RUNX1-BMP2 promotes vasculogenic mimicry in laryngeal squamous cell carcinoma via activation of the PI3K-AKT signaling pathway

Abstract

Background: Laryngeal squamous cell carcinoma (LSCC) is one of the most common malignant tumors of the head and neck. Vasculogenic mimicry (VM) is crucial for tumor growth and metastasis and refers to the formation of fluid channels by invasive tumor cells rather than endothelial cells. However, the regulatory mechanisms underlying VM during the malignant progression of LSCC remain largely unknown.

Methods: Gene expression and clinical data for LSCC were obtained from the TCGA and Gene GEO (GSE27020) databases. A risk prediction model associated with VM was established using LASSO and Cox regression analyses. Based on their risk scores, patients with LSCC were categorized into high- and low-risk groups. The disparities in immune infiltration, tumor mutational burden (TMB), and functional enrichment between these two groups were examined. The core genes in LSCC were identified using the machine learning (SVM-RFE) and WGCNA algorithms. Subsequently, the involvement of bone morphogenetic protein 2 (BMP2) in VM and metastasis was investigated both in vitro and in vivo. To elucidate the downstream signaling pathways regulated by BMP2, western blotting was performed. Additionally, ChIP experiments were employed to identify the key transcription factors responsible for modulating the expression of BMP2.

Results: We established a new precise prognostic model for LSCC related to VM based on three genes: BMP2, EPO, and AGPS. The ROC curves from both TCGA and GSE27020 validation cohorts demonstrated precision survival prediction capabilities, with the nomogram showing some net clinical benefit. Multiple algorithm analyses indicated BMP2 as a potential core gene. Further experiments suggested that BMP2 promotes VM and metastasis in LSCC. The malignant progression of LSCC is promoted by BMP2 via the activation of the PI3K-AKT signaling pathway, with the high expression of BMP2 in LSCC resulting from its transcriptional activation by runt-related transcription factor 1 (RUNX1).

Conclusion: BMP2 predicts poor prognosis in LSCC, promotes LSCC VM and metastasis through the PI3K-AKT signaling pathway, and is transcriptionally regulated by RUNX1. BMP2 may be a novel, precise, diagnostic, and therapeutic biomarker of LSCC.

Keywords: Bone morphogenetic protein 2; Laryngeal squamous cell carcinoma; Metastasis; Runt-related transcription factor 1; Vasculogenic mimicry.

Fig. 1

Construction of the prognostic VMRG model in LSCC. A Screening for VMDEGs using the TCGA database. B Distribution of TMB according to DEGs. C Twelve prognostic genes extracted from the TCGA and GEO databases using univariate Cox regression analysis. D Correlation analysis of prognosis-related genes. E–F A prognostic model was constructed using LASSO Cox regression. G Risk genes in the VMRG prognostic model. H Distribution of risk scores. I Distribution of the survival status of patients. J A heatmap of the three prognostic signature genes. KM survival curves and the AUC for the entire set (K, O), in the training (L, P), testing (M, Q), and validation (N, R) sets

Fig. 2

Independent prognostic analysis of the VMRG prognostic model. A–D Univariate and multivariate Cox analyses of OS and PFS in LSCC. ROC curves for OS (E) and PFS (F) according to the risk score. C-index values of the risk scores for OS (G) and PFS (H). I KM curves of PFS based on the risk score. J OS nomogram, and (K) calibration curve for the nomogram; (L) ROC analysis of the nomogram; (N) Histogram statistics of clinical features

Fig. 3

TMB and TME infiltration. A Relationship between TMB and risk score. B Comparison of TMB scores between risk groups. C TMB analysis of high- and low-risk groups. D CIBERSORT analysis of immune cell infiltration in high- and low-risk groups. E ssGSEA analysis in high- and low-risk groups. F Comparison of immune check loci in high- and low-risk groups. G Relationship of the risk score and abundance of immune cells in LSCC

Fig. 4

Clustering analysis of VMRGs in LSCC. A–B LSCC samples (TCGA and GSE27020) were divided into 2 clusters using the consensus clustering method. C PCA analysis of cluster 1 and cluster 2. D t-SNE analysis of clusters 1 and 2. E Differential expression of prognostic genes in clusters 1 and 2. F–G The KM plot of the two clusters of patients with LSCC for determination of OS and PFS. H Differences in immune cell infiltration between the two clusters using ssGSEA. I–J The distribution of risk scores in the two clusters. K–L GVSA and GSEA enrichment analysis of differential genes between the two clusters

Fig. 5

BMP2 as the core prognostic VMRG for LSCC progression. A KM curve based on VM scores in LSCC. B Differentially expressed genes in high-and low-VM score groups. C–D LASSO and SVM-RFE screening of candidate diagnostic genes. E WGCNA of differentially expressed genes. F Correlations of VM scores with WGCNA modules. G Correlation between the blue module and VM scores. H Five algorithmic Venn diagram screening genes. I–J The AUC of BMP2 and TWIST1. K Heatmap of gene correlation analysis. L Differences in the expression of two core genes between high- and low-VM score groups. M–N Differential expression of two core genes between normal and tumor samples. O–P KM survival analysis of BMP2 and TWIST1 in LSCC. Q ROC curves of survival. R Differential expression of two core genes in GSE51985 data

Fig. 6

The expression of BMP2 in LSCC and its relationship with clinical features. A Expression analysis of BMP2 using qRT-PCR. B WB analysis of the expression of BMP2. C Multivariate Cox regression analyses. D, G Representative figures of IHC staining for BMP2. E, H Statistical analysis of IHC staining intensity. F Histogram of chi-square test. I KM survival curve of BMP2 on OS in TCGA and GEO cohorts. J VM vessels (CD3-/PAS +) and BMP2 were observed by staining LSCC. K Statistical analysis of the number of VM. L The analysis of the correlation between the number of VM and BMP2. *P < 0.05, **P < 0.01, ***P < 0.001, Student’s t-test and one-way ANOVA

Fig. 7

In vitro knockdown of BMP2 inhibits LSCC VM and metastasis. A-C Validation of BMP2 knockdown by qRT-PCR and western blotting. D Tubule formation assay of TU212 and TU686 cells. E 3D cell spheroid invasion assay of TU212 and TU686 cells. F–G Column graph in index of tubule formation and spheroid invasion ratio. H Immunofluorescence (IF) revealed that changes in the expression of BMP2 altered the expression of EMT markers. I Western blot analysis of EMT markers. Scratch wound-healing (J) and Transwell migration and invasion (K) assays performed in TU212 and TU686 cells. ***P < 0.001, ****P < 0.0001, Student’s t-test and two-way ANOVA

Fig. 8

In vivo knockdown of BMP2 inhibits LSCC VM and metastasis. Representative Matrigel plugs are shown, n = 3. B ImageJ software was used to count and analyze tubules. C Images of solid tumors from nude mice. D Curves of xenograft tumor growth in nude BALB/c mice (n = 4). E Western blot analysis of the expression of BMP2 in xenograft tumors. F IHC staining for BMP2 and CD34/PAS, CD34 − PAS + vasculogenic mimicry (VM) tubes. G IHC analysis of EMT markers in TU212 xenografts. H–J) Lung metastasis model. K Statistical analysis of nodules. L, M A zebrafish model was used to analyze the dissemination and metastasis of TU212 cells. **P < 0.01, ***P < 0.001, Student’s t-test

Fig. 9

BMP2 promotes LSCC VM and metastasis by activating the PI3K-AKT pathway. A KEGG enrichment analysis of differentially expressed genes. B Western blot analysis of p-PTEN, p-PDK, p-Akt, and Akt in LSCC. C Overexpression efficiency of BMP2 was verified using qRT-PCR. D, F–H Transwell migration, Transwell invasion, and scratch wound-healing assays were performed in three groups (DMSO, BMP2 + DMSO, and BMP2 + MK2206), respectively. E, I, J Tubule formation and spheroid invasion assays were conducted in the same three groups. K, L Western blot analysis of BMP2, p-PTEN, p-PDK, p-Akt, and Akt in the three groups in TU212 and TU686 cells. ***P < 0.001, ****P < 0.0001, Student’s t-test and two-way ANOVA

Fig. 10

BMP2 upregulation in LSCC results from increased transactivation by RUNX1. A CNV analysis of 12 prognostic VMRGs. B Venn diagram of predicted transcription factors. C Heatmap of correlation analysis. D, E QRT-PCR and western blotting shows the overexpression of RUNX1 in LSCC cells. F Diagram of RUNX1 binding sites. G qRT-PCR of the expression of RUNX1 and BMP2. I Western blot analysis of BMP2. J Immunofluorescence assay, RUNX1 (red fluorescence); BMP2 (green fluorescence). K Upon shearing by sonication, chromatin fragments ranged from 100 to 500 bp in size. L Western blotting was used to detect the RUNX1 in the ChIP assay. M DNA-binding sites electrophoresed on agarose gels. N Dual-luciferase reporter assays were used to analyze luciferase activity. O–P The levels of expression of BMP2 and RUNX1 were correlated according to IHC staining. Q Multivariate analysis of RUNX1 **P < 0.01, ***P < 0.001, ****P < 0.0001, one-way ANOVA and two-way ANOVA

Fig. 11

RUNX1 regulates BMP2 to promote vasculogenic mimicry in LSCC. A-B IHC results of RUNX1, BMP2 and CD34/PAS in LSCC tissue. C, E, F Tubule formation and spheroid invasion assays. D, G, H Transwell migration assay and transwell invasion assay. I Three groups (NC + NC, NC + shRUNX1, shRUNX1 + BMP2) of solid tumors were observed. J Curves of xenograft tumor growth in nude BALB/c mice (n = 4). K IHC staining for RUNX1, BMP2 and CD34/PAS in tumor tissues of nude mice. L Western blot analysis of EMT-related proteins and the AKT signaling pathway. M, N Kaplan–Meier curve for the 5-year overall survival rate of RUNX1 and BMP2. **P < 0.01, ***P < 0.001, ****P < 0.0001, one-way ANOVA and two-way ANOVA