Clinical-proteomic classification and precision treatment strategy of chordoma

Abstract

Chordoma is a rare and heterogeneous mesenchymal malignancy, with distinct clinical and biological behaviors. Till now, its comprehensive clinical-molecular characteristics and accurate molecular classification remain obscure. In this research, we enroll 102 patients with chordoma and describe their clinical, imageological, and histopathological features. Through tandem mass tag-based proteomic analysis and nonnegative matrix factorization clustering, we classify chordoma into three molecular subtypes: bone microenvironment-dominant, mesenchymal-derived, and mesenchymal-to-epithelial transition-mediated pattern. The three subtypes exhibit discrete clinical prognosis and distinct biological attributes of osteoclastogenesis and immunogenicity, oxidative phosphorylation, and receptor tyrosine kinase activation, suggesting targeted therapeutic strategies of denosumab, S-Gboxin, and anlotinib, respectively. Notably, these approaches demonstrate positive treatment outcomes for each subtype in vitro and in vivo. Altogether, this work sheds light on the clinical-proteomic characteristics of chordoma and provides a candidate precision treatment strategy for chordoma according to molecular classification, underscoring their potential for clinical application.

Keywords: S-Gboxin; anlotinib; bone tumor; chordoma; denosumab; molecular classification; precision treatment strategy; proteomic analysis; targeted therapy.

Graphical abstract

Figure 1

Clinical-proteomic analysis of chordoma (A) Heatmap of clinical data for the constructed chordoma cohort. Kaplan-Meier curve of PFS (B), OS (C), and CSS (D) of the chordoma cohort. Abbreviations: PFS, progression-free survival; OS, overall survival; CSS, cancer-specific survival.

Figure 2

Imageological-proteomic analysis of chordoma (A) Representative CT manifestation of sclerotic and osteolytic chordoma. Protein levels (B) and Representative IHC images (C) of RETN in chordoma with or without bone destruction. GSEA enrichment analysis of DEPs between chordoma with or without bone destruction revealed statistic difference in degradation of the ECM (D) and osteopontin signaling (E). (F) Representative T1-weighted images of MRI for chordoma (hypointensity, isointensity, heterogeneous hyperintensity). (G) Kaplan-Meier curve of OS between hypo-/iso-intensity and heterogeneous hyperintensity. Protein levels (H) and representative IHC images (I) of AHNAK2 in chordoma hypo-/iso-intensity or heterogeneous hyperintensity. GSEA enrichment analysis of DEPs between hypo-/iso-intensity and heterogeneous hyperintensity revealed statistic difference in mTORC1 signaling (J) and G2M DNA damage checkpoint (K). Abbreviations: CT, computed tomography; IHC, immunohistochemistry; OS, overall survival; DEPs, differentially expressed proteins; GSEA, gene set enrichment analysis; ECM, extracellular matrix; MRI, magnetic resonance imaging.

Figure 3

Histopathological-proteomic analysis of chordoma (A) Representative H&E staining of conventional and dedifferentiated chordoma. (B) Kaplan-Meier curve of OS between conventional and dedifferentiated chordoma. Protein levels (C) and representative IHC images (D) of NDUFS3 in chordoma with conventional or dedifferentiated chordoma. GSEA enrichment analysis of DEPs between conventional and dedifferentiated chordoma revealed statistic difference in establishment of protein location to mitochondrial membrane (E) and notochord development (F). (G) Representative IHC images of Ki-67 > 5% and ≤5%. Protein levels (H) and representative IHC images (I) of PRPF4 in chordoma with Ki-67 > 5% or ≤5%. GSEA enrichment analysis of DEPs between Ki-67 > 5% and ≤5% revealed statistic difference in DNA replication (J) and degradation of the ECM (K). Abbreviations: H&E, hematoxylin-eosin; OS, overall survival; IHC, immunohistochemistry; DEPs, differentially expressed proteins; GSEA, gene set enrichment analysis; ECM, extracellular matrix.

Figure 4

Molecular classification of chordoma based on proteomic analysis and their clinical attributes (A) NMF clustering to identify molecular subtypes was performed using the proteomic data. (B) PCA plot shows the t-SNE distribution of chordoma samples in the identified three clusters. (C) Kaplan-Meier curve of OS among the identified three clusters in all patients with chordoma. (D) Forest plot of multivariate Cox regression analysis for the OS of patients with chordoma in our cohort. (E) Kaplan-Meier curve of OS among the identified three clusters in patients without distant metastasis and subtotal resection. Proportion of patients with chordoma with different tumor location (F), previous surgical treatment (G), and signals on T1-weighted images (H) among the identified three clusters. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Abbreviations: NMF, nonnegative matrix factorization; OS, overall survival.

Figure 5

Biological features of molecular subtypes (A) NMF clustering based on the function enrichment analysis of DEPs among the identified three clusters. (B) Representative multicolor IF of CD8, CD14, SP7, TRAP, as well as IHC of pan-cytokeratin in the chordoma subtypes reflecting different bone microenvironment and epithelial cells. The protein expression (log2-intensity) of key immune microenvironment markers (C) and bone microenvironment markers (D) among the identified three clusters. Mito protein expression (E) and Mito protein expression of CI (F) among the chordoma subtypes (log2-intensity). (G) The protein expression (log2-intensity) of previously reported mutant genes of chordoma among the identified three clusters. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Abbreviations: NMF, nonnegative matrix factorization; DEPs, differentially expressed proteins; IF, immunofluorescence; IHC, immunohistochemistry; CI, complex I.

Figure 6

Precision treatment strategy for chordoma subtypes (A) The correlation between chordoma cell lines and the identified three clusters. (B) Tumor volumes of denosumab- and PBS-treated PDX model (cluster 1, SA47). (C) IC50 of S-Gboxin for CH22 cells. MTT assay (D), clone formation assay (E), and OCR (F) of CH22 cells after the treatment of S-Gboxin. (G) Tumor volumes of S-Gboxin- and PBS-treated PDX model (cluster 2, SA20). (H) Tumor volumes of S-Gboxin- and PBS-treated CH22-bearing nude mice. (I) IC50 of anlotinib for U-CH2 cells. MTT assay (J) and clone formation assay (K) of U-CH2 cells after the treatment of anlotinib. (L) Tumor volumes of anlotinib- and PBS-treated PDX model (cluster 3, SA50). (M) Tumor volumes of anlotinib- and PBS-treated U-CH2-bearing nude mice. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Abbreviations: IC50, inhibitory concentration 50; PDX, patient-derived xenograft; PBS, phosphate buffer saline; OCR, oxygen consumption rate.