CITMIC: Comprehensive Estimation of Cell Infiltration in Tumor Microenvironment based on Individualized Intercellular Crosstalk

Abstract

The tumor microenvironment (TME) cells interact with each other and play a pivotal role in tumor progression and treatment response. A comprehensive characterization of cell and intercellular crosstalk in the TME is essential for understanding tumor biology and developing effective therapies. However, current cell infiltration analysis methods only partially describe the TME's cellular landscape and overlook cell-cell crosstalk. Here, this approach, CITMIC, can infer the cell infiltration of TME by simultaneously measuring 86 different cell types, constructing an individualized cell-cell crosstalk network based on functional similarities between cells, and using only gene transcription data. This is a novel approach to estimating the relative cell infiltration levels, which are shown to be superior to the current methods. The TME cell-based features generated by analyzing melanoma data are effective in predicting prognosis and treatment response. Interestingly, these features are found to be particularly effective in assessing the prognosis of high-stage patients, and this method is applied to multiple high-stage adenocarcinomas, where more significant prognostic performance is also observed. In conclusion, CITMIC offers a more comprehensive description of TME cell composition by considering cell-cell crosstalk, providing an important reference for the discovery of predictive biomarkers and the development of new therapeutic strategies.

Keywords: cell infiltration; cell‐cell crosstalk; individualized analysis; network analysis; tumor microenvironment.

Figure 1

Schematic overview of the CITMIC method.

Figure 2

Performance evaluation of CITMIC. A) The correlation coefficient for the comparison of silicon dissection methods and tissue and flow cytometry counts in the whole blood dataset (SDY144); B) The correlation coefficient for the comparison of silicon dissection methods and tissue and flow cytometry counts in the PBMC dataset (SDY67); C–H) Correlation of CyTOF fractions and cell type of Th17 cells, Activated CD4⁺ T cells, B cells, CD4⁺ cells, CD8⁺ cells, and Naive B cells in whole blood dataset (SDY144); I,J) Correlation of CyTOF scores and cell type plots of B cells, Monocytes in PBMC dataset (SDY67). CyTOF, Cytometry by Time‐Of‐Flight.

Figure 3

Validate the effectiveness of CITMIC in identifying single‐cell types. Heatmap of the InScore profiles calculated from a single‐cell data set (GSE86363) containing 1159 cells. The cells are classified into three categories: myeloid, lymphoid, and stromal cells.

Figure 4

Prognostic ability of the cellular InScore on melanoma patients. A) The cell risk factors are based on single‐factor Cox regression analysis. In this bubble diagram, 71 cells were divided into five groups according to their different cell functions: lymphoid, myeloid, stem cell, stromal cell, and other cells. The color of the bubble indicates whether the cell is a risk or protective factor, with red indicating a risk factor and green indicating a protective factor. The size of the point represents the size of the significance p‐value. If the p‐value is larger than 0.05, the symbol “X” is employed; B) Comparison of AUROC values over time for 1–10 year overall survival between CITMIC and ssGSEA risk models at TCGA melanoma; C) Comparison of AUROC values over time for 1–9 year overall survival between CITMIC and ssGSEA risk models at GSE19234 melanoma; D) Comparison of AUROC values over time for 1–4 year overall survival between CITMIC and ssGSEA risk models at GSE22155 melanoma.

Figure 5

Performance of CT‐TME Risk Model in High‐Stage Melanoma. A) Kaplan–Meier survival curves of patients classified into high‐risk and low‐risk groups using the CT‐TME risk model in high‐stage melanoma; B) Scatter plot depicting risk score and survival time; C) Gene set enrichment analysis of KEGG pathway of differential genes between high‐risk and low‐risk groups; D) Heatmap of cell‐cell correlation coefficients for high‐stage TCGA melanoma patients based on cell InScores.

Figure 6

Comparison of the prognostic performance of the CT‐TME risk model in adenocarcinoma between all patients and high‐stage patients, and molecular and clinical features associated with CT‐TME Score in breast cancer. A–D) Comparison of AUROC values over time for 1–5 year overall survival between CITMIC and ssGSEA risk models at high‐stage adenocarcinoma (TCGA‐BRCA, TCGA‐LUAD, TCGA‐READ, TCGA‐STAD). “Highly stage” refers to the ssGSEA risk model for highly staged adenocarcinoma; E) An overview presents an analysis of the association between known molecular and biological processes and CT‐TME Score in high‐stage BRCA. Columns represent patients sorted by CT‐TME Score from low to high (top row). Rows represent molecular and biological processes associated with the CT‐TME Score; F) Correlation of immune score and CT‐TME Score in high‐stage BRCA; G–K) Boxplots of CT‐TME score in high‐stage samples, stratified by molecular and clinical features which include PAM50 subtypes, histological types, status of ER, PR, and mutation of TP53. Basal, Basal‐like; Her2, Her2‐enriched; LumA, Luminal A; LumB, Luminal B; Normal, Normal‐like; ILC, invasive lobular carcinoma; IDC, invasive ductal carcinoma; Mixed, mixed Invasive Ductal and Lobular Carcinoma.

Figure 7

Molecular and clinical features associated with CT‐TME Score in breast cancer, and Portrayal of the tumor microenvironment landscape. A) An overview presents an analysis of the association between known molecular and biological processes and CT‐TME Score in high‐stage STAD; B) Correlation of immune scores and CT‐TME Score in high‐stage STAD; C–F) Boxplots of CT‐TME score in high‐stage samples, stratified by molecular and clinical features which include mutation of PI3CA and ARID1A, molecular subtype, status of EBV; G) t‐SNE plots of 10328 cancer patients from TCGA (containing 33 cancer types) and targets colored by cancer types. t‐SNE plots were generated using enrichment scores from cell InScores for 86 cell types. EBV, Epstein‐Barr virus; MSI, microsatellite instability; GS, genomically stable; CIN, chromosomal instability.