Aberrant retinal structure and vasculature in mouse models of dominant retinopathies caused by CRX homeodomain mutations

Abstract

CRX is a transcription factor essential for photoreceptor differentiation and functional development. Missense mutations in CRX homeodomain, CRX^E80A and CRX^K88N, are linked to early-onset dominant retinopathies. Molecular studies have revealed distinct profiles of perturbed gene expression in differentiating photoreceptors of knock-in mouse models, resulting from altered DNA binding activities of mutant CRX proteins. This study characterizes concurrent morphological alterations in knock-in mouse models. Fated cones are present in heterozygous and homozygous Crx^E80A and Crx^K88N mutants at birth, but subsequent cone differentiation is rapidly compromised. Expression of rod marker rhodopsin (RHO) is absent in Crx^K88N/N retinae but present in other mutants through adulthood. Notably, as compared to wildtype controls, RHO expression is prematurely activated in neonatal Crx^E80A mutants. Only Crx^E80A/+ retinae elaborate rod outer segments but still lose visual function by young adulthood. The presence of irregular retinal rosettes displaces the localization of inner neurons without affecting their cell numbers during retinal development. Retinal vessels develop close contact with rosette structures. In summary, disrupted photoreceptor differentiation leads to the loss of visual function and formation of retinal rosettes, secondarily impairing the localization of inner neurons and vasculature. A deeper understanding of these cellular underpinnings will inform pathogenesis of CRX homeodomain mutations.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-60825-y.

Introduction

Inherited retinal diseases (IRDs) represent a group of genetic disorders characterized by progressive degeneration of photoreceptors. IRDs are caused by pathogenic gene mutations that are inherited in autosomal dominant, autosomal recessive, or X-linked patterns. Mutations in genes coding for photoreceptor-specific transcription factors, such as the cone-rod homeobox (CRX) gene, are associated with early-onset IRDs, notably Leber congenital amaurosis (LCA) and cone-rod dystrophy (CoRD)^1,2. CRX is an essential transcription factor that regulates photoreceptor gene expression during retinal development and for functional maintenance^2,3. Substitution mutations within the coding region for CRX homeodomain are generally associated with autosomal dominant LCA and CoRD. For example, several studies have evaluated knock-in mouse models harboring three of these mutations, namely CRX^E80A, CRX^K88N, and CRX^R90W. These mutations can alter the DNA-binding capability of mutant CRX proteins, causing distinct pathogenic mechanisms^4–6. p.E80A binds to the same sites as wildtype CRX protein (p.WT) but enhances transactivation activities on targets genes such as rhodopsin (RHO) at neonatal period⁴. However, p.E80A fails to promote sufficient expression of these genes for subsequent photoreceptor differentiation, with cones being more severely affected^4,6. p.K88N shows altered DNA-binding specificity, resulting in binding at ectopic sites and perturbation of photoreceptor gene expression during early development⁴. p.R90W exhausts the DNA-binding capability and compromises transcriptional regulation for photoreceptor differentiation. Thus, CRX^E80A and CRX^K88N mutations show antimorphic effects, causing dominant CoRD in CRX^E80A/+ patients and LCA in CRX^K88N/+ patients^7,8. The hypomorph CRX^R90W mutation is associated with autosomal recessive LCA⁹. A detailed genotype-phenotype analysis is needed to determine the progression of photoreceptor degeneration and critical windows for potential therapeutic interventions.

Although molecular analysis has gained significant knowledge on how disease mutations disrupt CRX-regulatory activity, the resulting photoreceptor biology and retinal development has not been comprehensively documented in these models, particularly for the detailed course of photoreceptor defects. Defective photoreceptor localizations in many IRDs extend to secondary morphological abnormalities in inner neurons, retinal glia and vasculature^10–14. This study characterizes the spatiotemporal progression of aberrant retinal morphology in selected mouse models of CRX homeodomain mutations. A key observation is that cone differentiation is largely compromised in Crx^E80A and Crx^K88N mutant retinae at young ages. Moreover, Crx^E80A and Crx^K88N mutant retinae develop structural undulations, whorls or folds within neuronal layers, adversely impacting the cellular organization of photoreceptors as well as localizations of inner neurons and retinal vasculature. The analysis of progressive photoreceptor pathologies and secondary morphological defects in these disease models not only provides new insights into IRD pathogenesis but also serves as a paradigm for how misregulated photoreceptor gene expression can disrupt the development of retinal structure and function.

Results

Aberrant retinal morphology and development of retinal rosettes

H&E staining of retinal sections is used to evaluate retinal morphology and photoreceptor structure. In the wildtype (WT) mouse retina (Fig. 1A), H&E staining showed that retinal lamination is established upon the completion of cell proliferation around postnatal day 10 (P10)¹⁵. The elaboration of outer segments (OS) becomes discernible following eye opening around P14. Synaptic integration and phototransduction is established by approximately P21^16–18, after which the retinal morphology is preserved through adulthood (after 1 month-old (MO)). In contrast, structural alterations were found in every model of mutant retinae (Fig. 1B-E), including focal undulations (Fig. 1B), multifocal whorls (Fig. 1C,D) and afocal folds (Fig. 1E). In the following context, these structural alterations were named as retinal rosettes for simplicity. The Crx^E80A/+ retinae (i.e. E80A/+, Fig. 1B) displayed a developmental trajectory comparable to WT controls. However, the wavy architecture emerged at apical outer nuclear layer (ONL) around P14, forming definitive retinal rosettes in adulthood. The Crx^E80A/A retinae (i.e. E80A/A, Fig. 1C) developed structural turbulence at basal ONL at P10, leading to retinal rosettes with more immense undulations as compared to those in E80A/+ retinae. Similarly, retinal rosettes were detected at ONL after P10 in the Crx^K88N/+ retinae (i.e. K88N/+, Fig. 1D). Retinal rosettes were pronounced in the Crx^K88N/N retinae (i.e. K88N/N, Fig. 1E) before P10. Overall, retinal rosettes formed during photoreceptor differentiation, persisted throughout adulthood, and exhibited irregular morphologies and variable sizes. Furthermore, ONL and inner nuclear layer (INL) remained distinct and well-separated in E80A/+, E80A/A and K88N/+ retinae, while ONL and INL became assimilated in K88N/N retinae from 1 month-old (MO) onward. These morphological features collectively suggested that retinal lamination was initiated during early development in mutant retinae but extensively disrupted by the formation of retinal rosettes. In addition, the structure of OS was missing in E80A/A, K88N/+, and K88N/N retinae, implying the profound defect in photosensitivity and phototransduction.

Fig. 1.

Retinal morphology of Crx^E80A and Crx^K88N mutants. Hematoxylin and eosin (H&E) cross-section staining of (A) wildtype, (B) Crx^E80A/+, (C) Crx^E80A/A, (D) Crx^K88N/+, and (E) Crx^K88N/N retinae at postnatal day 0 (P0), P10, P14, P21, 1 month-old (MO) and 3MO. NBL: neuroblast layer. GCL: ganglion cell layer. ONL: outer nuclear layer. INL: inner nuclear layer. OS: outer segment. Scale bar = 100 μm.

Impaired photoreceptor differentiation and visual function

Photoreceptor morphology, abundance, and localization was examined to determine the condition of photoreceptor differentiation in Crx^E80A and Crx^K88N mutant retinae. Since molecular analysis suggested that cone differentiation was more severely disrupted than rod differentiation, phenotypic characterizations of cone differentiation were reported at first. Retinoid X receptor gamma (RXRG) is expressed in fated cones of neonatal mouse retinae^19,20. The immunohistochemistry (IHC) staining of RXRG detected cones at apical regions of WT and mutant retinae at P0 (Fig. 2A). Quantification revealed comparable cone numbers among all genotypes (Supplemental Fig. 1A), indicating that cone genesis was unaffected by Crx^E80A and Crx^K88N mutations. However, at P2, RXRG+ cones were scarcely detected in E80A/A, K88N/+, and K88N/N retinae, and only a few were observed in E80A/+ retinae (Fig. 2B, Supplemental Fig. 1B). At P10, when cell proliferation concludes in the retina, IHC staining of CRX was localized to the ONL and apical side of INL where bipolar cells reside in mutant retinae, a comparable pattern observed in WT controls (Supplemental Fig. 2A). IHC staining of the cone-marker arrestin 3 (i.e. cone arrestin, ARR3) was detectable only in WT and E80A/+ retinae (Fig. 2C). Cell bodies of ARR3 + cones were localized to apical ONL of WT retinae but positioned within the middle ONL of E80A/+ retinae. Similar cone phenotypes were observed in WT and mutant retinae at P14 (Fig. 2D, Supplemental Fig. 2B). Notably, E80A/+ retinae displayed a complete loss of ARR3 + cones by P21, prior to the completion of photoreceptor differentiation in WT retinae. Altogether, this data suggested that p.E80A and p.K88N do not disrupt cone fate specification or the cone precursor pool, instead impair early-stage cone differentiation. Despite their distinct DNA-binding specificities of mutant proteins, Crx^E80A and Crx^K88N mutations both lead to a rapid loss of cone populations shortly after birth in E80A/A and K88N/N mutant retinae. Even though K88N/+ retinae carried one wildtype Crx allele, the ectopic DNA binding of p.K88N impaired early-stage cone differentiation during the critical window of P0 to P2. E80A/+ mutant retinae showed a progressive loss of cones during retinal development (Supplemental Fig. 2C), supporting the molecular evidence that p.E80A fails to establish the proper chromatin landscape for sufficient expression of cone phototransduction genes for later differentiation⁶.

Fig. 2.

Defective cone differentiation in Crx^E80A and Crx^K88N mutants. Immunohistochemistry (IHC) staining of RXRG (in red) in (A) P0, (B) P2 WT and mutant retinae. IHC staining of ARR3 (in red) in (C) P10, (D) P14, (E) P21 WT and mutant retinae. Nuclei are stained by DAPI (in blue). Scale bar = 100 μm. Asterisks indicate examples of labelled cones.

In contrast to cones, diseased rods in mutant retinae underwent a different phenotypic trajectory. IHC staining of rhodopsin (RHO) was sporadically detected in WT retina but found significantly enriched in the E80A/+ and E80A/A retinae at P3 (Fig. 3A), suggesting that p.E80A caused premature rod differentiation relative to p.WT and p.K88N⁴. At P10, RHO expression was localized in ONL of WT, E80A/+, E80A/A, and K88N/+ retinae, but became absent in K88N/N retinae (Fig. 3B). At P21, RHO expression was found within OS of WT and E80A/+ retinae, while RHO expression was localized within the disorganized ONL of E80A/A and K88N/+ retinae (Fig. 3C). Furthermore, a Nrl promoter-driven cytoplasmic GFP reporter (pNrl-eGFP, a rod lineage marker) was detectable in ONL of E80A/+, E80A/A and K88N/+ retinae at P21; GFP expression was lost in K88N/N retinae, suggesting that the rod lineage commitment was disturbed with minimal activity of Nrl promoter (Fig. 3D). In summary, cone differentiation was more severely affected than rod differentiation in mutant retinae, besides K88N/N retinae showing no sign of rod differentiation. RHO expression appeared precociously in neonatal E80A/+ and E80A/A retinae; RHO expression was later trafficked to OS of E80A/+ retinae but retained in ONL of E80A/A retinae that failed to develop OS. K88N/+ retinae also had RHO expression mislocalized in ONL.

Fig. 3.

Incomplete rod differentiation in Crx^E80A and Crx^K88N mutants. IHC staining of RHO (in green) in (A) P3, (B) P10, (C) P21 WT and mutant retinae. (D) Green cells labelled by pNrl-eGFP in P21 WT and mutant retinae. Nuclei are stained by DAPI (in blue). Scale bar = 100 μm.

Apoptosis analysis with cleaved caspase 3 (CC3) staining at P2, P10, P14 and P21 showed a significant increase only in the K88N/N retina at P21, but strikingly, found non-significant difference in other comparisons (Supplemental Fig. 3A-D, F). These data suggested that initial photoreceptor loss in developing mutant retinae was implausibly driven by cell death. Apoptosis analysis at 3MO reported increased cell death in mutant retinae (Supplemental Fig. 3E, F). Notably, apoptotic cells could be occasionally found in apical ONL bearing retinal rosettes.

Electroretinography (ERG) was performed to determine if disrupted photoreceptor differentiation and retinal structure impaired visual functions in adult mice. Previous studies reported that E80A/A, K88N/+, and K88N/N mice lacked ERG responses at 1MO⁴. As expected, ERG response remained undetectable in these mutants at 3MO (Supplemental Fig. 4A-C). Rod-driven responses (dark-adapted A and B waves) were reduced but detectable in 1MO E80A/+ mice, but cone-driven responses (light-adapted B waves) were however missing⁴. Although H&E staining revealed the presence of OS structure, 3MO E80A/+ retinae showed greatly reduced to near-complete loss of rod-driven responses as compared to WT controls (Supplemental Fig. 4A, B). Cone-driven responses remained nought in 3MO E80A/+ mice (Supplemental Fig. 4C). A longitudinal comparison of rod-driven ERG responses in E80A/+ samples reported a progressive loss of rod function between 1MO and 3MO (Supplemental Fig. 4D). IHC staining was employed to determine cellular mechanisms underlying the reduced visual function in 3MO E80A/+ retinae. RHO expression was localized to OS of 1MO and 3MO E80A/+ retinae (Supplemental Fig. 5A), including within regions containing retinal rosettes. However, GNAT1, a key transducing protein of the rod phototransduction cascade^17,21,22, was detected in OS of 1MO E80A/+ retinae but significantly decreased by 3MO (Supplemental Fig. 5B). VGLUT1 expression is usually detected in the presynaptic sites^23,24. The protrusion of retinal rosettes in E80A/+ retinae impeded VGLUT1 expression (Supplemental Fig. 5C, white arrows), in contrast to adjacent regions without retinal rosettes (Supplemental Fig. 5C, yellow arrows). These data suggested that the loss of ERG responses in E80A/+ mutant retinae likely resulted from the compound defects in phototransduction and synaptic connectivity.

Quantitative PCR (qPCR) analysis provided molecular validation for cellular and functional phenotypes. At neonatal ages, all mutant genotypes showed comparable expression of Crx and Nrl transcripts to those of WT controls (Supplemental Fig. 6A), except a moderate downregulation of Nrl expression in P5 K88N/N retinae. Rxrg expression was comparable between all mutant genotypes and WT controls at P0 but became severely reduced in mutant retinae at P3 (Supplemental Fig. 6A). During photoreceptor differentiation, Rho and Gnat1 expression underwent a progressive downregulation, dropping to less than 50% of WT controls in 1MO E80A/+ retinae. This transcriptional downregulation mechanistically correlated with the reduced rod-driven ERG responses⁴(Supplemental Fig. 6B). Cone opsin expression was rapidly ablated during photoreceptor differentiation (Supplemental Fig. 6B), confirming the degeneration phenotypes and prior sequencing findings⁶.

Unaffected development of inner neurons

Since photoreceptors closely interact with inner neurons, the next question asked if impaired photoreceptor differentiation caused secondary morphological changes in inner neurons of the E80A and K88N mutant retinae. The cell abundance and localization of inner neurons was examined by IHC staining at key age points during photoreceptor differentiation. Horizontal cells were labelled by calbindin D28K staining²⁵, rod bipolar cells by protein kinase C alpha (PKCα)²⁶, amacrine cells by calretinin²⁷, retinal ganglion cells by calretinin and RNA-binding protein with multiple splicing (Rbpms)²⁸. At P10 prior to the formation of retinal rosettes, inner neurons were correctly organized in mutant retinae (Supplemental Fig. 7A-C), and their cell numbers were comparable to those in WT controls (Supplemental Fig. 7D). Occasionally, mislocalized cells were observed in K88N retinae (Supplemental Fig. 7B, stars). These data suggested that the developmental programs of inner neurons were minimally affected by impaired photoreceptor differentiation in E80A and K88N mutant retinae. At P21, the protrusion of retinal rosettes impacted the localizations of horizontal cells and rod bipolar cells in the E80A/+, E80A/A, and K88N/+ retinae, although the cell count analysis found non-significant difference between WT and mutant retinae (E80A/+, E80A/A, and K88N/+ in Fig. 4A, B, E). Interestingly, fewer labelled cells were found within regions in contact with protruded ONL (Fig. 4A, B, white arrows) as compared to adjacent regions without retinal rosettes in the E80A/A and K88N/+ retinae (Fig. 4A, B, yellow arrows). Horizontal cells and rod bipolar cells were present in K88N/N retinae, although their cell numbers were reduced as compared to those in WT controls (K88N/N in Fig. 4A, B, E). Mislocalized cells were again evident in P21 K88N/N retinae (Fig. 4A, B, stars). In contrast, the cell abundance and localization of amacrine cells and ganglions cells were unaltered in mutant retinae (Fig. 4C, D, E). In conclusion, the presence of retinal rosettes induced morphological alterations in neighboring horizontal cells and rod bipolar cells in developing E80A/+, E80A/A, and K88N/+ retinae; amacrine cells and ganglions cells were relatively distant from retinal rosettes, thus remaining unaffected. K88N/N retinae showed a reduction in cell numbers of horizontal cells and rod bipolar cells, possibly due to the worse structural disorganizations.

Fig. 4.

Inner neurons in Crx^E80A and Crx^K88N mutants. IHC staining of (A) calbindin D28K (in green), (B) PKCα (in green), (C) calretinin (in red), (D) rbpms (in red) in P21 WT and mutant retinae. White arrows indicate regions lacking labelled cells. Yellow arrows indicate clusters of labelled cells. Nuclei are stained by DAPI (in blue). Scale bar = 100 μm. (E) Cell counts for different retinal cell types between 500 μm and 1500 μm cross-sections of P21 WT and mutant retinae. Error bars represent SD on mean values (n = 4). Statistical analysis by pairwise t-test happens between WT and each mutant group. HC: horizontal cells. AC: amacrine cells. RGC: retinal ganglion cells. Asterisks (**) denote p ≤ 0.05, ns means not significant.

Moderate activation of retinal glia

Reactive gliosis can be triggered by photoreceptor injury during retinal development¹⁴. Müller glia and microglia were analyzed to assess the potential of glial activation in Crx^E80A and Crx^K88N mutant retinae at key age points during photoreceptor differentiation using immunohistochemistry. Glutamine synthetase (GS) is an enzyme specifically localized in Müller glia, and its immunoreactivity covers the radial morphology that traverses from outer limiting membrane (OLM) to inner limiting membrane²⁹. At P21, GS immunoreactivity was readily detected in mutant retinae (Fig. 5A), implying the presence of Müller glia. In particular, GS immunoreactivity in E80A/+, E80A/A, K88N/+ retinae extended apically and basally to span the retinal width, following the shape of undulations. However, GS immunoreactivity appeared misoriented and did not display the radial extension in K88N/N retinae. Cell bodies of Müller glia were labelled by P27^Kip1 staining³⁰. Müller glia were mislocalized in E80A/A and K88N/+ retinae (E80A/A and K88N/+ in Supplemental Fig. 8A), mirroring the disarray of horizontal cells and rod bipolar cells, i.e. fewer Müller glia within regions in contact with protruded ONL (Supplemental Fig. 8A, white arrows) but more at adjacent regions without retinal rosettes (Supplemental Fig. 8A, yellow arrows). Cell bodies of Müller glia appeared clustered in K88N/N retinae (K88N/N in Supplemental Fig. 8A), and cell count analysis found a significant decrease in cell numbers of Müller glia in K88N/N retinae (Supplemental Fig. 8B). In response to acute injury or stress, Müller glia become reactive and accumulate glial fibrillary acid protein (GFAP)¹⁴. At P21, GFAP expression was found at nerve fiber layer in all mutants, likely attributed to astrocytes. K88N/N retinae displayed uneven upregulation of GFAP expression, whereas GFAP expression was weakly detectable in other mutant retinae (Fig. 5B). Overall, the moderate GFAP expression indicated a limited reactive gliosis in mutant retinae.

Fig. 5.

Retinal glia in Crx^E80A and Crx^K88N mutants. IHC staining of (A) glutamine synthetase (in green), (B) GFAP (in red), (C) IBA1 (in red) in P21 WT and mutant retinae. Green cells in (C) are labelled by pNrl-eGFP. Asterisks in (C) indicate examples of IBA1 + cells. Nuclei are stained by DAPI (in blue). Scale bar = 100 μm.

Activated microglia phagocytose damaged, dying, misconnected, or mislocalized retinal neurons¹³. Ionized calcium binding adaptor molecule 1 (IBA1) functions as a microglia and macrophage-specific calcium binding protein and becomes a specific marker for microglia in the retina^31,32. pNrl-eGFP was employed to accurately identify ONL (especially for regions of retinal rosettes) from INL in WT, E80A/+, E80A/A, and K88N/+ retinae. Although the mutant retinae contained numerous mislocalized cells within retinal rosettes, microglia were detected exclusively in outer plexiform layer (OPL), INL, and inner plexiform layer (IPL) (Fig. 5C, stars). In addition, P2Y12 receptor (P2Y12R) staining confirmed that microglia were localized outside retinal rosettes (Supplemental Fig. 9A, stars). Interestingly, CD68 staining found infiltrated cells, likely macrophages, at the outer surface of retinal rosettes in E80A/A and K88N/+ retinae and within neuronal layer of the K88N/N retinae (Supplemental Fig. 9B, stars), suggesting that phagocytic macrophages may be involved in the structural formation of retinal rosettes.

Abnormal neurovascular networks

In a retina, superficial vascular plexus (SVP), intermediate vascular plexus (IVP), and deep vascular plexus (DVP) form three layers of the retinal capillary network. SVP contains a network of blood vessels located in the retinal ganglion cell layer and nerve fiber layer. IVP is located within IPL, branching out to connect SVP and DVP. DVP resides in OPL and contacts with photoreceptors. IVP and DVP develop concurrently with photoreceptor differentiation^33,34. To this end, vascular development was analyzed in Crx^E80A and Crx^K88N mutant retinae. Conjugated lectin staining revealed the vascular endothelium at key age points during photoreceptor differentiation. pNrl-eGFP reporter was employed to distinguish ONL from INL in the WT, E80A/+, E80A/A, and K88N/+ retinae. At P10, vessels at SVP and DVP were correctly localized in E80A/+, E80A/A, K88N/+ retinae, as compared to those in WT controls (Fig. 6A). In contrast, the retinal vasculature was profoundly disorganized in the K88N/N retina. Vascular elements were ectopically localized within retinal rosettes; vessels traversed the neuronal layer to reach OLM (Fig. 6A, yellow arrow). Interestingly at P21, vessels at DVP remodeled in parallel with the protrusion of retinal rosettes in the E80A/+, E80A/A, and K88N/+ retinae (Fig. 6B, white arrows). Vessels in K88N/N retinae remained in a disorganized clutter at P21, with aberrant vessels penetrating the entire neuronal layer (Fig. 6B, yellow arrow). The wholemount image at OLM identified ectopic vascular networks in K88N/N retinae (Supplemental Fig. 10, yellow arrows). The 3D reconstruction of vasculature between INL and ONL confirmed the presence of DVP and IVP in WT, E80A/+, E80A/A, and K88N/+ retinae (Fig. 7A-D). Rather than forming flat plexuses, vessels aberrantly wrapped around the protrusion of retinal rosettes in mutant retinae (Fig. 7B’-D’). Besides vessels at regions of retinal rosettes, the structural components of vasculature in E80A/+, E80A/A, and K88N/+ mutant retinae resembled those in WT retinae. However, neither DVP nor IVP could be found in the vascular landscape of K88N/N retinae. Many vessels instead grew along the apical-basal axis (Fig. 7E, E’) in K88N/N retinae. Vascular endothelial growth factor receptor 2 (VEGFR2) is known to regulate endothelial migration and proliferation^35,36. At P14, VEGFR2 expression was localized to regions with evolving retinal rosettes in E80A/A, K88N/+, and K88N/N retinae (Supplemental Fig. 11, white arrows), indicating that VEGFR2-mediated signaling may be involved in coupled pathologies of DVP disorganization and the formation of retinal rosettes. Collectively, these data support a close 3-way non-autonomous interaction between photoreceptor differentiation, the formation of retinal rosettes, and the aberrant organization of retinal vasculature in mutant retinae.

Fig. 6.

Vasculature in Crx^E80A and Crx^K88N mutants. IHC staining of conjugated lectin (in red) in (A) P10, (B) P21 WT and mutant retinae. Green cells are labelled by pNrl-eGFP. White arrows point to vessels wrapping around retinal rosettes. Yellow arrows indicate vessels running through neuronal layers of Crx^K88N/N retinae. Nuclei are stained by DAPI (in blue). Scale bar = 100 μm.

Fig. 7.

Vasculature landscape in Crx^E80A and Crx^K88N mutants. (A–E) Intermediate vascular plexus (IVP) and deep vascular plexus (DVP) in cross-sections of P21 WT and mutant retinae. ONLs in (A–D) are labelled green. Retinal rosettes are pointed in (B–D). (A’–E’) Top view oriented from INL towards ONL.

Unaffected retinal lamination and vasculature in CRX-loss-of-function mutants

The next question asked if the formation of retinal rosettes and aberrant vasculature was specific cellular signatures of the specific Crx homeodomain mutations Crx^E80A and Crx^K88N, or a common consequence of incomplete photoreceptor differentiation. A published study has shown that Crx^−/− retinae maintain the normal morphology of inner neurons at early stages of ONL degeneration³⁷. The loss-of-function CRX^R90W mutation produced a mutant protein devoid of DNA-binding activity, similar to the case of Crx^−/−4,5. Photoreceptor differentiation was absent in Crx^R90W/W (i.e. R90W/W) and Crx^−/− retinae (Supplemental Fig. 12A). These retinae did not form retinal rosettes. At P21, horizontal cells (Supplemental Fig. 12B), rod bipolar cells (Supplemental Fig. 12C), amacrine cells, retinal ganglions cells (Supplemental Fig. 12D), and Müller glia (Supplemental Fig. 12E) were present in R90W/W and Crx^−/− retinae, suggesting the development programs of these cell types were unaffected. GFAP was moderately activated in R90W/W and Crx^−/− retinae (Supplemental Fig. 12F), and some IBA+ cells were found at OPL, INL and IPL in mutant retinae (Supplemental Fig. 12G). Lectin staining detected three plexuses in R90W/W and Crx^−/− retinae (Supplemental Fig. 12H). Therefore, these loss-of-function mutations result in ablated photoreceptor differentiation but do not disrupt the developmental programs or organizations of inner neurons, glia, and retinal vasculature.

Discusion

CRX^E80A, CRX^K88N, and CRX^R90W mutations cause distinct effects on photoreceptor differentiation. In E80A/+ retinae, cone genesis appears unaffected, but the subsequent differentiation is progressively diminished. In contrast, rod differentiation initiates earlier in E80A/+ retinae, and subsequent elaboration of the OS structure develops normally. However, rods in E80A/+ retinae fail to maintain functionality in adulthood.

Cone differentiation is absent in K88N/+ retinae shortly after birth, and rod differentiation fails to develop functional rods. These phenotypes in mouse models recapitulate the clinical manifestations, i.e. dominant cone-rod dystrophy in CRX^E80A/+ patients and dominant Leber congenital amaurosis in CRX^K88N/+ patients^7,8. E80A/A retinae show signs of incomplete rod differentiation but ablated cone differentiation, while K88N/N retinae exhibit no photoreceptor differentiation. These phenotypes reveal cell-type specific effects on transcriptional regulation: p.E80A still regulates Rho expression but fails to activate cone differentiation; p.K88N represses photoreceptor differentiation in general.

Photoreceptor pathologies in Crx^E80A and Crx^K88N mutant retinae suggest important features of photoreceptor differentiation. Firstly, cone differentiation has a critical time window shortly after the event of fate choice to establish and maintain cone identity. Secondly, cone differentiation requires a fine-tuned and precise transcriptional program, for which neither hyperactive (p.E80A) nor ectopic (p.K88N) activity can be tolerated⁶. Thirdly, rod differentiation can be defined by three stages, namely, the establishment of cell identity, OS elaboration, and phototransduction. Results of RHO staining and pNrl-eGFP fluorescence suggest that rod identity can be retained even under hyperactive transcriptional activity. However, OS elaboration needs more stringent control on transcriptional activity: at least a copy of wildtype allele has to be present. Phototransduction demands an orchestra of various components in cell body and OS, which can only be maintained by sufficient activity of p.WT. Future work will be devoted to further investigation of affected components of OS, phototransduction and synaptic connectivity due to these mutations. Fourthly, photoreceptor differentiation is completely lost in K88N/N and R90W/W retinae. However, R90W/W retinae undergo a rapid ONL degeneration to a complete loss before 3MO⁵; cells survive much longer in the merged layer of K88N/N retinae. This difference in cell survival suggests that affected cells may adopt distinct cellular states or responses to stress. Future work will also examine cell death mechanisms in Crx^E80A and Crx^K88N mutant retinae. Lastly, since heterozygous mutants of each genotype show less severe phenotypes than homozygous ones, CRX augmentation may represent a potential treatment strategy³⁸. Ectopic expression of wildtype CRX may provide the essential transcriptional activity to drive photoreceptor differentiation in CRX^E80A and CRX^K88N mutant retinae. Due to altered binding activities of mutant CRX proteins, optimal efficacy may require a dual strategy of combining CRX augmentation with genetic suppression of the mutant alleles to counteract the antimorphic effects. All in all, this study provides comprehensive phenotypic evidence to established molecular mechanisms on CRX^E80A, CRX^K88N mutations^4,6 and identifies critical windows for potential therapeutic interventions.

Phenotypes of retinal rosettes in CRX^E80A and CRX^K88N mutant retinae are pronounced. In Nrl^−/− and Nr2e3^rd7/rd7 mice, retinal rosettes are formed in cone-dominant ONL or arise from disrupted rod differentiation^39,40. Interestingly, E80A/A and K88N/+ retinae develop retinal rosettes in rod-dominant ONL. Regardless of ONL cellular identity, Nrl^−/−, Nr2e3^rd7/rd7, Crx^E80A/A, Crx^K88N/+ mice develop retinal rosettes at approximately the same age. A question thus arises: why does impaired photoreceptor differentiation in transcription factor mutants lead to the formation of retinal rosettes? Although the detailed mechanisms remain to be elucidated, this study provides cellular insights to further address this question. Firstly, OS elaboration in E80A/+ retinae concurrently occurs with ONL undulations around P14, while a few cones are still detectable at this age. Hence, at least in the cases of CRX^E80A and CRX^K88N mutant retinae, a cone-dominant or rod-dominant composition is not a specific characteristic to the formation of retinal rosettes. Secondly, K88N/N retinae exhibit a complete loss of photoreceptor identity and differentiation, developing a more severe form of retinal rosettes before P10. In contrast, CRX-loss-of-function R90W/W and Crx^−/− retinae do not develop retinal rosettes, even though their ONL cells retain immature photoreceptor identity. These findings suggest that the formation of retinal rosettes cannot be explained solely by the conditions of photoreceptor identity or maturation. Therefore, the formation of retinal rosettes is likely associated with misregulated transcription by the mutant p.E80A and p.K88N. Retinal pigment epithelium cells remain as monolayers in these mutants without invading into retinal rosettes, suggesting that they are unlikely to play a role in formation of retinal rosettes. It is worth mentioning that this study does not specifically address if perturbed nuclear condensation contributes to the formation of retinal rosettes during photoreceptor differentiation. Based on H&E and DAPI staining, E80A/+ retinae show signs of nuclear condensation, yet still develop retinal undulations. It is unclear if E80A/A and K88N/+ retinae undergo any form of nuclear condensation, despite showing RHO expression within retinal rosettes. Nuclear condensation appears lost in K88N/N retinae. Future studies will comprehensively analyze the nuclear morphology of ONL neurons in these mutants during photoreceptor differentiation.

Phenotypes of inner neurons, retinal glia and vasculature are of particular interest and significance. Firstly, the localizations of Müller glia, horizontal cells and rod bipolar cells are ‘pushed’ aside by the presence of retinal rosettes. Secondly, mislocalized cells within retinal rosettes can escape microglia scavenger during photoreceptor differentiation and induce only moderate GFAP reaction. Lastly, vascular outgrowth around retinal rosettes interacts with nearby neurons via local signaling pathways in E80A/+, E80A/A, and K88N/+ retinae. The aberrant vasculature in mutant retinae likely provides nutrients and oxygen to mislocalized cells within retinal rosettes, thereby supporting their survival. Future studies will analyze the molecular mechanisms underlying the remodeling of inner neurons, retinal glia and vasculature in these models and the crosstalk between defective photoreceptors and these cells.

Vessels in K88N/N retinae become disorganized and penetrate through the neuronal layers as early as P10. A similar phenotype of vascular penetration can be seen in ONL of Vldlr^−/− retinae during adulthood⁴¹. To this end, K88N/N retinae may serve as a new model to study retinal angiomatous proliferation in absence of photoreceptor differentiation. Retinal choroidal anastomosis is commonly seen in K88N/N retinae, which requires further studies to identify the cellular underpinning.

An unanswered question remains if the rosette formation and vascular defects observed in this study are recapitulated in human pathology. It is known that IRDs, especially early-onset ones, cause progressive vision loss in conjunction with significant vascular complications^42,43. This study identifies a potential cellular link between incomplete photoreceptor differentiation and vascular defects. These findings suggest that early intervention targeting vascular abnormalities should be considered in IRD therapeutic strategies.

In summary, this study reports non-autonomous disease complications that impaired photoreceptor differentiation drives the formation of retinal rosettes, which in turn secondarily disrupts the organization of the retinal vasculature. These findings highlight the necessity of considering vascular pathologies for designing potential therapeutic strategies.

Materials and methods

All methods were performed in accordance with the relevant guidelines and regulations by Washington University in St. Louis (WUSTL) Institutional Animal Care and Use Committee (IACUC) policies.

Transgenic mouse lines

All mice in this study were on the genetic background of C57BL/6J (the Jackson Laboratory, Stock No: 000664). CRX^E80A, CRX^K88N, and CRX^R90W mutant mice, as well as pNrl-eGFP mice were generated for previously published studies^4,5,44. Both male and female mice were used in experiments.

Euthanasia of experimental mice was carried out according to approved protocol and AVMA Guidelines for the Euthanasia of Animals. In brief, introduction of CO₂ was controlled by a pressure regulator and flow meter that generated the recommended displacement rates. Narcosis was followed by a secondary method of cervical dislocation.

Retinal histology

Eyes were dissected at various ages with the superior/dorsal side marked on the cornea and fixed in 4% paraformaldehyde at 4 °C overnight. Tissues were processed with paraffin embedding. Each retinal cross-section was cut by 4 microns on a microtome. Hematoxylin and eosin (H&E) staining was performed to examine retinal histology.

Immunohistochemistry (IHC) staining

A summary of cell markers and antibodies involved in this study is detailed in Table 1. IHC staining was performed in paraffin-embedded and cryosections based on the protocol used in the previous study⁴⁵.

Table 1.

Antibody information.

Procedures of IHC staining with paraffin-embedded sections are outlined as follows. Slides were treated with hot citrate buffer for antigen retrieval and subsequently blocked with a blocking buffer of 5% donkey serum, 1% bovine serum albumin, 0.1% Triton X-100 in 1X phosphate-buffered saline (PBS) (pH 7.4) for 2 h. Slides were then incubated with primary antibodies (listed in Table 1) at 4 °C overnight. Slides were washed with 1X PBS containing 0.01% TritonX-100 (PBST) for 30 min and then incubated with specific secondary antibodies for 1 h. Slides were washed and mounted with hard set mounting medium with DAPI (Vectashield, Vector Laboratories).

Procedures of IHC staining with cryosections are outlined as follows. Dissected eyes were fixed in 4% paraformaldehyde for 1 h, followed by the removal of lens. Tissues were subsequently immersed in increasing concentrations of phosphate-buffered sucrose solutions and incubated overnight at 4 °C in 20% phosphate-buffered sucrose solution. Tissues were incubated in a 1:1 solution mixture of OCT (optimal cutting temperature; Sakura Finetek) and 20% phosphate-buffered sucrose for 1 h and then embedded in OCT and snap frozen. Cryosections were cut by 10 microns per section on a Cryotome E (Thermo Fisher Scientific). Blocking and antibody staining were performed as described above.

For vascular immunostaining and imaging of vasculature, mice were anesthetized with a cocktail of ketamine (86.9 mg/kg) and xylazine (10 mg/kg), then cardiac perfused with ice cold PBS with heparin (3U/mL) prior to surgical dissection of eyes. Whole-amount staining of conjugated lectin was performed with P21 retinae. Retinal samples were fixed overnight in 4% paraformaldehyde, then blocked and stained.

All cross-section slides were imaged taken on a Leica DB5500 microscope. All images were acquired at 1000 μm from ONH for ≥ P10 samples and at 500 μm from ONH for < P10 samples.

For 3D reconstruction of retinal vasculature, lectin-positive vasculature and pNrl-eGFP-positive photoreceptor nuclei of the retina were visualized using confocal laser scanning microscopy (Zeiss LSM 800, 20x) and Z-stacks that encompassed full depth of the vascular network (IVP-DVP) and ONL (0.5 μm slices). Vascular and photoreceptor elements captured within images were reconstructed using Imaris 9.9 software (Oxford Instruments). Following automatic surface renderings of vascular and photoreceptor elements, manual editing was performed to delete erroneous process segments.

Cell count analysis

Number of fluorescent objects were tallied in defined ranges as described in figure legends. 5 biological replicates of each genotype were used in the statistical analysis. Pairwise t-test was performed with p < 0.05, CI:95% using Graphpad Prism 10 (GraphPad Software).

Electroretinography (ERG)

Mice were dark-adapted overnight prior to ERG tests. ERG tests on 3MO mice were conducted on the same UTAS-E3000 Visual Electrodiagnostic System (LKC Technologies Inc) that was employed in the published study with 1MO mice⁴, according to the same setting and protocol used in the previous study⁴⁵. In short, for dark-adapted responses, electrode measurements were recorded with full-field light flashes (10 µs) of increasing intensity, the maximum flash intensity was 0.895 cd*s/m². For light-adapted responses, mice were exposed to 10 µs light flashes of increasing intensity, maximum flash intensity for light-adapted testing was 2.672 cd*s/m². ERG responses of biological replicates were recorded, averaged, and analyzed using GraphPad Prism 10 (GraphPad Software). The mean peak amplitudes were plotted against log values of light intensities (cd*s/m²). The statistics were obtained by using analysis of variance (ANOVA) or pairwise t-test.

Quantitative PCR (qPCR)

For P0, P3 and P5 samples, each RNA sample was extracted from 4 retinae of 2 mice using the NucleoSpin RNA Plus kit (Macherey-Nagel). For samples older than P10, each RNA sample was extracted from 2 retinae of a mouse. 2 µg of RNA was used to produce cDNA using First Strand cDNA Synthesis kit (Roche). Primers used in this study were listed in Table 2. The reaction master mix consisted of EvaGreen polymerase (Bio-Rad Laboratories), 1 µM primer mix, and diluted cDNA samples. Samples were run using a two-step 40-cycle protocol on a Bio-Rad CFX96 Thermal Cycler (Bio-Rad Laboratories). Pairwise t-test was performed with p < 0.05, CI:95% using Graphpad Prism 10 (GraphPad Software).

Table 2.

qPCR primers.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

The principal investigators, SC and RSA, acquired the funding to support the project and oversaw the progress. CS performed IHC staining, imaging, ERG, qPCR and associated analysis. CP performed cardiac perfusion and whole-amount staining of conjugated lectin, 3D reconstruction imaging analysis. YZ assisted in sample collection. CS, CP and SC prepared the manuscript. All authors edited and revised the manuscript.

Funding

The following grants were involved in this study: Knights Templar Pediatric Ophthalmology Career-Starter Research Grant (CS), NIH R01 EY012543 (SC), NIH R01 EY032136 (SC), Widłak Family CRX Research Fund (SC), Jeffrey T. Fort Innovation Fund (RSA), Siteman Retina Research Grant (RSA), NIH P30 EY002687 (WUSTL DOVS), an unrestricted grant from Research to Prevent Blindness (WUSTL DOVS).

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information.

Declarations

Competing interests

The authors declare no competing interests.