Gastric cancer is one of the deadliest cancers worldwide. An accurate prognosis is essential for effective clinical assessment and treatment. Spatial patterns in the tumor microenvironment (TME) are conceptually indicative of the staging and progression of gastric cancer patients. Using spatial patterns of the TME by integrating and transforming the multiplexed immunohistochemistry (mIHC) images as Cell-Graphs, we propose a graph neural network-based approach, termed Cell−GraphSignatureorCGSignature, powered by artificial intelligence, for the digital staging of TME and precise prediction of patient survival in gastric cancer. In this study, patient survival prediction is formulated as either a binary (short-term and long-term) or ternary (short-term, medium-term, and long-term) classification task. Extensive benchmarking experiments demonstrate that the CGSignature achieves outstanding model performance, with Area Under the Receiver Operating Characteristic curve of 0.960 ± 0.01, and 0.771 ± 0.024 to 0.904 ± 0.012 for the binary- and ternary-classification, respectively. Moreover, Kaplan–Meier survival analysis indicates that the “digital grade” cancer staging produced by CGSignature provides a remarkable capability in discriminating both binary and ternary classes with statistical significance (P value < 0.0001), significantly outperforming the AJCC 8th edition Tumor Node Metastasis staging system. Using Cell-Graphs extracted from mIHC images, CGSignature improves the assessment of the link between the TME spatial patterns and patient prognosis. Our study suggests the feasibility and benefits of such an artificial intelligence-powered digital staging system in diagnostic pathology and precision oncology.

The tumor microenvironment (TME) comprises distinct cell types, including stromal types such as fibroblast cells and macrophage cells, which have recently become a critical factor in tumor development and progression. Here, we identified the TME-related gene, plexin domain containing 2 (PLXDC2), in a high-stromal-score population. And we revealed that this gene was related to poor survival and advanced (tumor-node-metastasis) stage in gastric cancer (GC) patients from The Cancer Genome Atlas database. An integrated gene profile and functional analysis of the proportions of tumor-infiltrating immune cells revealed that the expression of the M2 macrophages cell marker CD163 was positively correlated with PLXDC2 expression. In addition, the M2 macrophages gene signature and high PLXDC2 expression were associated with the inflammatory signaling pathway and the epithelial-to-mesenchymal transition (EMT)-related gene signature. Single-cell study of GC identified PLXDC2 was enriched specifically in fibroblasts and monocytes/macrophages populations, which supported its important role in the stroma. Furthermore, according to a tissue microarray immunohistochemistry analysis, the expression of PLXDC2 elevated in human GC stromal specimens compared to tumor tissue specimens. Moreover, PLXDC2 overexpression in the stromal compartment was associated with CD163-positive regulatory M2 macrophages, and its functions were related to the pathogenesis of GC. Multiplexed immunohistochemistry verified PLXDC2’s correlation with EMT markers. Our data suggested that PLXDC2 was expressed in stromal cells and that its crosstalk with tumor-associated macrophages could contribute to cancer biology by inducing the EMT process.

Given the complexity and highly heterogeneous nature of the microenvironment and its effects on antitumor immunity and cancer immune evasion, the prognostic value of a single immune marker is limited. Here, we show how the integration of immune checkpoint molecule expression and tumor-associated immune cell distribution patterns can influence prognosis prediction in non-small-cell lung cancer (NSCLC) patients. We analyzed tissue microarray (TMA) data derived from multiplex immunohistochemistry results and measured the densities of tumor-infiltrating CD8+ and FOXP3+ immune cells and tumor cells (PanCK+), as well as the densities of programmed cell death 1 (PD-1)+ and programmed cell death ligand 1 (PD-L1)+ cells in the peritumor and intratumor subregions. We found a higher density of infiltrating CD8+ and FOXP3+ immune cells in the peritumoral compartment than in the intratumoral compartment. In addition, unsupervised hierarchical clustering analysis of these markers revealed that the combination of high CD8/FOXP3 expression, low PD-1 and PD-L1 immune checkpoint expression, and lack of epidermal growth factor receptor (EGFR) mutation could be a favorable predictive marker. On the other hand, based on the clustering analysis, low CD8/FOXP3 and immune checkpoint (PD-1 and PD-L1) expression might be a marker for patients who are likely to respond to strategies targeting regulatory T (Treg) cells. Furthermore, an immune risk score model was established based on multivariate Cox regression, and the risk score was determined to be an independent prognostic factor for NSCLC patients. These results indicate that the immune context is heterogeneous because of the complex interactions of different components and that using multiple factors in combination might be promising for predicting the prognosis of and stratifying NSCLC patients.