The public data were mainly collected from Gene Expression Omnibus (GEO) and the Cancer Genome Atlas (TCGA) databases. Extensive expression profiles and associated clinical data were obtained from the TCGA-GDC program and subsequently incorporated into the TCGA database (TCGA-LUSC and TCGA-LUAD projects). The “STAR-Counts” format was used for each patient’s expression file, whereas the “bcr xml” format was used for clinical information. The perl and R codes of the authors were utilized for data collation. The probe annotation used the GRCh38.p13.gtf, which can be downloaded from the Ensembl website and corresponds to the human genomic reference file. The genes with a median value <0.1 were removed. Meanwhile, the gene expression value was converted into log2 to reduce the impact of extreme value. For the GEO database, the GSE30219 was selected, whose annotation platform is GPL570 [HG-U133_Plus_2] Affymetrix Human Genome U133 Plus 2.0 Array. Patients’ expression profiles and clinical information in GSE30219 were downloaded from the link “Series Matrix File(s).” Before analysis, the expression profile data of GSE30219 were pro-precessed using the limma package, including missing value completion, correction, normalization, and so on. Human Protein Altas (HPA) project can provide representative immunohistochemical data and corresponding protein level evaluation of encoding genes. The immunohistochemical image of FADD and its subcellular localization information was obtained from the HPA website through online searching. The expression data of FADD in pan-cancer was downloaded from the website USCS-Xena (https://xenabrowser.net/datapages/). The limma package was utilized to identify differentially expressed genes (DEGs) (Ritchie et al., 2015). For DEGs analysis between patients with high and low FADD expression, the threshold was set as |logFC| > 0.6 and p-value <0.05.

The genes involved in the PANoptosis process were collected from previous studies, including ADAR, MEFV, AIM2, NLRP1, NLRC4, NLRP9, NLRP3, ZBP1, TNFRSF1A, PYCARD, FADD, CASP10, CASP1, CASP2, CASP12, CASP4, CASP3, CASP6, CASP5, CASP7, DFNA5, CASP8, MLKL, GSDMD, RIPK3, RIPK1, NAIP, TNF, NLRP6, GSDMA, GSDMC, GSDMB, APAF1, BAK, BAX, DIABLO, DCN, CASP9, and FAS (Wang and Kanneganti, 2021).

The STRING database (https://cn.string-db.org/) was used to construct the PPI network. Herein, the “Organism” used was “Homo,” and the “minimum required interaction score” was “medium confidence.” Cytoscape was used for visualization of the PPI network. Cytoscape software was used for visualization and clueGO plug-in was used for the biological enrichment analysis of nodes (Shannon et al., 2003).

The prognosis difference between the two groups was compared using the KM (Kaplan-Meier) survival curves and p < 0.05 was regarded as statistically significant. We used univariate and multivariate logistic regression to identify variables that can independently affect patients’ survival (p < 0.05 was regarded as statistically significant).

Using univariate cox-regression, prognosis-related genes were determined for use as the input genes. Subsequently, the LASSO regression analysis was applied to reduce data dimension and identify optimization variables. Finally, the multivariate Cox regression was utilized for signature identification of prognosis with the following equation “Risk score = (Gene_A × A) + (Gene_B × B) + (Gene_C × C) + (…) + (Gene_N × N), where A-N are the coefficients” (Wu et al., 2021). KM and receiver operating characteristic (ROC) curves were utilized to evaluate the prognosis prediction efficacy. For TCGA cohort, the patients were randomly divided into training and validation cohort in proportion of 1:1. The GSE30219 was selecte as the external validation corhort.

The rms R package was utilized to establish a nomogram plot combining clinical parameters and the risk score. To evaluate the fit between the actual survival and nomogram-predicted survival, the calibration and decision curve analysis (DCA) were generated.

Gene Ontology (GO) analysis was conducted using the clusterprofiler software (Yu et al., 2012). The biological variations between the two groups were determined using Gene Set Enrichment Analysis (GSEA) on the basis of the set pathway reference set (Subramanian et al., 2005). The terms with p < 0.05 were regarded as statistically significant.

On the basis of gene expression profile, the tumor immune microenvironment of patients with lung cancer was quantified using multiple algorithms, including CIBERSORT, EPIC, MCPCOUNTER, QUANTISEQ, TIMER, and XCELL algorithms (Chen et al., 2018; Plattner et al., 2020; Racle and Gfeller, 2020).

Patient immunotherapy responses were assessed utilizing the TIDE (Tumor Immune Dysfunction and Exclusion) program (Fu et al., 2020). Each patient was given a TIDE score, with scores >0 indicating poor immunotherapy response (or responders) and a score <0 indicating no success (non-responders). Drug sensitivity testing was performed using the Genomics of Drug Sensitivity in Cancer (GDSC) database (Yang et al., 2013).

The BEAS-2B, A549, H1299, and H441 cells were routinely stored in the laboratory. The cell line authentication was carried out using STR detection. They were cultured in the 10% heat-inactivated fetal bovine serum at 37°C with 5% CO₂. Depending on the confluence of the cells, the cells were passaged every 3-4 days.

Utilizing an RNA extraction kit (TaKaRa Bio), total RNA was isolated. Then, cDNA was synthesized by reverse-transcribing utilizing a high-capacity cDNA Reverse Transcription Kit. Using TB Green Premix Ex Taq, we performed RT-PCR to measure the levels of mRNA expression levels. The PCR primers were synthesized by Tsingke (Beijing, China). The sequence of primers was as follows: FADD, forward, 5′-GTG​GCT​GAC​CTG​GTA​CAA​GAG-3′, reverse, 5′-GGT​AGA​TGC​GTC​TGA​GTT​CCA​T-3′; GAPDH, forward, 5′-TTG​TCT​CCT​GCG​ACT​TCA​ACA​G-3′, revers, 5′-GGT​CTG​GGA​TGG​AAA​TTG​TGA​G-3′.

GenePharma Co., Ltd. generated both FADD knockdown plasmids and control plasmids. H1299 (5 × 10⁵) cells on a 6-well plate were plated and cultured at 37°C overnight using RPMI-1640. These cells were incubated in serum-free Gibco™-OptiMEM (from Thermo Fisher Scientific, Inc.) for 6 h at 37°C after being treated with shRNA and Lipofectamine^®2000 Reagent (from Invitrogen, Thermo Fisher Scientific, Inc.). The cells were then added with Fresh RPMI-1640 containing 10% FBS. For the effect of FADD on the proliferation of H1299, the concentrations of FADD siRNA were 0, 10, 25, 50, 100, and 200 nM, and the concentrations of the total shRNA were made up to 200 nM with NC shRNA for those transfections with a FADD shRNA concentration <200 nM. After 48 h of shRNA transfection, H1299 cells were treated with Adriamycin to observe the effect of FADD on the cells’ drug resistance, where the concentration of FADD shRNA used was 100 nM.

The assessment of cell proliferation ability was conducted using the CCK8, colony formation, and 5-ethynyl-2′-deoxyuridine (EdU) assays. To perform the CCK8 assay, cell culture media was introduced with CCK-8 reagent at a mixed ratio of 1:10 (Beyotime, Shanghai, China). Cell proliferation was assessed by measuring absorbance at wavelength 450 nm immediately after incubation at 37°C for 2 h. Colony formation assays involve cultivating lentivirus-infected cells for 14 days after seeding 2000 of them into six-well plates. After 10 min of treatment in 4% paraformaldehyde, the cells were stained for 30 min with crystal violet (from Beyotime, Shanghai, China). For the EdU assay, cells were inoculated on coverslips of 12-well plates at suitable cell numbers. The following day, EdU working solution was co-cultured with the cells for 2 h. Subsequent fixation and staining were performed according to the EdU (from Beyotime, Shanghai-China) instructions and photographed by fluorescent microscope.

According to the manufacturer’s instructions, the cells were stained with annexin V/fluorescein isothiocyanate (FITC) kit (BD Biosciences, Franklin Lakes, NJ, United States). In short, H299 cells were cultured in 6-well plates. We collected cells and added 100 μL combined buffer and 5 μL fluorescein isothiocyanate (FITC) labeled annexin V (20 μ G/mL) and incubate at room temperature in the dark for 15 min. Then, the mixture were added 5 μL propidium iodide (PI, 50 μ G/mL) and incubated in the dark for 5 min. After that, we added 400 μL combined with buffer solution and immediately performed FACscan for quantitative detection by flow cytometry.

TUNEL apoptosis detection kit (Suzhou Yuheng Biotechnology Co., Ltd.) was used to detect cell apoptosis. Cells (1 × 10⁵) were fixed with 4% polyformaldehyde, and 0.2% Triton X-100 was used for penetration. We then added 50 μL of TUNEL reaction mixture and incubated it in dark at 37°C for 60 min, and then used DAPI solution to re-stain the nucleus. The cells were observed under fluorescence microscope (Zeiss) and the apoptosis rate was calculated by ImageJ software.

Cells were lysed in pre-cooled RIPA lysis buffer (Beyotime) with freshly added protease inhibitor mixture (Hoffman-La Roche Ltd., Basel, Switzerland). The protein concentration was determined by the Dioctanoic acid determination kit (Thermo Fisher Scientific). The same amount of protein was separated by SDS-PAGE and transferred to PVDF membrane. The membrane and the first antibody (Bcl-2, Bax, pro-caspase-1, NLRP3, GAPDH) were incubated at 4°C overnight. After incubation with HRP-labeled secondary antibody at room temperature for 2 h, the protein bands were analyzed using ECL Hypersensitive Chemiluminescence Kit (Beyotime).

All the analysis was performed using the R, SPSS, GraphPad Prism eight and ImageJ software. Statistically significant was reached when p < 0.05. The t-test was utilized for assessing normally distributed data. The Mann–Whitney-U test was utilized for assessing non-normally distributed data.

The flow chart of whole study was shown in Supplementary Figure S1. According to the previous studies, we collected the molecules involved in the PANoptosis, which is shown in Figure 1A. These genes are involved in PANoptosis, Pyroptosis, Necroptosis, and Apoptosis. The PPI network of these PANoptosis-related genes is depicted in Figure 1B. GO analysis indicated that these genes were mainly involved in the execution phase of apoptosis (GO:0097199), cysteine-type endopeptidase activity (GO:0004197), endopeptidase activity (GO:0004175), cysteine-type peptidase activity (GO:0008234), peptidase activator activity (GO:0016504), positive regulation of cysteine-type endopeptidase activity (GO:2001056), positive regulation of endopeptidase activity (GO:0010950), positive regulation of peptidase activity (GO:0010952), activation of cysteine-type endopeptidase activity involved in the apoptotic process (GO:0006919), regulation of cysteine-type endopeptidase activity (GO:2000116), positive regulation of cysteine-type endopeptidase activity involved in the apoptotic process (GO:0043280), inflammasome complex (GO:0061702), membrane raft (GO:0045121), membrane microdomain (GO:0098857), membrane region (GO:0098589), cysteine-type endopeptidase activity involved in the apoptotic process (GO:0097153) (Figure 1C). Univariate Cox regression analysis was applied to identify the genes significantly correlated with the patient’s survival (Supplementary Material S1). The results indicated that the genes FADD, TNFRSF1A, CASP9, MLKL, and CASP4 were the risk factor for lung cancer patients (Figure 1D, p < 0.05). Among these genes, FADD had the smallest p-value and aroused our interest. We noticed that FADD was also highly expressed in lung cancer tissue compared to paracancerous tissue (Figure 1E). Meanwhile, it is also an adaptor for PANoptosis and Apoptosis processes. Therefore, the FADD was selected for further analysis.

Subsequently, using the HPA database, we analyzed the subcellular location of FADD. The result indicated that FADD was primarily localized in the nucleoplasm and cytosol based on the A-431 and U-251 MG cell lines (Figure 2A). Representative immunohistochemistry images obtained from the HPA database indicated a higher protein level of FADD in the lung cancer tissue (Figure 2B). KM survival curves demonstrated that patients having higher-level FADD expression tend to have a poor prognosis performance in terms of disease-free survival (DSS), overall survival (OS), and progression-free survival (PFI) (Figures 2C–E). The prognosis prediction efficiency of FADD in OS was shown in Supplementary Figure S2. Next, we assessed the correlation between FADD and the clinical stage. The result depicted that the FADD could be related to the worse N classification but not the T and M classifications (Figures 2F–I). Statistical univariate and multivariate regression analyses indicated that FADD is an independent factor of risk in the prognosis of a lung cancer patient (Figures 2J, K, Univariate, p = 0.002, Multivariate, p = 0.02). Supplementary Figure S3 depicts the FADD expression pattern of the pan-cancer.

Then, we evaluated the biological role of FADD in lung cancer. Firstly, we conducted the DEGs analysis in high- and low-FADD expression patients with criteria of p < 0.05 and log FC > 0.6 (Figure 3A). ClueGO assessment indicated that these DEGs were primarily enriched in musculoskeletal movement, negative regulation in systemic arterial BP, morphogenesis regulation of a branching structure, response to zinc ion, regulation of organic acid transport, reflex, and camera-type eye morphogenesis (Figure 3B). GSEA analysis based on the Hallmark gene set indicated that the pathways of glycolysis, mitotic spindle, inflammatory response, mTORC1 signaling, interferon-gamma response, apical junction, TNFA signaling through NFKB, G2M check-point, epithelial-mesenchymal transition (EMT), and E2F target were activated in patients with high FADD expression (Figure 3C). Also, for the GSEA analysis based on GO and REACTOME gene sets, we noticed that the terms of nail development, hemidesmosome assembly, keratin filament, keratinization, formation of the cornified envelope, mitotic spindle check-point were enriched in (Figures 3D–I).

After that, we looked at how FADD affected the immune microenvironment of lung cancer. The result indicated that FADD could remarkably affect specific immune cells, including naïve B cells, plasma B cells, memory B cells, follicular helper T cells, activated NK cells, Tregs, activated mast cells, macrophages, and so on (Figure 4A). For immune function, we found that the activity of APC_co_stimulation, APC_co_inhibition, check-point, CCR, parainflammation, MHC_class_I, and type_I_IFN_response were upregulated in patients with high FADD expression (Figure 4B).

Then, we construct a nomogram combining the clinical features and FADD for better prognosis prediction efficiency (Figure 5A). A good agreement was found between observed and anticipated survival times using calibration curves based on the nomogram (Figure 5B). DCA curves showed that combining clinical features could improve the prediction ability of FADD on patient survival (Figure 5C). Immunotherapy is a vital therapy choice for lung cancer. A positive correlation was noticed between the FADD and TIDE score, indicating that patients with high FADD expression might respond worse to immunotherapy (Figure 5D). Meanwhile, we found that the non-responders could have a higher FADD expression (Figure 5E). Moreover, we found that those patients who expressed FADD at a higher level may have a stronger immune exclusion (Figure 5F). High-FADD patients may respond better to AICAR, bortezomib, docetaxel, and gemcitabine, as per the drug sensitivity analysis (Figure 5G).

We next evaluated the biological role of FADD through experiments. The results from qRT-PCR depicted that the FADD was overexpressed in the lung cancer cells (Figure 6A). Following qRT-PCR verification of knockdown efficiency, sh#1 was chosen for further experiments (Figure 6B). Inhibiting FADD reduced lung cancer cell proliferation, as assessed by CCK8 and colony formation assays (Figures 6C, D). EdU assay also demonstrated a lesser percentage of EdU-positive cells in the cells with FADD knockdown (Figure 6E). We next identified the molecules regulated by FADD with |Cor| > 0.3 and p < 0.05, defined as FADD-related regulatory molecules (Figure 6F). Biological enrichment assessment depicted that FADD-regulated genes were primarily enriched in the replication of DNA that is DNA-dependent (GO:0006261), regulation of sister chromatid segregation (GO:0033045), DNA replication (GO:0006260), sister chromatid segregation (GO:0000819), condensed chromosome (GO:0000793), heterochromatin (GO:0000792), centromeric region (GO:0000775), chromosome, chromosomal region (GO:0098687), Prion disease (hsa05020), Cell cycle (hsa04110), apoptosis (hsa04210) and Alzheimer disease (hsa05010) (Figure 6G).

The result of flow cytometry showed a higher apoptosis ratio of H1299 cells in sh-FADD cells compared to the control cells, indicating that the knockdown of FADD could promote apoptosis process of lung cancer cells (Figure 7A). The result of TUNEL staining also showed that the same result (Figure 7B). Subsequently, we further used Western blot assay to detect the expression level of apoptosis-related markers (Bcl-2, Bax) and pyroptosis markers (Caspase-1, NLRP3) in H1299 cells. Result indicated that the expression level of Bcl-2 decreased significantly, while the expression level of Bax, Caspase-1 and NLRP3 increased significantly in FADD knockdown cells, indicating that the inhibition of FADD could promote the apoptosis and pyroptosis (Figure 7C).

Based on the molecule regulated by FADD, we attempted to identify a prognosis signature effectively depicting the patient’s prognosis. Univariate Cox regression was first utilized to determine the prognosis-related genes (Figure 8A, p < 0.1). Then, LASSO regression was used for data dimension reduction and optimized variable screening (Figures 8B, C). Ultimately, multivariate Cox regression unraveled a prognosis signature consisting of five genes, C11orf24, TMEM250, ANO1, LMNB2, and PLK1 (Figure 8D). The risk score was evaluated with the equation below: Risk score = C11orf24 * 0.256 + TMEM250 * 0.229 + ANO1 * 0.102 + LMNB2 * −0.317 + PLK1 * 0.294 (Figure 8D). The result indicated that our signature performed well in the training cohort. The KM survival curve showed that the high-risk patients might have a worse OS (Figure 9A). The AUC values over 1, 3, and 5 years were 0.758, 0.751, and 0.76, respectively (Figure 9A). The same result was also observed in the internal validation as well as external validation cohorts (Figures 9B, C, internal validation cohort: 1, 3, and 5 years were 0.747, 0.771, and 0.74, respectively; external validation cohort: 1, 3, and 5 years were 0.752, 0.76, and 0.68, respectively).