PROM1 AS during human retinal development
AS is a widely used mechanism used by many cell types to generate protein isoforms that are necessary for their function. The photoreceptors are characterised by a specific splicing program that is in part driven by Musashi I. This displays a switch-like pattern with high exon inclusion levels in photoreceptors and almost complete exclusion outside the retina37. In addition, specialization of the photoreceptor transcriptome by Srrm3-dependent microexons is required for OSs maintenance and vision2. In our previous study, we discovered that transition from foetal to adult retina was characterised by a large increase in the percentage of mutually exclusive exons that code for proteins involved in photoreceptor maintenance. Of those, PROM1, a gene involved in photoreceptor disks morphogenesis in the retina, was shown to undergo exon 4 inclusion and exon 24/25 skipping25.
To validate the previous results and assess in detail the dynamics of PROM1 AS, RT-PCR was performed using specific primers that amplified the alternative exons of interest, exon 4 and exon 25 (Fig. 1a) followed by Sanger sequencing of PCR products (Fig. S1a, b). Both human embryonic/foetal retina (7.5–21 PCW) and adult retina samples were used. In total, 41 samples were analysed, using at least three biological replicates per developmental stage, except from 18 PCW and 21 PCW time points due to the scarcity of the tissue. This analysis demonstrated a significant increase in exon 4 inclusion from 12 PCW (Fig. 1b), which is maintained in the adult retina. For exon 25, three different PCR products were detected from 12 PCW, demonstrating skipping of exon 25 as well as 24 and 25 (Fig. 1c). Semiquantitative analysis showed that exon 4 inclusion increased to 88% at 12–18 PCW and reached 99% in the adult stage (Fig. 1d). The exon 25 skipping was more gradual compared to exon 4 inclusion, increasing to 61% at 12–18 PCW and reaching 97% in the adult retina (Fig. 1e). In contrast, exon 24 skipping was much lower than exon 25 skipping and increased from 12–18 PCW reaching its maximum at 21 PCW and then decreased at the adult stage (Fig. 1f).
In vitro ROs recapitulate PROM1 AS pattern in vivo
PSCs-derived ROs have been shown to recapitulate retinal development and were used as a tool to investigate PROM1 AS in vitro. ROs derived from both hiPSCs and hESCs were generated using a previously established protocol29 and collected at different time points of differentiation, from day 0 to 210. In total, eight time points were analysed, using three biological replicates per time point. Our RT-PCR analysis showed that exon 4 inclusion increased significantly at day 60 and 90 of differentiation in hiPSC- and hESC-ROs, respectively (Fig. 1g,i; S1c,e), reaching maximum levels at day 150 of differentiation. Exons 24 and 25 were both included in PROM1 transcripts in the undifferentiated PSCs. Exon 25 skipping started from day 35 and increased alongside the differentiation process, reaching maximal levels at day 150 of differentiation (Fig. 1h,j; S1d,f).
Exon 24 skipping increased during ROs development and remained stable from day 150 until the end of differentiation (Fig. 1k). In case of hESC-ROs, there were some variations, but the overall trend was an increase from day 0 until the end of differentiation process at day 210 (Fig. S1g). Together these data demonstrate that AS in vivo can be recapitulated in vitro, establishing ROs as a useful tool to study PROM1 exon skipping and inclusion.
PROM1 splicing pattern in the eye is cell type-specific
Previous studies have shown natural exon 4 skipping in extraocular tissues such as saliva, blood or fibroblasts26 but did not address PROM1 AS in RPE or non-retinal tissues. To gain more insight about this phenomenon, PROM1 splicing was studied in RPE, cornea, lens and vitreous. Different stages of development were assessed namely 9, 12 and 18 PCW as well as the adult stage, using three biological replicates per time point.
In the RPE, exon 4 inclusion is apparent from 12 PCW, although the PROM1 transcripts without exon 4 are more abundant (Fig. 2a). Both the cornea and lens show a remarkably similar pattern, with almost 100% of exon 4 exclusion up to 18 PCW of development and 100% inclusion at the adult stage (Fig. 2c, e). In the case of the vitreous, there is the same amount of PROM1 transcripts with or without exon 4, except at 12 PCW, where exon 4 is excluded in ca. 60% of the transcripts (Fig. 2g). Concerning exon 25, the majority of PROM1 transcripts (except the vitreous) include both exons 24 and 25 during foetal development. Exon 25 skipping increases in the adult stage, with 55% of skipping in the RPE and 100% for both cornea and lens tissues. PROM1 exon 24 skipping is low but stable throughout development and adulthood for all tissues, with the RPE and cornea displaying lower levels of skipping compared to the lens and vitreous (Fig. 2b,d,f,h,j,k).
During the PSC differentiation, we observed the presence of “eye-like” ROs, which in addition to neural retina also contained conjunctival epithelium, limbal-corneal epithelium and corneal stroma cells38. PROM1 splicing in these more complex ROs nicely recapitulated the splicing patterns observed during retinal development in vitro and in vivo, with exon 4 inclusion and exon 25 skipping increasing until day 90 of differentiation (Fig. 2l-p). These “eye-like” ROs are not evident after day 90, most likely due to culture conditions being optimised for neural retinal differentiation.
Prominin-1 is localized to the photoreceptor OSs
To determine the localization of Prominin-1, immunohistochemistry was performed in the adult human retina. The following antibodies were used to mark different parts of the photoreceptors: Translocase of Outer Mitochondrial Membrane 20 (TOMM20) to mark the ellipsoid region of the photoreceptor inner segments (ISs), peripherin-2 (PRPH2) to mark the photoreceptor OSs and pericentrin (PCN) to mark the basal body of the photoreceptors. This analysis showed that Prominin-1 is localized in the photoreceptor OSs, colocalizing with PRPH2 but not TOMM20 or PCN (Fig. 3a).
To assess Prominin-1 expression throughout RO development, immunohistochemistry and western blot analyses were performed at various time points of differentiation (Fig. 3b, c). At the early stages of differentiation (day 35 and 60), Prominin-1 is barely present (Fig. S2a), but by day 90, Prominin-1 starts to localize at the apical site of the ROs, colocalizing with Arl13b, a connecting cilia marker (Fig. S2a). From day 180, Prominin-1 colocalizes with PRPH2 in the photoreceptor OSs, but not TOMM20 (Fig. 3b) or CRALBP (Fig. S2a), which mark the ellipsoid and myoid regions of the ISs, respectively. The same pattern was observed at day 210 ROs, with restricted localization of Prominin-1 to the OSs, recapitulating the pattern observed in the adult retina (Fig. 3a).
Western blot analysis demonstrated strong expression in neural retina in adult and day 180 ROs (Fig. 3c) but not in the RPE layer. Using both an N- and C-terminal Prominin-1 antibodies, we were able to demonstrate very low expression at the early stages of differentiation (day 35 and day 60) and a distinct increase in Prominin-1 expression from day 90, which was maintained until the last time point of differentiation, day 210 (Fig. 3d), corroborating the immunohistochemistry results.
Phosphorodiamidate morpholino oligos (PMOs) screening
Two PMOs were designed to cause exon 4 skipping with the first targeting the splice acceptor site between intron 3 and exon 4 of PROM1 (PMO1), and the second targeting the splice donor site between exon 4 and intron 4 (PMO2) (Table S1). Both PMOs were evaluated in iPSC-ROs at day 90 of differentiation, alongside a standard Control-PMO (Ctrl-PMO) (Fig. S3a). Exon skipping (ES) efficiency was estimated according to the percentage of exon-skipped products, relative to total products [skipped product/ (full-length + skipped product)]. RT-PCR analysis revealed that ES efficiency was low after 48 hours at 10 µM PMO dose in all the treated groups (Fig. S3b,c). Higher concentrations of PMOs (25 µM, 50 µM and 75 µM) were investigated, showing a dose-dependent exon skipping in the case of PMO1 treatment, however no differences were observed between the PMO2 and Ctrl-PMO treated groups (Fig. S3d,e). As the ES efficiency was still not high enough, higher concentrations of PMOs were assessed (100 µM, 200 µM and 250 µM). In this case, RT-PCR analysis showed a higher proportion of exon 4 skipped product-band, suggesting higher ES efficiency, with PMO1 being more efficient than PMO2 (Fig. S3f, g). Nonetheless, a dose dependent increase in toxicity was observed from 25 µM and higher, for this reason 10 µM was chosen as the most optimal concentration to cause exon skipping without inducing toxicity in the ROs (Fig. S3h).
Effect of PMO treatment on PROM1 expression
In order to increase ES efficiency without causing toxicity, ROs were treated at day 180 of differentiation with 10 µM of Ctrl- or PROM1-PMO with either one dose or several doses from 1 to 4 weeks (Fig. S3i). RT-PCR analysis showed that PROM1-PMO treatment caused exon 4 skipping, which increased both with time and the number of doses, reaching 33% after 4 weeks with a single dose and 57% after four doses (Fig. 4a,b). qRT-PCR analysis confirmed that the level of PROM1 transcripts with exon 4 was reduced by 55% after the 4 weeks-several doses treatment (Fig. 4c).
To predict the structure of the different Prominin-1 isoforms and the impact of exon 4 skipping, AlphaFold modelling was used (Fig. 4d)31. All the splice isoforms displayed similar overall structure with 5 transmembrane helices and a large extracellular extension. The large extracellular extension, often referred to as E2 and E33, is predicted to be α-helical and protrudes 15nm into the extracellular space. Whereas exons 24 and 25 are both found on the intracellular side in the C-terminal tail, exon 4 is predicted to fold into an amphipathic helix (coloured green in Fig. 4d) and is located on the extracellular side close to the position of the first transmembrane helix. This amphipathic helix is lost when exon 4 is skipped (Fig. 4d), which may potentially alter the position of the extracellular extension or the membrane-binding properties of Prominin-1. Although, this isoform, known as s1, lacks the exon 4 which encodes the internal segment of nine amino acids near the N-terminal, its structure is stable as the nine amino acids are in frame. To further assess the impact of exon 4 skipping on expression of Prominin-1, western blotting was carried out with two different antibodies targeting the N-terminal and the C-terminal domains, respectively. As expected, Prominin-1 expression recognised by N-terminal antibody was reduced after the 4 weeks-several doses PROM1-PMO treatment compared with the Ctrl-PMO, but surprisingly this downregulation was also observed with the C-terminal antibody (Fig. 4e).
Immunofluorescence analysis was also performed to examine Prominin-1 localization. In Ctrl-PMO treated ROs, Prominin-1 was located in the OSs of the photoreceptors, whereas in the PROM1-PMO treated ROs, Prominin-1 changed its expression pattern and showed a characteristic circular pattern around the ISs, more noticeable after 4 weeks-several doses PROM1-PMO treatment (Fig. 4f). qRT-PCR demonstrated that PROM1-PMO treatment did not affect the expression of genes marking mature cones (OPN1SW, OPN1MW, OPN1LW) and rods (RHO, GNAT1) or photoreceptor OSs (PRPH2 and ABCA4). However, we noticed a significant downregulation of the cone precursor marker ARR3, the mitochondria-rich ISs gene TOMM2O and basal body marker PCN in PROM1-PMO treated ROs compared to the control ROs (Fig. 4g).
PROM1-PMO treatment results in photoreceptor cilium defects
Four weeks treatment with several doses of PROM1-PMO resulted in 57% of ES efficiency (Fig. 4a), which was considered high enough to evaluate the impact of PROM1 mis-splicing in photoreceptors. As PROM1 is a cilia-related gene, effects in cilia were studied using immunofluorescence analysis. Photoreceptor OSs formation was confirmed in both groups of treated ROs by the presence of PRPH2+ structures located apically to the ISs (Fig. 5a). In addition, the length of the segments, from the outer limiting membrane (defined by the CRALBP staining) to the distal edge of the PRPH2+ OS-like structures, was measured. No significant changes were observed until the four weeks treatment, with shorter OSs in the PROM1-PMO treated ROs observed compared to the control treated group (Fig. 5a). PROM1-PMO treatment also influenced ciliation levels, with a significantly reduced cilia incidence (measured by Arl13b staining) in the treated ROs compared to the control (Fig. 5b), although in this case, the length of the connecting cilium was similar (data not shown).
Since the connecting cilium is crucial for protein trafficking, immunofluorescence analysis was performed to assess the impact of the decreased number of ciliated cells on the distribution of several proteins associated with mature photoreceptors. Both recoverin and cone arrestin were abundant across the apical layer of the PROM1-PMO treated ROs. Immunostaining with R/G cone opsin revealed a disorganized distribution of cone photoreceptors, with abnormal accumulation in the basal side of the PROM1-PMO treated ROs (76.6% ± 1.3) compared to the control PMO treated ROs (28.9% ± 1.3). In addition, the few cone photoreceptors situated in the apical aspect of the organoid presented an abnormal bulbous cilia morphology (Fig. 5c). No obvious abnormalities were observed for Rhodopsin localization although the fraction of rods was decreased in the PROM1-PMO treated ROs compared to the control (25% ± 1.4 vs. 34.3% ± 2.6, respectively; **p < 0.005). Scanning electron microscopy demonstrated the presence of photoreceptor connecting cilium and ISs in both groups, however a higher number of OSs were captured in the control group, confirming the immunohistochemistry results (Fig. 5d).
Bulk RNA-Seq analysis suggest downregulation of autophagy in response to PROM1 downregulation
To gain insights into the consequences of exon 4 skipping in the ROs, bulk RNA-Seq analysis was performed in both control and PROM1-PMO treated ROs. A total of 53 genes were significantly differentially expressed (DE) between PROM1-PMO and Ctrl-PMO treated ROs (fold-change > 1.5, Table S1), of which 20 and 33 genes were significantly down- or up-regulated, respectively (Fig. 6a).
Metascape was used to analyse upregulated and downregulated genes separately, identifying five enriched ontology clusters for the upregulated genes, related to “cytokine signaling in immune system” and “negative regulation of autophagy”. For downregulated genes, four different enriched pathways were identified, highlighting the “formation of primary germ layer” and “regulation of secretion by cell” (Fig. 6b). These enriched ontology clusters were also apparent in the enrichment network created by mapping overexpressed and downregulated genes to KEGG pathways (Fig. 6c).
Interestingly, bulk RNA-Seq data presented an outlier (with a log2FoldChange = 2.75) named PSMC1P1, Proteasome 26S subunit ATPase 1 pseudogene 1. To confirm the upregulation of PSMC1P1, qRT-PCR analysis was performed in both groups of ROs. PSMC1P1 (pseudogene) and PSMC1 (gene) have a highly similar coding sequence, but differ in their 3’ untranslated region (3’ UTR). Specific primers targeting the identical coding sequence and the unique 3’ UTR were used in the qRT-PCR showing very low expression of the pseudogene, but much higher PSMC1 expression in the PROM1-PMO treated group (Fig. 6d). PSMC1 gene encodes one of the subunits (26S ATPase 1) of the proteasome complex, hence RNA-Seq data related to ubiquitin-proteasome system (UPS) were studied in more detail revealing a global overexpression of proteasomal subunits, including the 20S (α and β) and the 19S- subunits, as well as the proteasome assembly chaperones (Fig. 6e). These results were corroborated by proteasome activity assays, which demonstrated higher chymotrypsin-like protease activity in the PROM1-PMO treated ROs (Fig. 6f).
The UPS contributes to the degradation of proteins together with the Endoplasmic-Reticulum-Associated Degradation (ERAD) system. Notably, there was an overexpression of several key components involved in this system such as derlin3, calnexin or COPII vesicle coat proteins (Fig. 6g). As UPS and autophagy are the two key pathways in proteostasis, the autophagy regulation was further studied via western blot. PROM1-PMO treatment significantly decreased p62 and LC3-II levels in the ROs and increased phospho-S6 Ribosomal protein (p-S6) expression, leading to activation of mammalian target of rapamycin complex 1 (mTORC1) (Fig. 6h). This increased expression matches the TP53 upregulation, which regulates the mTOR complex. These findings collectively suggest a downregulation of autophagy in the PROM1-PMO treated ROs, consistent with the identification of the enriched cluster “negative regulation of autophagy” in the RNA-Seq data.
Identification of Prominin-1 binding partners using the yeast two-hybrid (Y2H) system
In order to understand the physiological role of Prominin-1 and to identify novel binding partners, a GAL4-based interaction trap screen in yeast (Y2H system) was used. Using Gateway Cloning strategy, four different plasmids, which have in common the binding domain of GAL4, were fused with a different fragment of Prominin-1: Extracellular 1 (E1), 2 (E2), and 3 (E3) and intracellular 1 (I1) (Fig. 7a). Preliminary experiments showed that Prominin-1 sub-fragments were expressed in yeast without auto-activation, proving that they were good candidates for the Y2H screen. Thus, these four constructs were used in Y2H screening against human and bovine retinal cDNA libraries, respectively. No colonies were obtained for the E1, E3 or I1 fragments neither in the human or bovine retinal libraries.
Concerning the E2 fragment, a total of 195 clones and 59 clones were isolated during screening of human or bovine retina, respectively, and were subjected to two colorimetric assays: α-galactosidase (Fig. 7b) and filter lift β-galactosidase (data not shown). Positives clones for both assays were sequenced, uncovering 12 candidate interactors in the human screening and 5 in the bovine library screening, with SH3GL3 protein present in both (Fig. 7c).
Assessment of the resulting human and bovine candidate interactors revealed four proteins that are known to be involved in protein transport, namely SH3GL3, COG6, EXOC7 and KRT18, some of which are vesicle-mediated (SH3GL3 and COG6) or related to Golgi transport (COG6 and KRT18). Other interacting proteins are microtubule-associated, such as KNSTRN, or components of microtubule-based kinesin motors (KIFC3 and KIF2A). Another centrosome-associated protein was found, BLOC1S2, a component of the BLOC-1 complex which functions in the formation of lysosome-related organelles. One transcription coactivator, FRYL, involved in cell morphogenesis and neuron projection development was also identified.
As SH3GL3 was identified in both screenings and was highly present in the human screening (29/54 sequences analysed), it was further evaluated by immunofluorescence analysis. SH3GL3 localized in the ISs of the adult retina so no colocalization was observed with Prominin-1 (Fig. 7d). SH3GL3 expression was also assessed at different time points of ROs development, showing colocalization with Prominin-1 in the periphery of the ROs at days 90 and 150 in very close proximity to the ONL at day 90 (Fig. 7e, asterisks). At the latest time points of development (day 180 and 210), SH3GL3 distribution was restricted to the ISs, colocalizing with TOMM20 (Fig. 7f). Further work is needed to validate the interactions of Prominin-1 interacting proteins at the functional level.