Classification of PANoptosis-related subtypes in HCCThe study explored the PANoptosis-associated genes and observed a notable rise in gene expression of major regulators in HCC tissues in contrast to normal individuals (Fig. 1A). Subsequently, we took advantage of consensus clustering to stratify HCC patients into distinguished molecular clusters based on the expression of PANoptosis molecules. Our research showed that the CDF curve had a smooth pattern, and the area beneath the curve was maximized when k = 2. This observation suggested that the categorization was optimal, as depicted in Fig. 1B. Principal component analysis (PCA) indicated that there was a clear distinction between cluster 1 (n = 200) and cluster 2 (n = 171) (Fig. 1C). The heatmap was utilized to sort patients into two groups on the basis of PANoptosis (Fig. 1D), and GSVA scores signified that group 2 possessed higher PANoptosis degree, namely high-PANoptosis (C2) groups (Fig. 1E). The patients belonging to subgroup 2 showcased inferior survival outcomes (Fig. 1F). The circular diagram depicted the distinctive divergences in tumor grades and stages across two subcategories, suggesting PANoptosis may be correlated with neoplasm evolution (Fig. 1G). The GO analysis proved that these controllers primarily resided in the NF-kappaB signaling pathway and signature connected with inflammation. The KEGG analysis discovered substantial correlations between the acquired evidence and numerous biological events, including TNF signaling, measles human immunodeficiency virus 1 infection, apoptosis, necroptosis and platinum resistance, uncovering that PANoptosis served an integral part in the controlling of several immune signals. (Figure S1). The consensus clustering approach highlighted apparent disparities between two unique PANoptosis groups, each defined by discrete transcriptional profile characteristics. The innovative categorization system, which was based on PANoptosis traits, possessed the capacity to improve medical treatments and facilitate decision-making at an early stage.Fig. 1 The usage of consensus clustering was conducted within the framework of PANoptosis-associated controllers. (A) Transcriptional levels of PANoptosis-related genes in normal subjects and HCC were portrayed by heatmap according to TCGA data. (B) The process involved recognizing two individual clusters by using the technique of consensus clustering analysis. (C) Scatter plot of PCA scores of two clusters. Each dot indicated a sample and circles represented the confidence interval of each group. Horizontal and vertical coordinates represented two main components. (D) A heatmap depicting the manifestation levels of PANoptosis genes in two subtypes. Red meant higher values while blue indicated lower scores. (E) GSVA method predicted the PANoptosis scores in two clusters. (F) The Kaplan-Meier curve was employed to gauge the overall survival. (F). Distribution differences of clinical features between 2 clusters. Patients’ proportion of each trait was pictured in matching colors and statistical analysis was fulfilled by the chi-square test.Enrichment analysis and medication susceptibility testing between PANoptosis subgroupsRegarding the expected result and the various degrees of PANoptosis in two genetic subtypes, we undertook a GSVA analysis of tumor hallmark signaling pathways, BP and MF to exploit variations in organisms. The results implied that the high-PANoptosis group maintained a substantial increase in immunological and tumorigenic pathways, whereas the low-PANoptosis group exhibited extensive participation in metabolism-related pathways and oxidative phosphorylation. (Fig. 2A). The fraction that focused on low levels of PANoptosis seems to be fully engaged in many metabolic processes, as pictured in Fig. 2B. While the high PANoptosis subset displayed a hold on the regulation of the cell cycle. Moreover, the findings gained from the MF analysis demonstrated that enhanced PANoptosis exhibited notable advantages in interacting with large biological molecules. Conversely, the lower group focused on governing various bioenzymatic activities, as seen in Fig. 2C. Patients with high PANoptosis features were active in immune and inflammatory signals such as TNF-α signaling, interferon-α and -γ signaling. Nevertheless, oncogenic traits were also prominent for them, includng epithelial-to-mesenchymal transition (EMT), PI3K-AKT-MTOR signaling, Wnt-β catenin signaling and so on. The amalgamated computational observations disclosed that people in differing PANoptosis groups had extensive and intricate bioheterogeneity, which might account for their varied outcomes.Fig. 2 This research project was dedicated to setting up enrichment analysis in two independent PANoptosis groupings. The color red signified elevated scores and activities, whereas the color blue denoted diminished values. (A) The graph conveyed the contrasting GSVA scores of 33 cancer signature pathways in two subgroups. (B) GSVA scores of BP (C) and MF activities were visualized by heatmap. To display the leading property, the top 15 enriched terms in two subgroups were selected.Taking account of the obvious distinctions in signaling pathways between the two groups of PANoptosis, we also tested whether these subtypes could guide the practical use of medicines in a clinical setting. The IC50 values of 138 chemotherapeutic drugs were determined, and 117 of them exhibited considerable variations, constituting two separate clusters (Fig. 3A). In order to pinpoint possible drugs for two categories, a list of the ten most unique medications was suggested. Patients in the low PANoptosis cohort had increased susceptibility to several pharmaceuticals, including WH.4.023, SB590885, AZD6244, RDEA119, Temsirolimus, A.770,041, AZD.0530, Erlotinib, Dasatinib and NVP.TAE684 (Fig. 3B); whereas AG.014699, XMD8.85, Doxorubicin, Tipifarnib, Epothilone.B, BMS.754,807, Mitomycin.C, BI.2536, Bleomycin and Gemcitabine may be more appropriate for high PANoptosis subgroup (Fig. 3C). The investigation’s findings reaffirmed that the two genotypes presented noteworthy deviations in their biological properties and clinical management, necessitating the implementation of targeted treatment techniques.Fig. 3 Estimation of drug susceptibility of each individual between 2 clusters based on GDSC training set. (A) 138 agents’ IC50 were delineated by heatmap. The red color suggested elevated IC50 values while blue implied decreased IC50 scores. (B) The efficacy of the top 10 medications was noted in persons with low PANoptosis and (C) the proactive PANoptosis group was reflected. An average multiple of IC50 values ranked pharmaceuticals. The person with lower IC50 scores noted more effective to a drug and vice versa.The landscape of tumor microenvironments in PANoptosis clustersIntending to complete a rigorous exploration of the peculiarities of TME, we hired multiple approaches to quantify the relative abundance of immune cells. Cluster 2 of the work displayed an immense rise in immunological cells, including CD8 + T cells, B cells, neutrophils, and CD4 + T cells, as portrayed in Fig. 4A. The TIP technique evaluated subjects in the high PANoptosis subtype and noticed that they had increased immunological activity, involving the release and transmission of cancer antigens. The attracting capability of CD8 + T cells and Th1 cells was also energetic (Fig. 4B). Additionally, we handled an analysis of the prevalence of various immunological checkpoints, immune-activating factors, HLA molecules, chemokines, and chemokine receptors. The majority of these components were found to be augmented in the high-PANoptosis category (Fig. 4C-G). The ESTIMATE methodology additionally computed that cluster 2 presented higher stromal scores, immune scores, and ESTIMATE scores (Fig. 4H), whereas purity scores were observed to decline (Fig. 4I). The preceding details supplied evidence that individuals with upgraded levels of PANoptosis ported a peculiar tumor immunotype outlined by heightened immune activation, which encompassed the potential to identify patients who were more inclined to have appropriate responses to immunotherapy interventions13.Fig. 4 The current study aimed at interpreting the spatial distribution patterns of immune cells and immunological regulators within two unique molecular categories of PANoptosis. (A) A complex plot calculated the immune cell scores in two subtypes based on TIMER, MCPcounter, QUANTISEQ and EPIC approaches. Different colors showcased diverse immune infiltration degrees in four calculation methodologies. (B) The TIP algorithm predicted the immunophenotypes between 2 subsets. (C) Transcriptional levels of immune checkpoints, (D) immunostimulatory factors, (E) HLA, (F) chemokine and (G) chemokine receptors in two clusters. (H) ESTIMATE algorithm assessed the stromal scores, immune scores and aggregated ESTIMATE scores. (I) The purity scores within two distinct clusters. Higher purity scores indicated elevated percentages of tumor cells.Development and validation of PANoptosis-related prognostic signatureIn order to generate a solid forecasting model, we chose the TCGA dataset as the training set and utilized the ICGC file as the testing set. The C-index achieved its maximum value when examined with 101 resultants and we observed that RSF fulfilled the greatest average C-index in two different profiles (Fig. 5A). Following the deployment of RSF, a risk score was generated for each individual. Patients were divided into low- and high-risk groups according to the median score. The PANoptosis score and the survival state of HCC were presented in Fig. 5B, suggesting alive individuals were clustered in low scores both in TCGA and ICGC profiles. The KM curve indicated patients with high risk had prominent shorter OS than low-risk patients in both cohorts (Fig. 5C). ROC curves demonstrated the area under curve (AUC) values of 2 years, 3 years, and 4 years were 0.974, 0.981 and 0.987 in the TCGA cohort, respectively. In the ICGC cohort, the AUC values were 0.675, 0.685 and 0.636, respectively (Fig. 5D). The heatmap indicated that T and the clinical stage had remarkable diversity in the TCGA cohort, while tumor grade presented obvious distinctions in the ICGC set (Fig. 5E). Additionally, multivariate Cox regression analysis proved risk score was an independent prognostic factor (Table S2). The data provided above emphasized the significant projected values of the PANoptosis score, which may function as an alternative predictor for prognosis in patients with HCC.Fig. 5 A survey was designed to assess the predictive relevance of the PANoptosis score in HCC. (A) The C-index of each prognostic model was structured by 10 machine learning algorithms in TCGA and ICGC cohorts. (B) This study effort examined the allocation of risk scores and prognostic statuses among HCC patients in the training and testing datasets. The median scores split the individuals into two risk groups. (C) KM graphs highlighted the OS between low and high-risk individuals. (D) ROC analysis of risk score on OS at 2-year, 3-year and 4-year. (E) Dispersion of clinical profile indicators and clinical outcomes according to TCGA and ICGC profiles. The chi-square test was applied for statistical analysis.Certification of predicted efficacyIn glowing of the evident correlations between risk score and prognosis, we incorporated clinical characteristics to formulate a nomogram for the intent of prognostic forecasting in patients. These clinical characteristics encompassed gender, T stage, TNM stage, age and grade. The nomogram was put together for the objective of determining and quantifying the likelihood of survival at 2,3 and 4 years, as illustrated in Fig. 6A. The calibration curve indicated a satisfactory agreement with the ideal forecast (Fig. 6B). And ROC curves verified that our nomogram was superior to other parameters in forecasting patients’ prognoses, with an AUC of 0.916 (Fig. 6C). To gauge the feasibility of the scoring system, we integrated five independent datasets. The GSE109211 cohort revealed that HCC patients who responded to Sorafenib medication had lower scores compared to non-responders, with a predicted value of 0.917 (Fig. 6D). Patients who benefited from the TACE surgery had higher scores compared to those who failed to respond, with a predicted level of 0.618 (Fig. 6E). Differentiating HCA from well-differentiated HCC might be challenging in some instances14. Throughout our examination, individuals suffering from HCC encountered lower scores, which provided a meaningful predictive value in both groups (Fig. 6F). Concurrently, the accuracy of the index to determine the likelihood of vascular invasion was similarly encouraging, with an AUC of 0.684 (Fig. 6G). The above data indicated that our risk index may be a viable indication for rating risks and aiding clinical decision-making in HCC-related management and diagnosis.Fig. 6 Fabrication of a nomogram for prognosticating the OS of HCC patients. (A) The nomogram was constructed by merging a risk score with clinical criteria. (B) The calibration curve tested the envisioned performance of the nomogram at 2-year, 3-year and 4-year. (C) ROC curves were utilized to determine the clinical applicability of each indicator and the nomogram. (D) The predictive proficiency of the model in the group receiving Sorafenib therapy, (E) TACE curing profile, (F) HCC and HCA profiles and (G) HCC patients associated with vascular invasion dataset.Serological diagnostic value of HSP90AA1 for HCC patientsGiven the extraordinary anomalous expression and significant predictive meaning, we dictated HSP90AA1 as a latent circulating indicator for HCC diagnosis. The experimental findings unambiguously showed that the concentrations of serum HSP90α were substantially elevated in patients with HCC in comparison to HCs (Fig. 7A). The ROC plots for differentiating all HCC patients and early confirmed individuals were shown in Fig. 7B-C. The AUC values for separate groups were 0.9 and 0.842, with optimal cutoff values of 75.67 and 91.07 ng/ml, respectively. The diagnostic efficiency characteristics, including specificity, sensitivity, accuracy, PPV and NPV, were presented in Table S3. Impaired liver function predictors, such as ALT, AST, GGT and ALP were observed to be dominantly interrelated to HSP90AA1 (p < 0.05, Fig. 7D). Regretfully, little correlation was discovered between serum HSP90AA1 levels and clinical features, including age, gender, T stage, N stage, M stage and clinical stage (Figure S2).Fig. 7 In vitro diagnostic significance of HSP90AA1 for HCC patients. (A) The protein levels of HSP90α were juxtaposed between individuals with HCC and those without any pathological conditions. (B) ROC depicted the diagnosability in distinguishing HCs from HCC and (C) early malignant patients. (D) The association between HSP90α and liver function biomarkers, including ALT, AST, GGT and ALP, were plotted by scatter diagrams.Depiction of PANoptosis in scRNA-seq degreeThe stated above research highlighted the link between PANoptosis and the immune ecology at the RNA-seq level. In an attempt to gain a grasp of the heterogeneity in PANoptosis activity among various immune cells, we accomplished a scRNA-seq test based on 7 HCC tissues. The samples underwent a decrease in dimension and cell annotation (Fig. 8A-B). Among them, we identified 23 cell subsets and 6 cell types, including B cell, dendritic cell, endothelial cell, exhausted CD8 + T cell, macrophage and malignant cells. The cells were harvested for CytoTRACE analysis in order to ascertain their stem divergence. The outcomes suggested that malignant cells had greater stem scores and less differentiation capability as opposed to immune and stromal cells (Fig. 8C). Correspondingly, the Addmodulescore gadget illustrated the level of PANoptosis activity, and it appeared that immune and stromal cells had a more significant PANoptosis landscape compared with cancer cells (Fig. 8D), and PANoptosis was dramatically negative with stem value (Fig. 8E). Exhausted CD8 + T cells experienced the greatest level of PANoptosis intensity, and malignant cells showed the least intensity, which was consistent with our bulk RNA-seq findings that those with high PANoptotsis were in the presence of more abundant immune cells (Fig. 8F). Our findings have proven that PANoptosis varies considerably across immunological, stromal, and tumor cells, sending evidence of its pivotal participation in the evolution of tumors.Fig. 8 The scRNA-seq method disclosed the PANoptosis inherent in the immunological conditions. (A) A t-SNE plot was used to allocate colors to 7 HCC tissues, signifying 23 distinct cell clusters after quality control and batch removal. (B) 6 subgroups were categorized using the ssGSEA technique based on the existing cell markers. (C) The Cytotrace approach approximated the stem values of each cell. Higher scores noted less degree of cellular differentiation and increased aggressiveness. (D) The Addmodulescore strategy reckoned the total PANoptosis activity in all cells. Red suggested raised PANoptosis degrees. (E) The scatter plot pictured the correlation between PANoptosis values and the stemness of all cells. (F) The PANoptosis metrics in each cell type were portrayed by histogram.Afterward, we classified and defined groups with high and low levels of PANoptosis by implementing the median values within each grouping (Fig. 9A). To explore the biological discrepancy between two PANoptosis cell groups, scrutiny was conducted. Figure 9B displayed that the high PANoptosis cell subset was predominantly enriched in processes such as the conversion of ingested MeSeO2H into MeSeH, intracellular oxygen transport, interactions between VEGF ligands and receptors, proton-coupled neutral amino acid transporters, muscarinic acetylcholine receptors, dermatan sulfate biosynthesis and other modulations of biogenic functions. The low-value group endured higher scores in alanine metabolism, agmatine production, mitochondrial ABC transporters, and carnitine synthesis, which was similar to bulk-RNA results. By employing the pySCENIC approach, we perceived the primary transcription element in two distinct cell subgroups (Fig. 9C). Among them, ATF3, ETS2, FOS, FOSB, HES1, IRF1, JUN, JUNB, JUND, KLF4, MEF2C, NFE2L2, NFIA, SOX4 and STAT1 were highly expressed in a high score group, while CEBPD and XBP1 mostly emerged in low score cells. Ligand receptor pairs mediating cell communication in two PANoptosis subgroups confirmed APP-CD74 and MDK-SDC2 were the main interactions in the high and low-score groups, respectively. MDK-SDC2 and MDK-NCL were the leading interaction pairs between them. In total, cells in discrete subgroups of PANoptosis exhibited distinguishable biological variability. The deregulation of PANoptosis may partially drive the formation of tumors.Fig. 9 Researching the properties of molecules and the routes by which cells in low- and high-PANoptosis subgroups, leveraging the scRNA-seq strategy. (A) The cells were split into two subpopulations based on the median values. (B) Reactome saturation analysis was carried out in two distinct cell clusters. The prime biosignatures were displayed by heatmap. (C) pySCENIC algorithm presumed the transcriptional regulators and (D) each distribution density in two cell subgroups. Higher density implied more active transcriptional regulation cells. (E) The Cellchat instrument sorted the ligand-receptor pairing probabilities under two distinct categories and red meant increased chances.Identity of malignant cells in dissimilar PANoptosis subgroupsInitial investigations disclosed that cancerous cells got the lowest extent of PANoptosis. To peculiarly delve into the disparity between different PANoptosis malignant cells, we segregated them into 9 independent subcategories and discovered that cluster 2 had the most deficient scores, whilst group 3 conveyed the highest score level (Fig. 10A-B). Cytotrace disclosed that HCC cells with lower PANoptosis activity contained lower stemness and more mature differentiation levels (Fig. 10C). The Addmodulescore program computed the metastasis and proliferation scores, confirming that elevated PANoptosis tumor cells had exacerbated carcinogenicity (Fig. 10D-E). As illustrated in Fig. 10F, the preponderant transcriptional factors varied in each cluster. As for the low PANoptosis subset, NFIB, STAT2 and NFE2L1 hold the main position; while SREBF2, JUND, JATAD1, NFIA, ZBTB20 and SMAD5 played a dominant role in elevated PANoptosis cells. Yet, the major scale of interaction between each set of cells did not exhibit any noticeable divergences (Fig. 10G). Our data concluded that PANoptosis has the ability to impact the development of tumor cells to some degree, thereby affecting the outcomes for beings with HCC. Addressing high-PANoptosis tumor cells could represent a promising method for combating HCC.Fig. 10 Classification of cancerous cells based on PANoptosis behaviors. (A) The tSNE map organized and consolidated the tumor cells into 9 distinct groups. (B) The phenomenon of PANoptosis was shown using box line plots. (C) Cytotrace tool evaluated the stemness values in the lowest and highest PANoptosis cell subsets. (D) Addmodulescore estimated the metastasis and (E) proliferation scores in two diverse PANoptosis cell subgroups. (F) The pySCENIC program recognized the transcription elements in 9 specific cell subsets. Red indicated higher levels of corresponding transcription factors. (G) Cellchat accurately forecasted the cellular connections between the C2 and C3 cell clusters. The wider area demonstrated more vigorous interactions.
