Transcriptomic and proteomic profiling identifies feline fibrosarcoma as clinically amenable model for aggressive sarcoma subtypes

Abstract

Fibrosarcomas (FSA) are malignant mesenchymal tumors characterized by low chemo- and radiosensitivity. Development of novel treatment strategies for human adult FSA is hindered by the low incidence and the absence of suitable clinical models. Interestingly, aggressive FSA occur more frequently in domestic cats, hence potentially representing a clinically amenable model to assess novel therapies such as targeted imaging or theranostics. However, a lack of molecular characterization of FSA and adjacent normal tissue (NT) in both species hinders identification of tumor-specific targets and undermines the translational potential of feline FSA. Combining laser-capture microdissection, RNAsequencing and liquid chromatography-tandem mass spectrometry, we perform comprehensive profiling of 30 feline FSA and matched skeletal muscle, adipose and connective tissue. Clear inter-tissue differences allow identification of significantly upregulated and tumor-exclusive features that represent potential targets for diagnostic and therapeutic approaches. While feline FSA are characterized by hyperactive EIF2, TP53 and MYC signaling, immune-related and neuronal pathways emerge as modulators of tumor aggressiveness and immunosuppression. A high degree of molecular similarity with canine and adult FSA allows identification of tumor targets that are conserved across species. Significant enrichment in DNA repair pathways in feline FSA correlate with aggressive clinical behavior in human soft-tissue sarcoma. Finally, we leverage the molecular profiles to identify vulnerabilities, including sensitivity to ATR and PARP inhibition as potential treatment for feline FSA. In conclusion, this detailed landscape provides a rich resource to identify target candidates and therapeutic vulnerabilities within and across species and supports feline FSA as relevant models for the human disease.

Keywords: Comparative oncology; Disease model; Feline injection-site sarcoma; LC-MS/MS; Laser-capture microdissection; RNAsequencing; Soft-tissue sarcoma; Tumor targeting.

Fig. 1

Overview of proteins and transcripts identified in feline fibrosarcoma and surrounding peritumoral tissue of 30 cases. A) Pie chart illustrating the anatomical locations of the patients’ tumors. B) Experimental approach to isolate FSA, adipose tissue (AT), connective tissue (CT) and skeletal muscle (SM) from FFPE sections using LCM, followed by LC-MS/MS and RNAseq. C) Number of samples per tissue type analyzed using LC-MS/MS or RNAseq. D) An overview of the total count of identified proteins and transcripts across all samples and tissue types. Parts in panels A and C were created with BioRender.com.

Fig. 2

Transcriptomic analysis of FSA and matched NT highlights unique expression patterns of tumors compared to the surrounding tissue. A) 3D-Principal Component Analysis plot of genes detected in tumor and each NT from 30 feline FSA. PCA was performed using all identified genes. B) Volcano plots featuring DEG between the tumor and each NT. Cutoff values for significance (|log₂FC| ≥ 1 and FDR < 0.05) are indicated using grey dashed lines. P-values were calculated using ANOVA. C) Venn diagram illustrates all upregulated genes in the tumor compared to each NT, using a threshold of log₂FC of ≥ 1 and FDR < 0.05. D) ORA analysis using KEGG pathway of the 436 commonly upregulated genes in tumor. E) Top 20 most differentially expressed genes in tumor compared to each NT (log₂FC ≥ 1, FDR < 0.05). F) Violin plot of selected transcripts from E to illustrate expression in tumor and adjacent NT (****P < 0.0001, calculated using ANOVA). The dashed lines indicate the interquartile range and the dotted line represents the median. G) (left) Heatmap of all T samples using all transcripts. (right) GSEA using HALLMARK and REACTOME pathway databases of highly (top) and lowly (bottom) expressed RNAs in FSA. T = tumor, AT = adipose tissue, CT = connective tissue, SM = skeletal muscle.

Fig. 3

Proteomic analysis of FSA and its NT highlights unique expression patterns of tumor compared to the surrounding. A) 3D-Principal Component Analysis plot of proteins detected in tumor and each NT from 27 feline FSA. PCA was performed using all identified proteins. B) Volcano plots featuring differentially expressed proteins between the tumor and each NT. Cutoff values for significance (|log₂FC| ≥ 1 and FDR < 0.05) are indicated using grey dashed lines. P-values were calculated using ANOVA. C) Venn diagram illustrates all upregulated proteins in the tumor compared to each NT, identified with a threshold of log₂FC of > 1 and FDR < 0.05. D) GSEA using HALLMARK pathways on all commonly upregulated proteins in tumor compared to NT. E) Top 5 most abundant proteins detected in T compared to each NT (log₂FC > 1, FDR < 0.05). F) Venn diagram showing proteins identified across all tissue types. G) List of the 6 tumor-exclusive proteins that were detected in every single tumor sample. H) Kaplan-Meyer curves for high/low IKBIP, MARCKSL1 and COPZ1 expressing tumors using the TCGA-SARC dataset. I) Venn diagram of the overlap of proteins and transcripts significantly upregulated in T vs all NT. J) Top canonical pathways were detected in T vs AT, CT and SM using ingenuity pathway analysis (IPA). Values on the top indicate the number of detected targets / number of targets attributed to the respective dataset; values in brackets represent p-values. K) Top upstream regulators detected in T vs AT, CT and SM using IPA. The top indicates the predicted activation status, and the values in brackets represent p-values. L) The activation status of the top 20 upregulated canonical pathways in T vs AT, CT and SM.

Fig. 4

Identification of transcriptomic and proteomic features in feline FSA associated with differences in clinical behavior. A) Volcano plot of differentially expressed genes between highly aggressive (HA) and lowly aggressive (LA) tumors. Cutoff values for significance: |log₂FC| ≥ 2 and p = 0.01. B) Heatmap showing unsupervised clustering of significant differentially expressed genes from A. C) Ridge plot showing Gene Set enrichment analysis of differentially expressed genes from B. D) Top 20 pathways with the highest variance between HA and LA tumors in the proteomic dataset assessed using single sample GSEA and Wikipathway. E) PCA of all tumors using all transcripts with the three molecular clusters C1 (fibroblastic), C2 (neuronal-like) and C3 (inflammatory) indicated in dashed lines. F-H) ORA of PC1 loadings > 0.02 (F), PC2 loadings > 0.02 (G), or PC2 loadings <-0.02 (H) from E using Gene Ontology of Biological Processes. I) Heatmap of unsupervised clustering results, based on transcripts restricted to neuronal and immune-related features, differentiates between the three clusters C1 – C3.

Fig. 5

Cross-species assessment of feline, human and canine FSA. A) Competitive gene set testing to compare proteomic data from feline FSA and ‘human other FS’. B) ORA of the 199 shared highly expressed proteins between feline FSA, human other FS and human MFS. C) Competitive gene set testing to compare feline FSA proteomic data to canine FSA. D) Top 20 pathways from ssGSEA using Wikipathway for proteomic data from all canine and feline FSA. E) Venn diagram showing overlap of proteins identified in feline and canine FSA. F) ORA using KEGG pathways for the unique feline proteins in E. G – H) Kaplan-Meyer plot showing overall survival (G) and disease-free survival (H) in the TCGA-SARC cohort based on high/low expression of the 10-gene signature identified in F. I – L) Sensitivity of feline FSII (I, K) and FSIII (J, L) cells to various drugs after exposure for 72h (I, J) or 6 days (K, L). Data shown are mean from n = 6 assays ±SD. The colored lines indicate nonlinear fit IC50 curves for each cell line. M) Table detailing IC50 for both cell lines, as calculated from I – L.