Retinoblastoma is a childhood retinal eye cancer caused by biallelic inactivation of the tumor suppressor gene RB1. This tumor is normally detected in children under the age of five years. Current models for retinoblastoma are based on animal models or established retinoblastoma cell lines. Modelling retinoblastoma in mice is not comparable, as the genetic background of biallelic Rb1 inactivation is not sufficient to induce tumor development. Retinoblastoma cell lines however, only display later stages of tumorigenesis. Especially for the analysis of early stages of retinoblastoma development, a model based on the differentiation of human embryonic stem cells (hESCs) into retina using 3D organoid technology was designed.

hESCs were successfully differentiated into retinal organoids, showing the formation of all seven retinal cell types with similar structural layering and arising in the same chronological order as observed for in vivo retinogenesis. To model retinoblastoma in vitro, this model was used to differentiate genetically modified hESCs with mono- (RB1^het hESCs) and biallelic inactivation of the RB1 gene (RB1^ko hESCs).

As expected, comparative differentiation showed the same phenotype for RB1 wildtype (RB1^wt organoids) and heterozygous RB1 organoids (RB1^het organoids), as heterozygous inactivation of RB1 does not lead to retinoblastoma formation. In organoids with background of biallelic RB1 inactivation (RB1^ko organoids) however, an increase in proliferating cone photoreceptors at day 152 (d152) of differentiation was detected, pointing towards immature and maturing cone photoreceptors as cells-of-origin in retinoblastoma. Transcriptome analysis showed an increase in the expression of retinoblastoma signature genes in RB1^ko organoids. However, between d126 and d152, the retinal layer of the RB1^ko organoids disintegrated. Mixing H9 RB1 wildtype cells (RB1^wt hESCs) and RB1^ko hESCs did not result in better structural integrity of the retinal layer during differentiation. Similar studies of colleagues showed that these disintegrated cells harbor a tumorigenic potential upon transplantation into the vitreous space of immunodeficient mice. Optimizing the model using thyroid hormone triiodothyronine (T3), hyperoxic culturing conditions and a later onset of inactivating the second RB1 allele might lead to prolonged structural integrity and therefore retinoblastoma development in vitro.

In heritable retinoblastoma, one RB1 mutation is already inherited from one parent. A few mutations were described resulting in parent-of-origin effects, as the development of retinoblastoma differs upon maternal or paternal inheritance. In 2009, it was discovered that RB1 is imprinted. A differentially methylated CpG island within intron 2 of RB1 (CpG85) serves as a promoter for an alternative RB1 transcript (RB1-E2B)and leads to allelic differences. RB1-E2B is only expressed from the paternally unmethylated CpG85. On the maternal allele, CpG85 is methylated and therefore, RB1-E2B expression is silenced. The hypothesized transcriptional interference of regular RB1 and RB1-E2B on the paternal allele could lead to skewing of regular RB1 expression in favor of the maternal allele.

If and how parent-of-origin effects are linked to the allelic differences established by genomic imprinting was addressed in the second part of the project. To study this hypothesis in a stem cell model, two modifications need to be introduced stepwise. As CpG85 gains methylation in some tissues with increasing age and in most hESCs and induced pluripotent stem cells (iPSCs), the cells need to be genetically modified first, to exhibit differential methylation of CpG85 and expression of RB1-E2B. In a second step, a mutation showing a strong parent-of-origin effect should be introduced on one of the RB1 alleles. Upon introduction of a second inactivating RB1 mutation on the remaining RB1 wildtype allele, differences in retinoblastoma development due to parent-of-origin dependent first mutation should be analyzed by subsequent comparative differentiation into neural retina.

First, hESCs were genetically modified by exchanging CpG85 by the EEF1A1 promoter on one allele. Expression of RB1-E2B under the EEF1A1 promoter should mimic the paternal allele, whereas the remaining methylated CpG85 allele mimics the maternal allele. The exchange of CpG85 by the EEF1A1 promoter resulted in onset of RB1-E2B expression in hESCs. Unfortunately, the EEF1A1 promoter gained methylation after exchange during culturing. As an alternative, two iPSC lines showing differentially methylated CpG85 could be used. Interestingly, these iPSC lines were generated from patients with Angelman- (AS) and Prader-Willi syndrome (PWS) due to imprinting defects which manifest in errors of establishment and/or maintenance of DNA methylation. Such an imprinting defect could also have an influence on the DNA methylation at CpG85. In these two iPSC lines with differentially methylated CpG85, RB1-E2B expression was detectable. However, these iPSC lines failed to differentiate reliably into retinal organoids and could not be used in this project.

Unfortunately, the analysis of parent-of-origin effects in retinoblastoma using a stem cell model seemed impossible and a more artificial system by overexpressing RB1-E2B might be more successful.