CRX haploinsufficiency delayed retinal stratification and disrupted photoreceptor pattern

To confirm whether CRX haploinsufficiency causes dominant pathogenicity, we first constructed an ESCs (H9) cell line with a CRX monoallelic deletion. A N-terminal knock-in strategy was used as previously described [36], the tdTomato fragment with polyA-tail was inserted behind CRX promoter, encoding to create a modified ESC line with monoallelic transcriptional of CRX (Fig. 1A). In the early and middle stages of differentiation, the bright field images of CRX monoallelic retinal organoids (CRX±-ROs) looked like the same as those of wildtype H9-ROs (Additional file 1: Fig. S1A). Immunostaining of early ocular development marker (RAX), retinal progenitor cell markers (SOX2, PAX6, VSX2, OTX2) showed no differences between two groups, indicating that CRX monoallelic transcription does not affect the directional differentiation from pluripotent stem cells to retina (Additional file 1: Fig. S1B). Quantitative PCR of CRX showed there was a 2-to threefold reduction in CRX±-ROs compared to control (Fig. 1B), in agreement with the monoallelic deletion of CRX.

Differences in organoid morphology do not appear until D120, when control group showed a highlighted reflective ring on the outer edge of the organoid in bright field images, which in our and others' experience indicated the outer nuclear layer (ONL) stratification. The CRX±-ROs did not show the typical ONL structure highlighted at D120 but at D225, suggesting a delay in ONL stratification. In D225 ROs of control group, the appearance of extended brushes containing the photoreceptor inner segments (ISs) and OSs indicated that the photoreceptor was gradually maturing. However, the D225 CRX±-ROs contains significantly many fewer outer/inner segments (Fig. 1C), suggesting that the ability to generate inner/outer segments was impaired.

The generation of photoreceptor opsins marks the maturation of photoreceptor cells. We examined opsin expressions in rod and cone cells in two groups (Fig. 1D–G). Rhodopsin, S-opsin and M-opsin, cell markers for rod, blue cone and green cone, respectively, can all be detected at D120 ROs in control group. In contrast, Rhodopsin and S-opsin could not be detected until D180 and D225 in CRX±-ROs, respectively, and did not increase their positive cell production even if the differentiation time is further prolonged. M-opsin+ cells cannot be detected in CRX±-ROs at all stages. Therefore, CRX haploinsufficiency is sufficient to cause destructive photoreceptor disorders in CRX±-ROs, which is similar to the severe LCA disease phenotype in human.

Altered gene expressions in CRX ± organoids at late differentiation

In order to dissect gene transcriptome changes along with photoreceptor OS/IS absence, we performed RNA-seq analysis of D120, D180 and D220 ROs (Fig. 2). We found that most of the genes downregulated in CRX± ROs of these stages were associated with Gene Ontology terms of “visual perception”, “photoreceptor outer segment”, “phototransduction”, “regulation of rhodopsin mediated signaling pathway” and other visual related processes (Fig. 2B). Opsin genes (OPN1LW, OPN1MW, OPN1MW2, OPN1SW, RHO) were significantly downregulated in all three differentiation stages, which was consistent with the results of immunostaining. The expression of guanylate cyclase activator genes (GUCA1A, GUCA1B, GUCA1C) and phosphodiesterase genes (PDE6A, PDE6C, PDE6G, PDE6H) in phototransduction, cell–cell adhesion and skeleton-related genes (RS1, IMPG1, TIMP3, SPTBN5), retinoic acid metabolism and calcium-dependent biological processes related genes (RBP3, ABCA4, RDH12, CABP4), and other phototransduction regulator genes (GRK1, PPEF2) was also decreased at D180 and D220 organoids. These downregulated genes involve in photoreceptor maturation and functional performing, partially overlapping with altered genes in Crx mutant mice (13,25).

Fig. 2

Transcriptomic analysis in wildtype and CRX±-ROs. A Heatmap comparing expression of typical genes across wildtype and CRX±-ROs at D120, D180 and D220. Values represented the averages of 3 samples in all stages of CRX±-ROs, and 2 in all stages of Control after quality assessment of RNA sequencing. B The Gene Ontology annotation analysis of differentially expressed genes (DEGs) between wildtype and CRX±-ROs

On the other hand, a small set of downregulated genes at D180 were subsequently upregulated to some extent at D220. These included the genes for catalyzing the synthesis of cyclic GMP in the photoreceptors (GUCY2D), for synaptic transmission between cones and bipolar cells (LRIT3), and for calcium sensor to light in photoreceptor (RCVRN). Interestingly, the photoreceptor transducin (GNAT1, GNGT1) and cyclic nucleotide gating ion channel (CNGA1, CNGB1), which were presented in rod cell phototransduction, were partially recovered at D220. However, GNGT2, CNGB3, ARR3, and other important genes in cone development did not recover in D220 organoids. The different performances in rod and cone maturation after CRX monoallelic deletion indicate that CRX runs different rules in two types of cells.

In addition, we found a set of genes related to cell cycle were upregulated in D180 CRX±-ROs (Additional file 2: Fig. S2). They involve in a series of biological processes accompanying mitosis, such as centromeric sister chromatid cohesion, mitotic nuclear division, spindle organization and cell cycle phase progress. We speculate that certain types of cells in organoids may go through excessive mitosis at this stage.

More Müller cells were generated at late differentiation of CRX ±-ROs

To examine the proliferative cell types in D180 organoids, we performed immunofluorescence staining of KI67, a pan-proliferation cell marker. KI67+ cells existed in both wildtype and CRX±-ROs, and most of them located in the inner nuclear layer (INL) of organoids (Fig. 3A). All KI67+ cells were also SOX9+ (a marker of retina progenitors and Müller cells at the stage). The percentage of KI67+ cells (7.279 ± 0.4827) in CRX±-ROs was significantly increased compared with that (2.673 ± 0.2089) of control (Fig. 3B), indicating an extra cell proliferation in late retinal progenitor cells or Müller precursor cells.

Fig. 3

Generation of more Müller cells in CRX±-ROs at late stage of differentiation. A Immunostaining with anti-KI67 and anti-SOX9 in D180 CRX±-ROs and Control. Scale bar, 50 μm. B Quantification of anti-KI67+ cells showing significant cell proliferation in D180 CRX±-ROs. C Immunostaining with anti-GFAP in D120, D180 and D227 CRX±-ROs and Control. Scale bar, 50 μm. D Quantification of GFAP+ area between CRX±-ROs and Control. E Quantification of average fluorescence intensity of anti-GFAP between CRX±-ROs and Control. F Immunocytochemistry with anti-CRALBP in D120, D180 and D227 CRX±-ROs and Control. Scale bar, 50 μm. G Quantification of average fluorescence intensity of anti-CRALBP between CRX±-ROs and Control

For further cell type identification, we examined glial fibrillary acidic protein (GFAP), a marker that is mainly present in the activated Müller end feet and upregulated in retinal injury. Immunofluorescence staining for GFAP showed a punctate or discontinuous fibrillar distribution in D120 organoids, with a marked increase in coarse fibrillar-like structures as differentiation continued, particularly in CRX±-ROs (Fig. 3C). The GFAP+ area increased in D227 CRX±-ROs (Fig. 3D), and the average fluorescence intensity of GFAP+ areas was significantly enhanced in D120, D180 and D227 CRX±-ROs compared with WT (Fig. 3E), suggesting an increase in activated Müller cells in CRX±-ROs. However, the cellular retinaldehyde binding protein (CRALBP), a marker widely expressed in the cytoplasm of mature Müller cells, did not enhance their average fluorescence intensity at all timepoints examined (Fig. 3F and G), implying that the number of mature Müller cells was unaffected in CRX±-ROs.

Temporary abnormal distribution of CRX+ cells during early differentiation of CRX ±-ROs

We have observed a delayed ONL stratification in CRX±-ROs (Fig. 1C). The ONL stratification is a complex process precisely orchestrated photoreceptor development, migration and synaptic growth, still with many unsolved mysteries [37]. To quantify abnormalities in ONL stratification, we measured ONL thickness in two groups. The results indicated that the ONL in CRX±-ROs was thinner than that of control at all stages (Fig. 4A and B). The CRX+ cells represent the development of postmitotic precursors into photoreceptors, which should eventually translocate into the ONL during development. A careful assessment of the positional information of the CRX+ cells is therefore necessary. For this purpose, the degree to which the fluorescent signal (tdTomato) matched the pixels of CRX was firstly checked. We verified the complete colocalization of CRX expression with tdTomato (Additional file 3: Fig. S3A), indicating that the tdTomato reporter could faithfully represent CRX+ cells. Two-photon microscopy characterized the distribution of tdTomato+ cells in the whole organoids, which showed an initial basal distribution and a gradual broadening to the apical side (Fig. 4C).

Fig. 4

Differential spatial distribution of postmitotic cells between CRX±-ROs and CRX+/+-ROs. A The ONL stratification at the indicated stages in two groups. Immunocytochemistry of Anti-RCVRN and DAPI was used to exhibit ONL formation. Scale bar, 50 μm. B Quantification of ONL thickness at the indicated stages between two group. Significance tested with t-test: ***, p < 0.001; ****, p < 0.0001. C Two-photon microscopy showing the overview of tdTomato+ cell distribution in CRX±-ROs. D Schematic diagram of the construction of CRX+/+ hESCs. E Schematic diagram of tdTomato+ cell distribution in apical, middle or basal regions. F The probability density histogram showed the overall spatial distribution pattern of tdTomato+ cells in D45, D90 and D120 CRX±-ROs and CRX+/+-ROs. Data were displayed by setting the binwidth to 0.01. Green color stands for CRX±-ROs and brown for CRX+/+-ROs. G Quantification of tdTomato+ cell percentages at the indicated stages in demarcated regions. Significance tested with t-test: *, p < 0.05

As a control, another tdTomato reporter stem cell line was constructed with an alternative knock-in strategy that keeps biallelic CRX expression as well as concurrent tdTomato expression with the P2A sequence as a spacer (Fig. 4D). In the following retinalization, the tdTomato+ cells emerged in these CRX+/+ organoids (CRX+/+-ROs) around 28 days and gradually increased with differentiation continues, same as that of CRX±-ROs (Additional file 4: Fig. S4A). We analyzed the number of tdTomato+ cells in two groups. Flow cytometric analysis showed considerable consistency in D45, D60 and D90 organoids with positive cell rates of (5.17, 13.8, 32.5) and (5.63, 12.6, 31.8) in CRX±-ROs and CRX+/+-ROs, respectively (Additional file 4: Fig. S4B and C). Colocalization analysis exhibited a well follow of CRX with tdTomato in CRX+/+-ROs (Additional file 3: Fig. S3B). The expression of PAX6, RAX, VSX2, SOX2, OTX2 at D45 exhibited no difference in wildtype H9-ROs and CRX+/+-ROs (Additional file 5: Fig. S5A). The ONL appeared as expected in CRX+/+-ROs with unaffected brush structure (Additional file 5: Fig. S5B). The results indicated that the tdTomato knock-in cell line does not affect its differentiation pattern to retinal organoids.

To quantify spatial distribution of CRX+ cell in organoids, we calculated the relative distance between fluorescent cells and apical of organoids in CRX±-ROs and CRX+/+-ROs (Fig. 4E). Probabilistic distribution analysis was performed to depict spatial distribution of tdTomato+ cells in both groups according to the distance. The probability density histogram showed that most of cells located at basal side in D45 organoids with including small number of cells adjacent to the apical in both groups (Fig. 4F). All tdTomato+ cells translocated at apical side in D120 organoids, while D90 organoids showed a transitional state with apical, basal and middle distribution in both groups. Changes in the distribution pattern of tdTomato+ cells display the location requirements of photoreceptor development during organoid differentiation.

According to the probability distribution, both groups exhibited fairly definite boundary of apical and basal. We demarcated the cells in the interval for statistics (Apical, 0–0.25; Basal, 0.75–1; Middle, 0.25–0.75). The D60 and D90 organoids showed that there is lower ratio of tdTomato+ cells at the apical and a higher ratio at the middle in CRX±-ROs compared with CRX+/+-ROs (Fig. 4G), suggesting that the CRX+ cells encountered a certain degree of translocation obstacles during early differentiation (i.e., before D90). The ratios were restored to the same level in D120 organoids. Cells distribution correction, but not the ONL formation as expected, suggests that the temporary disruption of the postmitotic cell position is attribute to a delay in ONL stratification.

Live image recording of postmitotic cell translocation in CRX ±-ROs and CRX +/+-ROs

Live-cell imaging provides us a new dimension beyond omics to trace the processes of organ development and disease occurrence. In order to investigate the translocation obstacles of CRX+ cells, we anchored D60 organoids to perform live-cell imaging. In order to minimize photobleaching and phototoxicity, a dose of laser as small as possible (0.2% of total power) was irradiated in CRX+/+-ROs and CRX±-ROs. During 28 h of continuous imaging, most of the tdTomato+ cells changed their positions in organoids (Additional file 6: Video S1; Additional file 7: Video S2). Cells translocated with elongated morphologies in a pause-motion alternate manner, which may be hindered by the narrow intercellular spaces.

To quantify the translocation, we depicted cell trajectories within the same sample thickness after image acquisition (Fig. 5A). We found that the direction of cell translocation is not uniform (Fig. 5B). In addition to some cells that did not move effectively, the other cells either translocated along apical-basal axis or perpendicular to the axis. In CRX+/+-ROs, about 60% of the cells translocated toward the apical side, while this proportion was only about 20% in CRX±-ROs (Fig. 5C). On the contrary, cells toward basal side and tangential direction were significantly increased in CRX±-ROs. The number of cells with invalid paths did not show a significant difference between two groups. Selected typical cell trajectories showed different translocation characteristics between two groups (Fig. 5D).

Fig. 5

Live-cell imaging of D60 CRX±-ROs and CRX+/+-ROs. A Overview of cell trajectories in the imaging area of CRX+/+-ROs. The color represents cell displacement over time with the live-cell imaging. B Representative images of basal- and apical-translocating cells. Scale bar, 5 μm. C Quantification of basal-, apical-, tangential-orientated and hesitant cells in CRX±-ROs and CRX+/+-ROs. Significance tested with t-test: *, p < 0.05; ns, not significant. D Representative cell trajectories in CRX±-ROs and CRX+/+-ROs. E Groups of paths were summed according to the total length of movement. F Groups of paths were summed according to distance between start and end positions. Significance tested with paired t-test: *, p < 0.05; **, p < 0.01. G Quantification of straightness between two groups. Significance tested with t-test: **, p < 0.01

To compare the translocation parameters of tdTomato+ cells, we grouped all cells into 9 sets according to time-length of each cell appeared in images for further differential analysis. Distance is used to represent the effect translocation between the start and end positions, while Length is used to sum all paths traveled by a cell (Fig. 5E and F). Both Distance and Length in the CRX±-ROs were shorter than those in CRX+/+-ROs, showing a relatively inefficient translocation pattern. We further analyzed the straightness of cell paths, which exhibiting significantly lower in CRX±-ROs than in CRX+/+-ROs. To sum up, CRX+ cells showed an unclear direction and relatively inefficient translocation due to CRX monoallelic transcription.

Overtension of actomyosin network restricted postmitotic translocation in CRX ±-ROs

To further investigate changes in underlying molecules and pathways to restrict translocation of CRX+ cells, we performed RNA-seq analysis of D60 organoids in CRX±-ROs and wildtype group. When setting significance threshold corresponding to an adjusted p value < 0.05, we obtained 118 upregulated and 58 downregulated differentially expressed genes (DEGs) in CRX±-ROs. Among which, we focused on genes with larger foldchanges, most significant differences, with higher expression levels (Fig. 6A). Genes of extracellular matrix (ELN, COL8A2, COL22A1, TENM3 and FBLN2), signal molecules and transcriptional regulators related with retina development (NOG, BMP2, PRDM16, PAX2, BMPR1B and SMOC1), cell adhesion and cytoskeleton (AHNAK, MYBPC1, PDPN, DOCK6 and TENM1) were significantly upregulated. In contrast, genes related to visual development (RCVRN, RD3, RS1, PDE6H, ARR3, GUCA1A, IMPG2, GNGT2 and MPP4), multifunction cellular process (ZNF717, HSPA6) and synaptic development and transmission (EGFLAM, CPLX4) were significantly downregulated.

Fig. 6

Overtension of actomyosin network in D60 CRX±-ROs and candidate regulators. A Differentially expressed genes (DEGs) between the CRX±-ROs and the wildtype group. The red plots represent the upregulated genes and the blue plots represent the downregulated genes. B Gene enrichment annotation analysis of DEGs and the Top 4 terms in up- and downregulation, respectively. C Protein–protein interaction analysis by DEGs and the Top 8 clusters. D Heatmaps depict Z-scores of FPKM of genes related to actomyosin contraction and extracellular matrix. E Immunostaining of pMLC2 in CRX+/+-, CRX±-ROs and wildtype groups. Scale bar, 50 μm. F Immunostaining of pMLC2 in CRX±-ROs with blebbistatin. BI: blebbistatin. Scale bar, 50 μm

To evaluate biological processes more systematically, we performed enrichment analyses on DEGs. For this purpose, we set significance threshold corresponding to a p value < 0.1, resulting in an enlarged set of DEGs with 316 upregulated and 134 downregulated. The top 4 upregulated genesets were associated with “striated muscle contraction pathway”, “regionalization”, “muscle structure development”, and “extracellular matrix organization”. The top 4 downregulated genesets were involved with “visual perception”, “NABA core matrisome”, “peptide cross-linking via chondroitin 4-sulfate glycosaminoglycan”, and “PID cone pathway” (Fig. 6B). Meanwhile, the simulated molecular complexes in protein–protein interaction analysis showed that there were 4 clusters contributing to muscle contraction pathway and extracellular matrix regulation in the top 8 dense regions (Fig. 6C). Previous studies have shown that actomyosin contractility and extracellular matrix organization affect multiple dynamics, such as migration, cell shape, cell division and tissue morphogenesis. Therefore, it is not surprising that actomyosin contraction and extracellular matrix-related genesets were present in downstream of CRX (Fig. 6D), suggesting that they became potential regulators involved in the abnormal translocation of photoreceptor precursors in CRX+/+-ROs. The data probably indicated a differentially activated actomyosin network between two groups.

Actin interacts with molecular motor myosin and crosslinkers to generate contractility. This activated process is energized by ATPase hydrolysis and characterized by myosin phosphorylation. To verify whether there was an overactivated actomyosin network, we examined the levels of phosphorylated myosin light chain 2 (pMLC2), a local indicator of actomyosin activation, in wildtype, CRX±- and CRX+/+-ROs (Fig. 6E). The CRX±-ROs had a significantly stronger accumulation of pMLC2 especially near the spherical surface than the other two groups.

Blebbistatin, a selective small molecule inhibitor of motor myosin, can slow down phosphate release in ATPase catalysis. We found that blebbistatin attenuated the accumulation of pMLC2 (low concentration, ≤ 2 μM) on the spherical surface of CRX±-ROs in a dose-dependent manner (Fig. 6F), confirming the presence of an overactivated actomyosin network in CRX±-ROs. The stronger contractility suggested that the overtension network in CRX±-ROs impeded the translocation of photoreceptor precursor cells.