Pan-cancer dissection of vasculogenic mimicry characteristic to provide potential therapeutic targets

Abstract

Introduction: Vasculogenic mimicry (VM) represents a novel form of tumor angiogenesis that is associated with tumor invasiveness and drug resistance. However, the VM landscape across cancer types remains poorly understood. In this study, we elucidate the characterizations of VM across cancers based on multi-omics data and provide potential targeted therapeutic strategies.

Methods: Multi-omics data from The Cancer Genome Atlas was used to conduct comprehensive analyses of the characteristics of VM related genes (VRGs) across cancer types. Pan-cancer vasculogenic mimicry score was established to provide a depiction of the VM landscape across cancer types. The correlation between VM and cancer phenotypes was conducted to explore potential regulatory mechanisms of VM. We further systematically examined the relationship between VM and both tumor immunity and tumor microenvironment (TME). In addition, cell communication analysis based on single-cell transcriptome data was used to investigate the interactions between VM cells and TME. Finally, transcriptional and drug response data from the Genomics of Drug Sensitivity in Cancer database were utilized to identify potential therapeutic targets and drugs. The impact of VM on immunotherapy was also further clarified.

Results: Our study revealed that VRGs were dysregulated in tumor and regulated by multiple mechanisms. Then, VM level was found to be heterogeneous among different tumors and correlated with tumor invasiveness, metastatic potential, malignancy, and prognosis. VM was found to be strongly associated with epithelial-mesenchymal transition (EMT). Further analyses revealed cancer-associated fibroblasts can promote EMT and VM formation. Furthermore, the immune-suppressive state is associated with a microenvironment characterized by high levels of VM. VM score can be used as an indicator to predict the effect of immunotherapy. Finally, seven potential drugs targeting VM were identified.

Conclusion: In conclusion, we elucidate the characteristics and key regulatory mechanisms of VM across various cancer types, underscoring the pivotal role of CAFs in VM. VM was further found to be associated with the immunosuppressive TME. We also provide clues for the research of drugs targeting VM. Our study provides an initial overview and reference point for future research on VM, opening up new avenues for therapeutic intervention.

Keywords: immunosuppressive microenvironment; immunotherapy; pan cancer analysis; therapy target; tumor microenvironment; vasculogenic mimicry.

FIGURE 1

Expression, Somatic alteration, methylation and interactions of vasculogenic mimicry genes (VRGs). (A) Expression patterns of VRGs in pan-cancer. (B) Expression changes of VRGs in tumor tissue and para-tumor tissue in different tumor types. (C) Protein–protein interactions network of VRGs. (D) The amplification ratio of VRGs in different tumor types. (E) The deletion proportion VRGs in different tumor types. (F) The relationship between the expression and methylation levels of vascular mimicry-related genes in different tumor types.

FIGURE 2

Construction and characterization of vasculogenic mimicry score (VM score) at pan-cancer level (A) pan-cancer distribution of VM score. (B) Correlation of VM level and 50 clearly defined Hallmark in different tumor types. (C) Summary of the correlation between different Hallmark and VM. (D–I) VM was associated with epithelial-mesenchymal transition, hypoxia, angiogenesis, NOTCH signaling pathway, WNT/β-catenin signaling pathway, and tumor mutation burden in pan-cancer. (J) VM level is positively correlated with cell stemness in pan-cancer.

FIGURE 3

Vasculogenic mimicry (VM) is associated with unfavorable tumor phenotypes and clinical outcomes (A) A forest plot showing the hazard ratio of VM score in different cancer types. (B) A pan-cancer Kaplan–Meier curve shows survival associated with quantile stratified VM score. The log-rank test was used for the test of survival differences between all four groups. (C) A pan-cancer box plot shows that VM score is associated with tumor stages. The higher the tumor stages, the higher the VM score. (D) The box plot shows higher VM score in metastatic tumors. *p < 0.05; **p < 0.01; ***p < 0.001.

FIGURE 4

Vasculogenic mimicry (VM) is associated somatic mutation at pan-cancer (A, B) The top 10 genes with the highest mutation frequency of high (A) and low VM score groups (B) and the distribution of different mutation types in pan-cancer. (C, D) The proportion of transitions and transversions, and the overall distribution of the six different substitution of a single base in high (C) and low VM score groups (D). (E, F) Somatic interactivity mutations are co-occurring in the high VM score group (E), while IDH has extensive exclusiveness in the low VM score group (F). (G, H) Oncogenic pathways mutations in high VM score group (G) and low VM score group (H).

FIGURE 5

Vasculogenic mimicry (VM) is associated with immunosuppressive microenvironment (A) Heatmap shows the infiltration levels of different cellular components in the tumor microenvironment (TME) of different VM score groups. The results are based on six different TME algorithms. (B) VM score is positively associated with immune score, stromal score and ESTIMATE score calculated by ESTIMATE algorithm in different tumor types. (C) Differences in M1/M2 ratio between high and low VM score groups in different tumor types. (D) Differences in activated CD8⁺ T/Regulatory T cell ratio between high and low VM score groups in different tumor types. (E) Correlation between VM score and TIDE results. (F) The rose chart shows that VM score is lower in the samples with CTL. flag is true. (G) The rose chart shows the immunotherapy prediction results of TIDE. Patients with immune response have lower VM score. (H) Regulator prioritization results of TIDE. (I) Cellchat identifies dominant senders, receivers, mediators and influencers in the TGFb1 intercellular communication network. (J) The relationship between TGFβ1 and immune cell dysfunction in KIRC. (K) TGFβ1 expression levels in different cell types in KIRC TME. (L) Sankey diagram shows the distribution of different immune subtypes. (M) Correlation between VM and immune response steps. The heatmap shows the correlation between each immune step. The line between the VM and each step represents the correlation. Red represents positive correlation, and green represents negative correlation. (N) Immunity-related factors, including chemokines, MHC molecules, immunostimulators, and immunoinhibitors were highly expressed in high VM score group.

FIGURE 6

Single-cell analysis explains potential regulatory mechanisms of vasculogenic mimicry (A) Correlations between gene modules inferred by NMF. Three highly correlated gene programs were identified. (B) The expression of genes in the three gene programs. Red represents high expression, while blue represents low expression. The darker the color, the higher or lower the expression. (C) UMAP plot shows various cellular components in the KIRC tumor microenvironment (TME). (D) Top 10 transcription factors enriched in VM cells. The higher the specificity score, the greater the association with VM. (E) The stength of signals emitted and received by different cellular components in the TME. The larger the value on the y-axis, the more signals are received, and the larger the value on the x-axis, the more signals are emitted. The size of the point represents the amounts of cells. (F) Cell communication circle diagram. Demonstrating the communication landscape of tumor and VM cells when they serve as senders and receivers. (G) Demonstrate cell communication pathways in the TME and the contribution of different cells to different signaling pathways. The darker the color, the greater the contribution. (H) Chord diagram shows the interaction pathways between CAFs, VM and tumor cells. The wider the channel, the higher the strength. (I) Communication possibility based on ligand receptor pairs. Red represents a higher possibility of communication. (J) Cellchat identifies dominant senders, receivers, mediators and influencers in Collagen, HGF, FN1 and CPSG4 signaling intercellular communication network. (K) The violin plot shows the expression of HGF and CSPG4 pathway ligand receptor pairs in different cells. (L) Positive correlation was observed between HGF/MET score and VM, EMT in KIRC.

FIGURE 7

Vasculogenic mimicry (VM) is associated with drug sensitivity and immunotherapy outcome (A) The relationship between VM score and drug sensitivity (IC50 value). Each row represents a drug and drug target. The length of the line represents the correlation coefficient. The blue represents a negative correlation. The size of the point represents the statistical significance. (B, C, D) Kaplan–Meier curves show the difference in survival prognosis between high and low VM score groups in two SKCM cohorts (B, C) and a NSCLC cohort (D). The stacked column chart shows the immune response. Higher response levels were investigated in low VM score group.

FIGURE 8

VRS can predict prognosis and immunotherapy outcomes in KIRC patient (A) C-index for 101 machine learning algorithm combinations in training and validation cohorts. (B) Kaplan–Meier curves show survival differences between high and low risk groups in the TCGA (left) and E-MTAB-1980 cohort (Right). (C) Distribution of survival status and survival time of patients in high and low risk groups in the TCGA (left) and E-MTAB-1980 cohort (Right). (D) Time-dependent ROC curves of the model in the TCGA (left) and E-MTAB-1980 cohort (Right). (E) Kaplan–Meier curve show survival differences between patients in high- and low-risk groups in immunotherapy cohort. (F) Stacked column chart showing the proportion of patients benefiting from immunity within different risk groups. ROC, receiver operating characteristic; AUC, area under the ROC curve. CB, clinical benefit; ICB, intermediate clinical benefit; NCB, no clinical benefit.