Culture supplements impact retinal organoid morphology
We compared two previously published retinal organoid differentiation protocols, referred to as Protocol 1 [10, 13] (Fig. 1A) and Protocol 2 [38] (Fig. 1B), and a novel supplemented protocol, Protocol 3 (Fig. 1C). For all protocols, differentiation was promoted by switching the hiPSC culture medium to E6, which was then supplemented with N-2 to direct anterior neural fate differentiation and promote the emergence of self-organised NR-like structures [14]. Following excision at D28, the NR-like structures were cultured transiently with FGF2 to promote proliferation and growth [16], and with B27 supplement to support long-term viability [39].
For Protocol 1, the floating NR-like structures remained in the B27-supplemented medium until maturation (Fig. 1A). As shown in Additional file 1, the Protocol 1 organoids expressed specific photoreceptor markers, such as recoverin (RCVRN) at D105 (Additional file 1: Fig. S1A), and NR2E3 (Additional file 1: Fig. S1B) and rhodopsin at D300 (Additional file 1: Fig. S1C), by immunofluorescence (IF) studies. Consistent with a previous report [10], a presumptive outer nuclear layer (ONL) did not form in these organoids, hence, marker expression was mainly concentrated in rosette-like structures. Therefore, to promote lamination of the retinal organoids, for Protocol 2 (Fig. 1B), we added FBS to the culture medium at D35 [8] and switched to B27 -VitA at D85 to enhance photoreceptor maturation [38, 40]. For Protocol 3 (Fig. 1C), we also added taurine at D42, to further promote lamination, structure and survival of the organoids [9]. Moreover, we switched to B27 -VitA at D65, at the same time as we added RA to promote photoreceptor differentiation; we removed RA at D120 to preserve photoreceptor maturation [8]. Lastly, we added N-2 at D85 to help survival of post-mitotic cells during long-term culture [39].
Both Protocols 2 and 3 gave characteristic retinal organoids with a presumptive ONL that was visible from D100 (Fig. 1D). Also at D100, we exclusively observed a nascent brush border, corresponding to the inner segments (IS) and outer segment-like (OS-like) structures of the photoreceptors, in Protocol 2 organoids; the brush border evolved and was clearly visible at D150. At D150, a nascent brush border was visible in Protocol 3 organoids (Fig. 1D). Quantitative analysis indicated that the brush border at D150 was significantly longer (p < 0.0001) in Protocol 2 (44.0 µm ± 2.4, n = 47) than Protocol 3 (23.1 µm ± 1.2, n = 60) (Fig. 1E) organoids. At D180, this difference remained visible (Fig. 1D) and significant (63.4 µm ± 3.4, n = 44, Protocol 2 vs. 45.3 µm ± 2.1, n = 44, Protocol 3; p < 0.0001) (Fig. 1E). In mature D225 organoids, no clear difference was visible between protocols (Fig. 1D) but quantitative analysis showed that the brush border was still significantly longer (p < 0.05) in Protocol 2 (72.2 µm ± 3.4, n = 29) than in Protocol 3 (65.9 µm ± 1.7, n = 79) organoids (Fig. 1E). We also measured the width of the ONL at D150, D180 and D225 (Fig. 1F). We did not observe differences between protocols until D225, when the Protocol 2 organoids showed a significantly (p < 0.01) thinner ONL (47.3 µm ± 2.4, n = 29) than the Protocol 3 organoids (54.2 µm ± 1.2, n = 67). This was not due to a significant difference in the overall organoid size (data not shown).
Taken together, culture supplements have an impact on organoid morphology, namely brush border formation during maturation and ONL width at maturity.
Retinoic acid delays photoreceptor gene expression
As Protocol 1 did not give laminated organoids, we continued our studies only with Protocols 2 and 3. For these protocols, the culture conditions were identical until D42, (Fig. 1B and C), therefore, to avoid introducing additional parameters that could affect comparisons, NR-like structures collected from the same hiPSC dish were cultured in parallel without (Protocol 2) or with (Protocol 3) culture supplements until D225 (latest time point tested). We first assayed gene expression by qPCR analyses. As shown in Additional file 1, the NR-like structures collected at D35 already expressed the transcription factors OTX2 (Additional file 1: Fig. S2A), SIX3 (Additional file 1: Fig. S2B), RAX (Additional file 1: Fig. S2C) and VSX2 (Additional file 1: Fig. S2D) involved in retinal specification, and the expression profiles were similar for both protocols.
The expression profiles of the transcription factors CRX (Fig. 2A), NR2E3 (Fig. 2B) and NRL (Fig. 2C) directing photoreceptor cell fate, were also comparable between Protocol 2 and Protocol 3 organoids, although differences were observed at specific time points. CRX (Fig. 2A), which drives photoreceptor differentiation, was expressed from D35 and levels increased rapidly up to D90. However, the addition of RA in Protocol 3 at D65 slowed CRX expression in Protocol 3 organoids compared to Protocol 2. Furthermore, at D120, when RA was removed from Protocol 3, CRX expression decreased in these organoids, whereas it increased in Protocol 2 organoids. The expression of NR2E3 (Fig. 2B) and NRL (Fig. 2C), which encode proteins that partner with CRX to specifically drive rod differentiation, began at D60, and interestingly, for the same protocol, the expression profiles of the two genes over time were almost identical, highlighting that they are closely regulated during retinal development [41]. Furthermore, in Protocol 3 organoids, NR2E3 and NRL expression continually increased from D60 to D120, i.e. during the RA-treatment window, consistent with previous studies showing that RA supplementation promotes rod differentiation [20]. By contrast, expression levels plateaued for Protocol 2 organoids from D90. Once RA was removed from Protocol 3 at D120, NR2E3 and NRL expression levels decreased in these organoids until D150, and then remained relatively stable. By contrast, NR2E3 and NRL expression levels in Protocol 2 organoids increased from D150 to D225.
The expression of the pan photoreceptor markers, recoverin (RCVRN; Fig. 2D) and rhodopsin kinase (GRK1; Fig. 2E), began between D35 and D60, and increased up to D225 in Protocol 2 and Protocol 3 organoids. However, after removal of RA from Protocol 3 at D120, RCVRN (Fig. 2D) and GRK1 (Fig. 2E) expression levels remained lower in Protocol 3 organoids than in Protocol 2 organoids. Expression of the rod-specific marker rhodopsin (RHO) was detected from D120 in both types of organoids, but it increased sharply in Protocol 2 organoids from D150 (Fig. 2F). Interestingly, the expression of the cone-specific marker, cone arrestin-3 (ARR3), which began at D35 regardless of the protocol, also increased sharply from D90 to D225 in Protocol 2 organoids (Fig. 2G), whereas the addition of RA in Protocol 3 stabilised ARR3 expression at low levels until D120. When RA was then removed, ARR3 expression moderately increased up to D150, and then remained stable until D225. The expression of the cone-specific red–green opsin (OPN1MW) began at D90 for Protocol 2 organoids and continually increased until D225 (Fig. 2H). By contrast, OPN1MW expression was undetectable in Protocol 3 organoids until removal of RA at D120, and from then on, it was expressed at lower levels than in Protocol 2 organoids.
As shown in Additional file 1, the expression profiles of other retinal cell type markers, RGCs (BRN3A; Additional file 1: Fig. S2E), amacrine (GAD2; Additional file 1: Fig. S2F), bipolar (PKCα; Additional file 1: Fig. S2G), Müller glia (GLAST1; Additional file 1: Fig. S2H) and horizontal (LIM1; Additional file 1: Fig. S2I) cells, were comparable between protocols although BRN3A, GAD2 and PKCα reached higher levels in Protocol 2 organoids, and GLAST1 and LIM1 in Protocol 3 organoids. By D225, the markers were expressed at similar levels in both types of organoids. Furthermore, regardless of protocol, BRN3A expression was lost from D120, in accordance with previous studies reporting the loss of RGCs in mature organoids [11, 14, 33].
Taken together, comparable expression profiles are observed in organoids generated using non-supplemented or supplemented culture conditions; however, under supplemented conditions, the presence of RA appears to restrain cone fate differentiation.
Supplemented media promotes rod-rich organoids and preserves ONL integrity
Due to the differences observed in the brush border length, ONL width and gene expression profiles between Protocol 2 and Protocol 3 organoids, we next analysed common photoreceptor-specific markers by IF studies at mid- (D150) and mature (D225) stages of differentiation (Fig. 3). CRX was mainly restricted to the nuclei of the photoreceptors, and at D150, the ONL of the Protocol 2 organoids (Fig. 3A) was less tightly packed than that of Protocol 3 organoids (Fig. 3B). This difference was even more pronounced at D225 when the ONL was visibly thinner in Protocol 2 organoids (Fig. 3C and D), consistent with quantification of bright-field images (Fig. 1F). RCVRN was expressed throughout the length of the photoreceptors, and at D150, RCVRN expression extended further from the ONL in Protocol 2 (Fig. 3A) than in Protocol 3 (Fig. 3B) organoids. This was in accordance with the longer brush border observed by bright-field microscopy (Fig. 1D). At D225, RCVRN expression in Protocol 3 organoids extended further from the ONL than at D150 (Fig. 3D), indicative of the prolongation of the IS and OS-like structures and consistent with the longer brush border measurements at this time point (Fig. 1E). However, at D225 the deterioration of the ONL in Protocol 2 organoids (Fig. 3C) did not allow the comparison of the length of the IS and OS-like structures by RCVRN expression with that of the Protocol 3 organoids (Fig. 3D).
Similar to the previous CRX results, NR2E3 was expressed in photoreceptor nuclei and the ONL appeared less tightly packed in Protocol 2 organoids at D150 (Fig. 3E), compared to Protocol 3 organoids (Fig. 3F), and this difference was even more flagrant at D225 (Fig. 3G and H). Quantification of the area of NR2E3 fluorescence, relative to Hoechst fluorescence in the ONL shown in Additional File 1: Fig. S3, showed no significant differences between the two types of organoids (Fig. 3M). We detected expression of the cone visual pigment red–green opsin (RG opsin) at D150 (Fig. 3E and F) and D225 (Fig. 3G and H) for both protocols and clearly observed more RG opsin-positive cells in Protocol 2 organoids (Fig. 3E and G), as compared to Protocol 3 organoids (Fig. 3F and H). Consistently, quantification of the relative area of RG opsin fluorescence within the ONL showed that it was significantly higher (p < 0.5; n = 3) for Protocol 2, compared to Protocol 3, organoids at both time points (eightfold at D150 and twofold at D225; Fig. 3N).
In accordance with the RG opsin results, we observed more cone arrestin-positive cells in the ONL of the Protocol 2 organoids at D150 (Fig. 3I) and D225 (Fig. 3K), compared to Protocol 3 organoids (Fig. 3J and L). This increase was confirmed to be statistically different (p < 0.5; n = 3) by quantification of the relative area of arrestin fluorescence relative to the Hoechst fluorescence (Additional file 1: Fig. S3) in the ONL (twofold at D150 and threefold at D225; Fig. 3O). Rhodopsin expression was abundant in both protocols and detected throughout the length of the photoreceptors at D150 (Figs. 3I and J) and D225 (Fig. 3K and L). However, at D225, the rhodopsin-positive cells for Protocol 3 organoids (Fig. 3L) appeared longer than those for Protocol 2 (Fig. 3K). Interestingly, the quantification of the area of rhodopsin fluorescence, relative to Hoechst fluorescence in the ONL shown in Additional File 1: Fig. S3, suggested that Protocol 2 had significantly twofold higher (p < 0.5; n = 3) rhodopsin levels at D150, whereas Protocol 3 organoids showed a tendency towards higher expression at D225 (Fig. 3P).
Lastly, we analysed the percentage of rods and cones in organoids for each protocol, and at D150, rods predominated (68.9% ± 2.58% Protocol 2 and 70.8% ± 5.49% Protocol 3) as compared to cones (31.1% ± 2.58% Protocol 2 and 29.3% ± 5.49% Protocol 3) regardless of the protocol used (Fig. 3Q). By contrast, at D225, the percentage of cones (50.3% ± 1.54%) in Protocol 2 organoids had increased to become equivalent to that of rods (49.7% ± 1.54%). Conversely, the percentage of cones (23.8% ± 2.70%) in Protocol 3 organoids had decreased, rendering the rods clearly predominant (76.2% + 2.70%) (Fig. 3R). We further confirmed the changing rod population over time semi-quantitatively using western blot analysis of another mature rod marker, PDE6B (Fig. 3S). At D150, PDE6B expression was significantly twofold higher (p < 0.5; n = 2) in Protocol 2, than Protocol 3, organoids. At D180, PDE6B expression levels in Protocol 2 organoids were closer to those of Protocol 3 but still significantly 1.2-fold higher (p < 0.5; n = 2). However, at D225, the situation had reversed and Protocol 3 organoids showed significantly 1.2-fold higher PDE6B expression levels (p < 0.5; n = 2) than Protocol 2 organoids. These data confirm the earlier differentiation of Protocol 2 organoids and the predominant rod population in mature Protocol 3 organoids.
The qPCR results suggested that RA plays a pivotal role in photoreceptor gene expression. Therefore, to assay whether the observed morphological changes were exclusively due to the presence of RA, we differentiated in parallel retinal organoids under Protocol 3 conditions with or without RA (Protocol 3 -RA). Similar to our observations in Protocol 2 organoids, we detected a brush border earlier in Protocol 3 -RA organoids that evolved more rapidly than that of Protocol 3 organoids (Fig. 4A). At D225, IF studies to assess CRX and RCVRN expression detected a less tightly packed ONL in the Protocol 3 -RA organoids as compared to Protocol 3 organoids (Fig. 4B). In addition, Protocol 3 -RA organoids contained more cones as determined by RG opsin and arrestin expression (Fig. 4B), compared to Protocol 3 organoids. Quantification of the area of arrestin fluorescence, relative to Hoechst fluorescence in the ONL shown in Additional file 1 (Fig. S4), confirmed a significant threefold difference (p < 0.5; n = 3; Fig. 4C). No significant differences in rhodopsin expression were observed for Protocol 3 and Protocol 3 -RA organoids (Fig. 4B and C), although Protocol 3 organoids showed a tendency towards higher levels. These data are reminiscent of the observations at D225 for Protocol 2 and Protocol 3 organoids (Fig. 3P). Lastly, we analysed the percentage of rods and cones and detected a higher percentage of cones (45.7% ± 9.17%) in Protocol 3 -RA organoids as compared to Protocol 3 organoids (21.4% ± 4.02%) (Fig. 4D). Conversely Protocol 3 organoids had a higher percentage of rods (78.6% ± 4.02%) compared to Protocol 3 -RA organoids (54.3, ± 9.17%). Notably, the cone and rod percentages of the D225 Protocol 3 -RA organoids were almost identical to those of Protocol 2 organoids.
In conclusion, under non-supplemented conditions, and specifically in the absence of RA, the ONL of the retinal organoids is less tightly packed and the photoreceptors appeared shorter in comparison to supplemented conditions. In addition, the absence of RA results in cone-rich organoids, whereas the presence of RA result in rod-rich organoids.
Supplemented media improves the structure of the photoreceptor layer
We next assayed the ultrastructure of the ONL and the photoreceptors of the retinal organoids using transmission electron microscopy (EM) at D180 (Fig. 5). As suggested by the IF studies, the nuclei and the cell bodies of the photoreceptors were less organised in Protocol 2 organoids (Fig. 5A), as compared to their perfect alignment forming a highly structured ONL in Protocol 3 organoids (Fig. 5E). Despite this organisational difference, we observed characteristic features of mature photoreceptors for both Protocol 2 and 3 organoids, such as the outer limiting membrane (OLM) and the connecting cilium (CC) linking IS and OS-like structures that contained rudimentary photoreceptor discs (Fig. 5B and F). In parallel we evaluated the retinal organoid brush border using scanning EM, which revealed differences in photoreceptor shape between protocols (Fig. 5C and G). The photoreceptors of Protocol 2 organoids were larger, rounder and shorter, consistent with the shape of cones (Fig. 5C), whereas the photoreceptors of Protocol 3 organoids were thinner and longer, consistent with the shape of rods (Fig. 5G). Similar to the transmission EM results, we observed the CC and the OS-like structures of the photoreceptors on the surface of both Protocol 2 and 3 organoids (Figs. 5D and H).
Taken together, under supplemented conditions, the photoreceptor layer of retinal organoids is more highly organised compared to non-supplemented conditions. In addition, the cone-rich (non-supplemented conditions) and rod-rich (supplemented conditions) organoids can be distinguished by their surface morphology.
Supplemented media promotes stratification and elongation of the photoreceptors
To better characterise the differential shape and spatial distribution of the photoreceptors at the surface of the retinal organoids observed by scanning EM, we performed IF studies on whole mature organoids (D225) using arrestin and rhodopsin as markers of cone and rod photoreceptors, respectively. Confocal analysis at low magnification of the surface of Protocol 2 organoids showed a well-distinguished, regular distribution of arrestin-positive cells, and intensely stained but irregularly distributed rhodopsin-positive cells (Fig. 6A). Analysis of a single confocal plane confirmed the regular layer (in distribution and width) of cones, in contrast to the irregular layer of rods, around the organoid (Fig. 6B). The DAPI-stained nuclei clearly formed a continuous layer on the inner side of the cones, whereas the rhodopsin signal was either interspersed among the cones or extended beyond them in patches. Higher magnification showed that within these patches, the rods extended the width of the brush border (Fig. 6C). Moreover, high resolution confocal microscopy using the Airyscan module confirmed that the rod OS-like structures had a distinctly thin and elongated shape, whereas the cones were more rounded (Fig. 6D), consistent with the scanning EM images. To achieve a better depth perception without the need for clearing, we analysed the retinal organoids by biphoton microscopy and Imaris 3D reconstruction, which confirmed a well-defined cone layer with interspersed rods, and the occasional rod OS-like structure extending past the cones (Fig. 6E and F), as well as the differential shape between rods and cones.
In stark contrast to Protocol 2 organoids, low magnification confocal analysis of the surface of Protocol 3 organoids showed a regular distribution of cones, as well as intensely stained and regularly distributed rods (Fig. 6G), which was confirmed by single plane analysis (Fig. 6H). At higher magnification, both cones and rods appeared longer than those observed in Protocol 2 organoids (Fig. 6C), and the rods consistently extended beyond cones the full width of the brush border (Fig. 6I). This was further confirmed using high resolution microscopy (Fig. 6J). Biphoton microscopy and Imaris 3D reconstruction confirmed a dense layer of rods, which were evenly distributed and extended beyond the cones (Fig. 6K), and the elongated rod OS-like structures clearly contrasted with the shorter bulbous cone structures (Fig. 6L).
Taken together, organoids cultured under supplemented conditions showed a regular and stratified distribution of photoreceptors, with longer cone and rod OS-like structures, as compared to organoids cultured under non-supplemented conditions. In both cases, rod and cone photoreceptors could be distinguished by their distinct shapes.