 rpm centrifugation for 5 min. Removed the PBS and resuspended cells with 10 ml lung cancer organoid medium, then added the cell suspension into a 10 cm ultra-low adsorption cell culture dish. The dish was subsequently positioned within a temperature-controlled cell incubator, maintained at 37 °C, and enriched with a 5% CO2 environment. The lung cancer organoid medium was refreshed every 3 days. While refreshing the medium, cell suspension was carefully collected and centrifuged to remove the old medium, and added 10 ml fresh medium to the cells.

Hematoxylin and Eosin (HE) staining

To prepare for HE staining, both the organoids were set in 2% agarose, and the fresh tumor samples were fixed using 10% formalin for 24 ~ 48 h. Subsequently, the samples were placed into embedding cassettes and subjected to a dehydration process using an ethanol series: starting with 75% ethanol overnight, followed by 85% ethanol for 3 h, then 95% ethanol for another 3 h, and finally two 1-hour sessions in 100% ethanol. Post-dehydration, the samples in the cassettes were cleared in two 30-minute xylene baths. Afterward, the samples were encased in paraffin. The FFPE specimens, which had been fixed in formalin and embedded in paraffin, were subsequently sectioned into 4 μm-thick slices and subjected to drying at 60 °C for 2 h. After drying, the samples can be effectively cooled to 20 °C and were preserved at 4 °C. The sections were then dewaxed in xylene for three intervals of 10 min each and rehydrated in a descending alcohol series: twice in 100% ethanol, once in 95% ethanol, and once in 70% ethanol, each step lasting 10 min. For HE staining, the slides underwent treatment with hematoxylin, were differentiated, reblued, and counterstained with eosin, and finally sealed using a specific HE staining kit (G11201, Solarbio, Beijing, China).

Immunofluorescence analysis (IF)

For the IF analysis, sections of tissue or LCOs were initially subjected to dewaxing and rehydration. Subsequently, a citrate buffer with a pH of 6.0 was utilized for antigen retrieval. Following that, the sections underwent permeabilization at 20 °C for 30 min using a 0.5% Triton X-100 solution in PBS. Afterward, they underwent blocking by incubating them in a solution consisting of 5% BSA and 0.2% Triton X-100 in PBS for another 30 min at 20 °C. After completing the blocking step, the sections were exposed to appropriate primary antibodies. TTF-1 (A18128, ABclonal), BRG1 (A2117, ABclonal), Cytokeratin 18 (CK18) (A19778, ABclonal), P63 (12143-1-AP, Proteintech), Cytokeratin5/6 (CK5/6) (28506-1-AP, Proteintech), CD56 (A7913, Abclonal), Synaptophysin (syn, A18127, Abclonal), Ki-67 (27309-1-AP, Proteintech), Actin (A12379, Invitrogen), and Integrin β1 (ab24693, Abcam), were utilized for the IF analysis at 4℃ overnight. The slides underwent a PBS wash followed by exposure to diluted secondary antibodies.

DNA extraction and WES (whole-exome sequencing) analysis

Genomic DNA from the tissue samples was extracted using the Tissue Genomic DNA Extraction Kit (DP304-03, Tiangen). The DNA was then fragmented with a Covaris M220 Focused-ultrasonicator (Covaris, Massachusetts), after which sequencing libraries were prepared. The Human Exome 2.0 Plus capture kit (Twist Bioscience) was utilized for exome capture according to the instructions provided by the manufacturer. Sequencing of the final library preparations was carried out on the Illumina NovaSeq 6000 Sequencing System (Illumina), generating 150 bp paired-end reads, at LC-Bio Technology Co., Ltd.

The software Fastp [34] was employed to trim sequencing adapters and remove nucleotides with a quality score below 20. For read alignment, the Burrows-Wheeler Aligner (BWA) [35] was used to map the reads to the hg19 reference genome. Following alignment, Picard tools were applied to identify and flag duplicate reads within the BAM files. The next step in post-alignment processing involved local realignment around indels to amend any misalignments. Before proceeding to variant calling, a base quality score recalibration was conducted to mitigate systematic errors. Mutect2 [36] was then used for the joint calling of somatic single nucleotide variants (SNVs) and insertions/deletions (InDels). ANNOVAR was engaged to annotate the variants with biological information [37]. Copy number variations were detected using the CNV kit [38]. The GC content of the sequences was factored in to normalize the distribution of reads, and this normalized read distribution across sliding windows was employed to determine copy number differences between tumor and normal samples.

RNA extraction and RNA sequence analysis

The TRIzol reagent (ThermoFisher, 15596026, USA) was utilized for the extraction of total RNA, followed by purification in accordance with the manufacturer’s provided protocols. Divalent cations were then used to cleave the mRNA into smaller pieces. These RNA fragments were used as templates for the synthesis of complementary DNA (cDNA) through reverse transcription. Subsequently, U-labeled second-strand DNA was synthesized using the cDNA as a template. Before the indexed adapters were attached, an additional A-base was appended to the blunt ends of the DNA strands. The fragments were then ligated with dual-index adapters and size-selected with AMPureXP beads. Post-ligation, the products were amplified by PCR, following the application of a heat-labile UDG enzyme (NEB-cat.m0280). The cDNA libraries produced had an average insert size of approximately 300 ± 50 bp. The libraries underwent sequencing using the Illumina Novaseq™6000 system (LC-Bio Technology CO., Ltd.) with a PE150 configuration, generating paired-end reads of 2 × 150 bp, in line with the vendor’s provided procedures. The DESeq2 software was utilized to perform differential expression analysis of genes, comparing two distinct groups. For comparing two samples, edgeR was employed. When comparing two individual samples, edgeR was the tool of choice. Genes that demonstrated a false discovery rate (FDR) less than 0.05 and an absolute fold change exceeding 2 were categorized as differentially expressed. These identified genes then underwent enrichment analysis through Reactome pathways, Gene Ontology (GO) functions, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways.

MSCs isolation and culture

MSCs were extracted from the human placenta using the same method described in previous studies [39, 40]. These MSCs were subsequently cultivated in Petri dishes using a DMEM medium (Gibco-Invitrogen) containing 10% FBS, along with a combination of antibiotics.

Production of viral vectors and transduction

The MSCs were incubated in a 37°C environment with 5% CO2, using DMEM supplemented with 10% FBS and a dual antibiotic solution containing penicillin/streptomycin at a concentration of 50 U/ml. Lentiviral vectors were created using the pLKO.1-TRC (Addgene, #10878), psPAX2 (Addgene, #12260), and pMD2.G(Addgene, #12259) plasmids to enable the expression of short hairpin RNA (shRNA) for inhibiting kindlin-2 (Sh-K2) or a non-targeting scrambled shRNA (Sh-con). The specific sequences used were as follows: Sh-K2, 5’-GAGGACCTATATGAATGG-3’; Sh-con, 5’-ACGCATGCATGCTTGCTTT-3’. Lentiviruses carrying these Sh-K2 and Sh-con were produced by co-transfecting HEK293T cells. To generate DNA expression vectors, kindlin-2 cDNA encoding the corresponding protein sequences were cloned into the 3×FLAG tagged pLVX-IRES-Hyg (3f). To generate lentiviral expression vectors encoding kindlin-2, 3×FLAG-kindlin-2(3f-K2) were co-transfected with psPAX2 and pMD2.G into HEK293T cells.

For the transduction process, MSCs or fibroblasts were grown in their standard growth medium until they achieved a cell density of 70%. Afterward, a fresh medium was introduced, which included the lentivirus at a multiplicity of infection (MOI) level of 100, and incubated for 24 h. The viral transduction was conducted with the addition of 8 µg/mL polybrene to enhance infection efficiency.

Western blot analysis

The western blot procedure was performed by methods previously reported [39, 40]. The total protein extracts were isolated using a 1% SDS lysis buffer (P0013G, Beyotime). 10 to 60 µg protein samples per lane were subjected to 10% SDS-polyacrylamide gel electrophoresis and subsequently transferred onto nitrocellulose membranes. The protein-loaded membranes were blocked for sixty minutes at room temperature using a solution comprising 5% skimmed milk. Following this, the membranes underwent an overnight incubation at 4 °C with primary antibodies: GAPDH conjugated with HRP (Santa Cruz, sc-365062HRP) and kindlin-2 (Proteintech, 11453-1-AP). After extensive washing, the membranes underwent incubation with secondary anti-rabbit antibodies (Jackson ImmunoResearch, #711-005-152) conjugated to HRP. The protein bands were detected using the Bio-Rad ECL kit, and the immunoblot was visualized using the Syngene G：BOXChemiXX9 automated digital imaging system.

CCK-8 test

The cells were processed following the procedures outlined in the “Ultra-low adsorption culture” and “MSCs isolation and culture” sections. Washed the cells with PBS which was centrifuged at 1500 rpm for 5 minutes and removed the PBS, repeated this step twice. Resuspended cells with melted Mi-gel to reach a concentration of 2 × 106 cells per milliliter gel. Then Mi-gel buffer B was added, and the ratio of Mi-gel and Mi-gel buffer B was 20:1. Fully mixed the cell suspension with a pipette. Added 10ul of cell suspension to every well in a 96-well plate for cell culture. Each group had 5–6 repetitions. After a 7-day culture, refreshed the medium and added 100ul medium, then added 10ul Cell Counting Kit-8 (C0037, Beyotime) into the wells. The plate was incubated at 37 °C for 60 min, before measuring the absorbance at 450 nm using a microplate reader (BioTek, SYNERGYLX).

Statistical analysis

Statistical comparisons between two independent groups were conducted using two-tailed unpaired parametric Student’s t-tests or Mann-Whitney test. The Pearson correlation coefficient was employed for assessing correlations. All results were presented as mean values ± SD. In the Figures, asterisks denoted statistical significance as follows: *, **, ***, and **** represent p values smaller than 0.05, 0.01, 0.001, and 0.0001 respectively; while n.s. indicates a p value larger than 0.05. The creation of graphs and the statistical analyses were performed using the software GraphPad Prism 9.0 and the R programming language for statistical computing.