Human iPSC differentiation to retinal organoids in response to IGF1 and BMP4 activation is line‐ and method‐dependent

Abstract Induced pluripotent stem cell (iPSC)‐derived retinal organoids provide a platform to study human retinogenesis, disease modeling, and compound screening. Although retinal organoids may represent tissue structures with greater physiological relevance to the in vivo human retina, their generation is not without limitations. Various protocols have been developed to enable development of organoids with all major retinal cell types; however, variability across iPSC lines is often reported. Modulating signaling pathways important for eye formation, such as those involving bone morphogenetic protein 4 (BMP4) and insulin‐like growth factor 1 (IGF1), is a common approach used for the generation of retinal tissue in vitro. We used three human iPSC lines to generate retinal organoids by activating either BMP4 or IGF1 signaling and assessed differentiation efficiency by monitoring morphological changes, gene and protein expression, and function. Our results showed that the ability of iPSC to give rise to retinal organoids in response to IGF1 and BMP4 activation was line‐ and method‐dependent. This demonstrates that careful consideration is needed when choosing a differentiation approach, which would also depend on overall project aims.


| INTRODUCTION
The development of in vitro retinal models has been driven by a lack of adequate animal models that recapitulate the structure and function of the human retina. Induced pluripotent stem cell (IPSC)-derived retinal organoids have been shown to have a wide range of applications, including the study of human retinogenesis, 1-3 disease modeling, 4,5 drug discovery, 6,7 and cell therapy. [8][9][10] Numerous protocols have been developed for the generation of retinal organoids that follow basic developmental principles of forebrain development and eye formation. Despite the ability of these protocols to give rise to laminated retinal organoids, variability in the propensity of iPSCs to give rise to various retinal cell types is often reported. Several groups including ours have reported variable laminar organization between samples differentiated from the same iPSC lines and the presence of non-neural cell types alongside the retinal structures. 7,11,12 Retinal development in vivo is controlled by a diverse set of signaling pathways and complex interactions between embryological tissues which affect the identity of the resultant cell population.
Forebrain development is orchestrated by a fine balance between a number of signaling pathways, including bone morphogenetic protein 4 (BMP4) and insulin-like growth factor 1 (IGF1). 13,14 Long-term maturation of retinal cells for an extended period of time is enhanced by retinoic acid (RA), taurine, and triiodothyronine (T3). 1,15,16 We used differentiation protocols that followed these principles in order to compare retinal differentiation efficiency of multiple iPSC lines across differentiation protocols that rely on activation of these pathways. 7,11,17,18

| RESULTS
To investigate the reproducibility of retinal organoid differentiation protocols across multiple iPSC lines, we differentiated three iPSC lines (WT1, WT2, and WT3. 7 ) from unaffected subjects using two differentiation protocols, designated as Method I and Method II (details are shown in Figure 1A). Key morphological features of

Significance statement
Retinal organoids were derived from three human induced pluripotent stem cell (iPSC) lines using two different differentiation approaches involving either bone morphogenetic protein 4 or insulin-like growth factor 1 signaling pathways. Retinal organoids were generated using both methods; however, the two different approaches produced bias toward certain retinal cell types. The results of this study suggest that careful consideration is needed when choosing a differentiation protocol and that overall efficiency to generate retinal organoids would depend on the signaling pathways that are modulated. Morphological observations at days 85 and 169 of differentiation showed that WT1 and WT2 organoids differentiated using both methods contained OV-like structures and in some cases presumptive RPE cells identified by their pigmented appearance. WT3 organoids were comparable with WT1 and WT2 in Method I but responded poorly to Method II ( Figure 1B,C). The capacity to give rise to neural retina with and without RPE differed across the lines and protocols ( Figure 1D). WT1 and WT2 produced more RPE with Method II, WT3 responded better to Method I overall, and the number of undefined structures was cell line-and methoddependent.
We performed quantitative PCR analysis to assess whether there was any difference in the expression of genes associated with various ret-  Figure S1). Expression of CRX and RHO was significantly higher in WT1 organoids differentiated with Method II comparing to Method I; similarly, differentiating WT2 with Method II resulted in significantly higher expression of NRL comparing to organoids differentiated with Method I. These results are unsurprising since Method II uses T3, which is known to encourage rod development, which is reflected in upregulation of NRL and RHO. 19 The largest difference between Methods I and II on differentiation outcome was observed in WT3 cells. Method I resulted in significant upregulation of all genes tested apart from RBPMS and RPE65, which corroborates the morphological observations reported in Figure 1D. it. Overall, WT3 cells did not respond well to Method II, which is reflected by gene and protein expression data (Figures 2 and 3 and Figure S1).
As a final test, we compared the functionality in these organoids by quantifying their ability to respond to light. We have recently shown that retinal organoids can respond to light, similar to the earliest light responses in mice at day 150 of differentiation. 7 Accordingly, we were able to record light-driven spikes from retinal organoids from WT1 and WT2 iPSCs. Based on gene expression and immunofluorescence data ( Figure S1, Figures 2A and 3A).  Figure 4A,B) or a decrease in spike rate (presumed OFF-center RGCs; Figure 4C). The medians of the calculated change of firing (COF) were not significantly different between the methods for WT1 but they were for WT2 ( Figure 4B, C). Indeed, WT2 organoids showed either a significant increase (Mann-Whitney test: P = .009) or decrease (P = .042) in COF medians with Method I.

| DISCUSSION
Generation of retinal organoids on a large scale is necessary in order to meet the growing demand for a model system which is predictive of human physiology and resembles key morphological and functional features. In this study, we used three human iPSC lines and differentiated them to retinal organoids using differentiation protocols activating either BMP4 (Method I) or IGF1 (Method II) signaling pathways; F I G U R E 4 Light-driven spiking activity recorded from presumed ON-Centre retinal ganglion cells (RGCs) and OFF-Centre RGCs. A, In the raster plot, each small vertical bar indicates the time stamp of a spike, where each row represents a different RGC. The left half illustrates the activity before stimulus onset and, separated by the red line, the right half the activity when exposed to light. B, The change of firing (COF) percentage values from presumed ON-Centre RGCs are scatter plotted for Method I and Method II of WT1-2 lines. The median is indicated as a red horizontal line and the interquartile range as a vertical line. Each symbol represents one RGC that showed more than 25% increased spiking after light onset. C, The COF percentage values from presumed OFF-Centre RGCs are scatter plotted in the same way as described above. We defined outliers as values which exceed three times the standard deviation (three-sigma rule). A total of five to six organoids were recorded from for each condition. Number of RGCs found to show increased firing in our data indicate that all cell lines were able to generate retinal cell types and the response to IGF1 and BMP4 was line-and methoddependent. Variability in the propensity of iPSCs to differentiate is widely reported in the literature and is thought to be a result of some factors including gene expression heterogeneity among stem cell populations, DNA methylation and histone modifications, and differences in endogenous signaling activities, such as BMP4. [20][21][22][23] This is unsurprising, since BMP4 has been shown to be involved in the differentiation of the anterior portion of the neural plate toward retinal neurones; in addition, IGF1 promotes induction of retinal fate. 14,24 Interestingly, it has also been reported that using IGF1 in combination with a BMP4 antagonist results in efficient generation of cone photoreceptors. 25 The interactions between these signaling pathways are complex, and it is plausible that depending on the endogenous levels of expression of key components of these pathways in the starting population of iPSCs it may be required to adapt differentiation protocols, which could include blocking BMP4 and simultaneously stimulating IGF1 in some cases. 23 This is further compounded by the addition of other components (eg, T3, N2, RA) after day 18 of differentiation, which can synergize or agonize the activation of BMP4 or IGF1 pathways in a different way in each of the two methods reported in this article. Furthermore, this variability is also reflected in the ability to establish light-sensitive signaling pathways. Only 7%-12% of all RGCs changed the activity when light activates phototransduction in photorecetors and bipolar cells relay that signal to RGCs, whereas 2%-5% of all RGCs were classified as potential ipRGCs. Except for WT2, there were no significant differences of light-induced RGC activity between BMP4 and IGF1 protocols. High proportion of ipRGCs could be a reflection of maturity of the retinal organoids. In neonatal mice, functional ipRGCs can be observed at birth (P0), which is earlier than the establishment of light-driven responses by RGCs (P10). Then, the number of ipRGCs decreases later in development. 26 In this study we found that the two differentiation approaches produced bias toward certain retinal cell types, which was iPSC lineand method-dependent. This shows that careful consideration is needed when choosing a differentiation protocol and that overall efficiency to generate retinal organoids would depend on the signaling pathways that are modulated.

CONFLICTS OF INTEREST
The authors indicated no potential conflicts of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available in the supplementary material of this article.