In-depth proteomic profiling captures subtype-specific features of craniopharyngiomas

Abstract

Craniopharyngiomas are rare epithelial tumors derived from pituitary gland embryonic tissue. This epithelial tumor can be categorized as an adamantinomatous craniopharyngioma (ACP) or papillary craniopharyngioma (PCP) subtype with histopathological and genetic differences. Genomic and transcriptomic profiles of craniopharyngiomas have been investigated; however, the proteomic profile has yet to be elucidated and added to these profiles. Recent improvements in high-throughput quantitative proteomic approaches have introduced new opportunities for a better understanding of these diseases and the efficient discovery of biomarkers. We aimed to confirm subtype-associated proteomic changes between ACP and PCP specimens. We performed a system-level proteomic study using an integrated approach that combines mass spectrometry-based quantitative proteomic, statistical, and bioinformatics analyses. The bioinformatics analysis showed that differentially expressed proteins between ACP and PCP were significantly involved in mitochondrial organization, fatty acid metabolic processes, exocytosis, the inflammatory response, the cell cycle, RNA splicing, cell migration, and neuron development. Furthermore, using network analysis, we identified hub proteins that were positively correlated with ACP and PCP phenotypes. Our findings improve our understanding of the pathogenesis of craniopharyngiomas and provide novel insights that may ultimately translate to the development of craniopharyngioma subtype-specific therapeutics.

Figure 1

Overall scheme of the quantitative proteomic analysis of craniopharyngioma. Schematic showing the proteomic analysis of craniopharyngiomas (ACPs and PCPs) analyzed by label-free quantification and TMT labeling experiments.

Figure 2

Results of the label-free quantification approach. (A) Volcano plot depicts differential expression between two craniopharyngioma subtypes (ACP and PCP) using label-free quantification data. (B) Principal component analysis of the 11-tumor label-free proteomic data. (C) The heatmap represents unsupervised hierarchical clustering of the 11 tumor samples using 923 proteins that were differentially expressed as identified by label-free quantification. (D) Gene ontology (GO) enrichment analysis of the label-free data. The red points indicate upregulated proteins in ACPs, and blue points indicate upregulated proteins in PCPs.

Figure 3

Results of—TMT quantification approach. (A) Volcano plot depicting the differential expression analysis of the two craniopharyngioma subtypes (ACP and PCP) using TMT quantification data. (B) Principal component analysis of the 9-tumor TMT quantification data. (C) The heat map represents unsupervised hierarchical clustering of the 9 tumor samples using 1474 proteins that were differentially expressed and identified in the TMT quantification data. (D) Gene ontology (GO) enrichment analysis of TMT quantification data. The red points indicate upregulated proteins in ACPs, and blue points indicate upregulated proteins in PCPs.

Figure 4

Comparison between label-free and TMT quantification methods. (A) Venn diagram showing the comparison of differentially upregulated proteins between ACPs in both datasets. (B) Venn diagram showing the comparison of differentially upregulated proteins between PCPs in both datasets. (C) Correlation analysis of fold changes in the label-free and TMT quantification data. (D) Validation of the proteomic data using protein expression. Western blotting was performed to measure EPCAM, P4HB and β-actin.

Figure 5

Comparison analysis of gene ontology enrichment (biological process category) between ACP and PCP. GO analysis of the common differentially expressed proteins. The bar graph represents the enrichment score, − log₁₀ (p-value), as heights. Each line illustrates the number of proteins enriched per GO term.

Figure 6

Network model of proteome characterization between ACP and PCP. A protein interaction network model was generated by integrating two proteomic datasets. Red nodes represent upregulated expression in ACPs, and blue nodes represent upregulated expression in PCPs. The color of the outer node indicates the differential expression levels revealed through the label-free quantification, and the color of the inner node indicates the differential expression level revealed through the TMT quantification. The gray line represents protein–protein interactions derived from the STRING database.

Figure 7

Prediction of the binding sites of transcription factors and the network of TFs and downstream targets. (A) Predicted TFs (MYC and ARNT) in ACPs. (B) Predicted TFs (ELK1 and ERR1) in PCPs. The color of the outer node indicates the differential expression levels revealed through label-free quantification, and the color of the inner node indicates the differential expression level revealed through TMT quantification.