Abstract
Purpose :
Retinoblastoma (RB) is the most common eye cancer in children under the age of 5 years. RB is caused by mutations of the RB1 gene, which leads to tumor development in the eye, resulting in blindness, loss of the affected eye, or death in uncontrolled cases. A key question that is hotly debated is the cell type that gives rise to RB, which is important to enable the development of better treatment for RB. To understand the pathological changes associated with retinoblastoma, we propose to use spatial transcriptomics to profile primary RB tumours.
Methods :
We first completed a set of tissue optimisation experiments to determine the permeabilization conditions needed for release of the RNA from human retinal tissue and without the diffusion caused by over permeabilization. Next, the primary RB tumour samples were cryosectioned and we performed spatial transcriptomics (10X Visium). Bioinformatic analysis was performed using the SCUtility and Seurat packages. For QC, we filtered out spots with low detected genes, high percentage of mitochondrial content, and did not co-localised under the tissue.
Results :
Using non-linear dimensionality reduction with UMAP, we identified 10 transcriptional distinct clusters in RB tumour samples, indicating the presence of heterogenous cell populations. Next, we assign a cell-cycle score to each cell based on its expression of G2/M and S phase markers. Interestingly, the 10 RB clusters have variable composition of proliferating cells in different cell cycles, which is indicative of varying degree of tumorigenicity within the RB sample.
To determine the cell origin of RB, we analysed marker genes for the major retinal cell types. Interestingly, our results showed high expression of photoreceptor precursor markers (i.e. CRX, RXRG, THRB). We performed further cell type analysis using a set of gene modules for retinal cell markers. All RB clusters showed low cell score for horizontal cells, bipolar cells, amacrine cells, RGC and rod photoreceptors. Notably, cluster 9 showed high cone cell score, which is reflective of a cone cell origin for this particular RB cluster. Altogether, our results support RB originated from photoreceptors, and pathological changes in the cones are, at least in part, involved in RB tumorigenesis.
Conclusions :
In summary, we have successfully generated a spatial gene atlas for human RB, which would provide an important tool to study the genetic landscape and understand the biology of RB.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.