A proteotranscriptomic approach to dissect the molecular landscape of human retinoblastoma

Abstract

Background: Retinoblastoma is a rare pediatric eye cancer caused by mutations in the RB1 gene, which regulates retinal cell growth. Early detection and treatment are critical for preventing vision loss and improving survival outcomes. This study aimed to perform an integrated proteotranscriptomic characterization of human retinoblastoma to provide a deeper understanding of disease biology and to identify novel therapeutic targets.

Methods: Paired tumor and adjacent retinal tissue samples were dissected from seven eyes. RNA sequencing and liquid chromatography-mass spectrometry were performed on the same samples. The spatially resolved cellular landscape was assessed using Imaging Mass Cytometry (IMC).

Results: The correlation between RNA and protein level was moderate with variations across different pathways, underscoring the value of an integrated proteotranscriptomic approach. IMC identified more than 67,000 single cells in 11 distinct clusters, including antigen presenting cells, T cells, stroma cells, vascular cells and two clusters of proliferating and CD44/c-Myc positive tumor cells. Antigen presenting cells expressed higher levels of CD68 in retinoblastoma compared to controls.

Conclusions: CD44+ and high-c-Myc-expressing tumor cells may represent cancer stem cells with possible involvement in metastasis, warranting further validation. Our multilayered approach could pave the way for enhanced molecular assessments and novel targeted therapies for human retinoblastoma.

Keywords: IMC; proteomics; proteotranscriptomics; retinoblastoma; transcriptomics; translational medicine.

Figure 1

Transcriptional profile of human retinoblastoma. (A): Unsupervised clustering using Principal Component Analysis (PCA). Each dot represents one sample. (B): Volcano plot visualizing differentially expressed genes (DEG) between retinoblastoma and retinal control tissue. Each dot represents one gene. The top ten DEG of both groups are labeled. (C): Heatmap visualizing DEG between retinoblastoma and retinal control tissue (Definition of DEG: log2FC > 1 or < -1 and adjusted p-value < 0.05). Basic demographic data is shown at the top. Each column represents one sample and each row one DEG. The number of DEG is given within the heatmap. The z-score represents a gene’s expression in relation to its mean expression by standard deviation units (red: upregulation, blue: downregulation). (D, E): Gene ontology (GO) analysis of up- (D) and downregulated (E) genes in retinoblastoma. The top ten enriched biological processes are shown in the dot plots. The size of the dots corresponds to the number of associated genes (count). The adjusted p-value of each GO term is indicated by color. The gene ratio describes the ratio of the count to the number of all DEG. (F, G) The bar plots visualize gene expression of the top 5 DEG involved in 6 of the most significantly (F) up- or (G) downregulated biological processes. The height of the bar represents mean expression and the error bar corresponds to standard deviation. Each dot represents one sample.

Figure 2

Proteotranscriptomic analysis of human retinoblastoma. (A) Experimental design. RNA-sequencing and liquid chromatography-mass spectrometry (LC–MS/MS) were applied to analyze the transcriptomic and proteomic profile of human retinoblastoma and retinal tissue specimens. (B) Comparison of the log2 fold change (FC) between retinoblastoma and retinal tissue on the transcriptomic (x-axis) and proteomic (y-axis) level. Each dot represents one gene/protein. Molecules with a significant difference between retinoblastoma and controls are shown in blue (proteomics or transcriptomics) or red (proteomics and transcriptomics). (C) Functionally grouped network analysis of enriched Gene ontology biological processes in which the up- or downregulated molecules were involved in. Enriched terms are visualized as nodes being linked based on the similarity of the factors associated with them. The node size represents the number of associated molecules. The pie charts visualize the percentage of molecules which were regulated on the transcriptomic (grey) or proteomic (orange) level. Each cluster is labeled with a representative term. (D) The top ten differentially expressed factors are shown for the three most affected groups of biological processes from (C) on the gene (grey) and protein (orange) level. Molecules with a significant change (adjusted p-value < 0.05) between retinoblastoma and controls are labeled with a red asterisk. (E) Density plots of Spearman’s correlation coefficients of RNA and protein levels of individual molecules between samples for retinoblastoma (yellow) and control retinal (blue) tissue. Dashed lines represent median Spearman correlation in each group.

Figure 3

Highly multiplexed spatially resolved single-cell proteomics of human retinoblastoma. (A) Experimental workflow. Post-enucleation tissue slices from whole eyes of 7 patients with retinoblastoma were analyzed using Imaging Mass Cytometry (IMC). Microscopically unaffected retinal tissue from the same eye was used as control tissue. ROI: Region of interest. (B) Uniform Manifold Approximation and Projection (UMAP) visualization showing Phenograph clustering of 67,058 single cells in 7 human retinoblastoma (58,299 cells) and 5 human control retinal tissue samples (8,759 cells). Cell type annotation is shown by color (see legend). UMAPs showing diagnosis or relevant marker proteins. The heatmap on the right demonstrates average marker expression in each cell cluster. (C) Violin plots visualizing protein expression of CD68 in antigen presenting cells (APC) in retinoblastoma (yellow) and controls (blue). ***: p<0.001. Representative images are shown on the right. Magnifications of the dashed white boxes in the image on the left are shown on the right (marked with a or b). Each scale bar corresponds to 100 µm. (D) c-Myc protein expression between the two tumor cell clusters from (B): proliferating tumor cells (pink) and CD44+ tumor cells (orange). ***: p<0.001. Representative images are shown on the right. Each scale bar corresponds to 100 µm.