Abstract
Polyglutamine (polyQ) expansion in Ataxin-7 (ATXN7) results in spinocerebellar ataxia type 7 (SCA7) and causes visual impairment. SCA7 photoreceptors progressively lose their outer segments (OSs), a structure essential for their visual function. ATXN7 is a subunit of the transcriptional coactivator Spt-Ada-Gcn5 Acetyltransferase complex, implicated in the development of the visual system in flies. To determine the function of ATXN7 in the vertebrate eye, we have inactivated ATXN7 in zebrafish. While ATXN7 depletion in flies led to gross retinal degeneration, in zebrafish, it primarily results in ocular coloboma, a structural malformation responsible for pediatric visual impairment in humans. ATXN7 inactivation leads to elevated Hedgehog signaling in the forebrain, causing an alteration of proximo-distal patterning of the optic vesicle during early eye development and coloboma. At later developmental stages, malformations of photoreceptors due to incomplete formation of their OSs are observed and correlate with altered expression of crx, a key transcription factor involved in the formation of photoreceptor OS. Therefore, we propose that a primary toxic effect of polyQ expansion is the alteration of ATXN7 function in the daily renewal of OS in SCA7. Together, our data indicate that ATXN7 plays an essential role in vertebrate eye morphogenesis and photoreceptor differentiation, and its loss of function may contribute to the development of human coloboma.
Introduction
Ataxin-7 (ATXN7) is a ubiquitously expressed protein harboring a polyglutamine (polyQ) tract in its N-terminus. Expansion of polyQ in ATXN7 results in adult-onset spinocerebellar ataxia type 7 (SCA7), which causes progressive neuronal loss in the cerebellum and associated structures, leading to cerebellar ataxia, dysarthria and dysphagia (1–4). In addition, SCA7 causes the loss of visual acuity in 83% of patients, primary due to cone–rod photoreceptor degeneration (5–10). Bipolar and retinal ganglion neurons are also affected, and damages of the Bruch’s membrane, retinal pigmentary epithelium (RPE) and optic nerve were reported (5,6,11,12). In SCA7 mouse models, photoreceptors show an atypical degenerative scheme characterized by a progressive disappearance of their outer segments (OSs), which correlates with the decreased expression of photoreceptor-specific genes (13–18). This suggests that the mutant ATXN7 primarily alters the differentiation of photoreceptors (15,17,19).
ATXN7 is a known member of the multiprotein Spt-Ada-Gcn5 Acetyltransferase (SAGA) complex, a highly conserved coactivator of transcription, which harbors chromatin-modifying activities (20,21). SAGA is organized into several functional submodules, including a histone acetyltransferase (HAT) module that contains either KAT2A or KAT2B acetyltransferases and a deubiquitination (DUB) module that contains USP22 ubiquitin protease, ENY2, ATXN7L3, as well as ATXN7 (22). ATXN7 was shown to anchor the DUB module within the core SAGA and may therefore coordinate the enzymatic activities of the complex (23).
SAGA functions as gene-specific and general cofactor for RNAPII transcription in yeast (24,25). In metazoans, the expression level of SAGA subunits appears to be critical for diverse neural development processes (22). Interestingly, subunits of the DUB module play a role in the proper development of the visual system in flies. Drosophila mutants for nonstop or sgf11 (the ortholog of USP22 and ATXN7L3, respectively) cause a similar loss of glial cells in the lamina plexus of the optic lobe, resulting in misprojection of photoreceptor axons into the medulla (26–28). While flies with loss of atxn7 ortholog die at prepupation, targeted RNAi knockdown of atxn7 expression within the retina and lamina leads to a progressive age-dependent retinal degeneration (29).
The retina of fly and vertebrate show marked anatomical and physiological differences. In flies, photoreceptors are the only retina-specific cell type and they directly project to interneurons in the optic lobes of the brain. In vertebrate, five different retinal neurons are present and photoreceptors indirectly transfer visual information to the brain via retinal interneurons and ganglion cells, the latter projecting to the optic lobes. To concentrate light absorption, photoreceptors use different strategies; Drosophila use actin-rich rhadomeric membranes, while vertebrates use microtubule-based ciliary OS. An initial report showed that moderate depletion of the atxn7 ortholog of zebrafish leads to reduced expression of rhodopsin gene (30), suggesting an altered differentiation of rod photoreceptors as in SCA7 disease.
Here, we show that inactivation of atxn7 in zebrafish—through antisense oligonucleotides or CRISPR/Cas9 approaches—primarily results in a characteristic ocular malformation named coloboma. Coloboma occurs when the choroid fissure fails to close, leaving a cleft in the inferonasal quadrant of the eye. In conjunction with coloboma, atxn7-deficient zebrafish showed interruption of the RPE monolayer and alteration of the optic nerve. Our data suggest that elevated Hedgehog (Hh) signaling is a major contributor to coloboma and other eye malformations in atxn7-deficient fish. Finally, at later stages of retinal development, photoreceptors show a defect in the morphogenesis of OSs, which explains the altered expression of rhodopsin seen in earlier studies (30). Consistently, atxn7-deficient embryos display strong decreased expression of crx, a key transcription factor involved in photoreceptor terminal differentiation and in the expression of rhodopsin (31). Taken together, our results indicate that Atxn7 plays an essential role in vertebrate eye morphogenesis and terminal differentiation of photoreceptors.
Results
High expression of atxn7 transcript in zebrafish eye and brain
The atxn7 zebrafish gene encodes a protein that shares 51.74% amino acids identity with the human one, including a polyQ stretch and conserved domains (Fig. 1A). Using reverse-transcription polymerase chain reaction (RT-PCR), the temporal expression profile indicates that zebrafish atxn7 mRNA is expressed throughout the development (Fig. 1B). atxn7 expression was detected from the cleavage period as early as 1 h post-fertilization (hpf), also known as 4-cell stage, and during the blastula period represented by the sphere stage, showing that atxn7 mRNA is maternally provided. The expression continued from the early to the late segmentation period (12 and 24 hpf, respectively), throughout the hatching period (48–72 hpf) and up until the early larval stage of 5 days post-fertilization (dpf).
Expression pattern of atxn7 transcript in zebrafish embryo development. (A) Modular structure of human ATXN7 and zebrafish Atxn7 proteins. The polyQ stretch is conserved in zebrafish and composed of five residues. It is polymorphic from 4 to 35 residues in humans. The C2H2 zinc-finger domain (ZnF Domain), the atypical Cys-X9–10–Cys-X5–Cys-X2-His motif, known as SCA7 domain, and a conserved C-terminal domain are also found in the zebrafish Atxn7. (B) Temporal expression of atxn7 mRNA in developing zebrafish using semi-quantitative RT-PCR. Beta-actin mRNA is used as quantity control. hpf and dpf, hours or days post-fertilization. (C–E’) atxn7 mRNA expression pattern in wild-type embryos using WISH of atxn7-specific antisense probe. Typical lateral views of embryos (C–E) and the corresponding dorsal views (C’–E’) of the head region. During early somitogenesis stage (C, C’), the atxn7 transcript is detected ubiquitously in the embryo with high intensity in the eye area (e). This expression is maintained through development as seen at 24 hpf (D, D’) and 48 hpf (E, E’; arrow indicates the absence of atxn7 in the fin bud). e, eye; L, lens. (F–F’) Dissected eye from a 72 hpf embryo with enlarged lateral view of photoreceptor layer (F’). Expression of atxn7 transcript in the photoreceptors is shown by the presence of labeling in the ONL.
To determine the spatial expression of atxn7, whole-mount in situ hybridization (WISH) was performed on selected stages ranging from 1 hpf to 3 dpf with a digoxigenin (DIG)-labeled antisense RNA probe specific for atxn7 (Fig. 1C–F’ and Supplementary Material, Fig. S1). The spatial expression of atxn7 appears to be dynamic during embryonic development. Consistent with the RT-PCR results, atxn7 is detected during the zygotic period as early as 4-cell stage and is ubiquitously distributed in the blastula (Supplementary Material, Fig. S1A and B). During the segmentation period (18 hpf), zygotic atxn7 mRNA is ubiquitously expressed with enrichment in the developing brain and eyes (Fig. 1C–C’ and Supplementary Material, Fig. S1C). This pattern is maintained at 24 hpf (Fig. 1D–D’ and Supplementary Material, Fig. S1D), and by 48 hpf, the expression becomes predominantly anterior, in the brain and eyes, but excludes the lens and the fin buds (Fig. 1E–E’ and Supplementary Material, Fig. S1E). Finally, by 72 hpf, expression is further restricted to the anterior portion of the larvae (Supplementary Material, Fig. S1F). We have observed an expression in photoreceptors of dissected eyes, as shown by staining in the outer nuclear layer (ONL) (Fig. 1F–F’). WISH, using a DIG-labeled sense RNA probe specific for atxn7, was used as a staining control (Supplementary Material, Fig. S1G).
Together, these results demonstrate that atxn7 is both maternally and zygotically expressed, has a highly dynamic expression pattern and becomes predominantly enriched in the brain and eyes during zebrafish embryogenesis.
Altered expression of atxn7 causes ocular coloboma
Previous studies in flies and zebrafish have suggested that depletion of atxn7 can result in photoreceptor alterations (29,30). However, atxn7 transcript showed a predominant expression in the ocular area already at 18 hpf, suggesting broader functions during eye development. In order to address this point, a morpholino antisense oligonucleotide (Mo) targeting the translation start site of atxn7 mRNA (named Mo1) was designed to knock down atxn7 expression in zebrafish embryos (Supplementary Material, Fig. S2A).
Since there is no known mutant to phenocopy, concentration-effect experiments were carried out to establish a working concentration of Mo1 with minimal toxicity levels. One- to 2-cell stage embryos injected with 3 ng or less of Mo1 always had very low mortality at 24 hpf and normal anteroposterior body axis at 72 hpf, as non-injected or embryos injected with control morpholino (6 ng) (Supplementary Material, Fig. S2B–D). Higher amounts of Mo1 (6 and 9 ng) induced mortality in more than 10% of embryos at 24 hpf, and at 72 hpf, survivors often displayed severe body anomalies with gnarled tails or shortening of the body axis, which could be the result of unspecific toxicity (Supplementary Material, Fig. S2E–F). Therefore, for the rest of the study, wild-type zebrafish embryos of the strain AB were injected with 3 ng of translation blocking Mo1 per embryo, hereafter referred as Mo1 morphants, and analyzed for eye phenotype.
Mo1 morphants displayed visible morphological anomalies of the eye cup at 24 hpf (Fig. 2A–A’). The formation of the eye cup involves key morphogenetic events. In particular, in the ventral area of the developing eye, the two edges of neuroectodermal layers meet at 24 hpf to form a narrow choroid fissure, a transient opening of the eye cup. Soon after, the choroid fissure closes to circumscribe the neural retina and the RPE in the eye cup (illustrated in Fig. 2B) (32,33). Compared to control embryos at 24 hpf, Mo1 morphant eyes presented larger opening of the choroid fissure. The extent of opening varied in severity, and in some cases, the neuroectodermal lip was bent backwards (red arrow in Fig. 2A’). To ensure that this larger fissure was not due to a developmental delay, embryos were followed throughout development. By 48 and 72 hpf, ocular defects in Mo1 morphants were more pronounced, and the choroid fissure clearly failed to close, leaving a cleft in the inferonasal quadrant of the eye (compare Fig. 2C–C’ for 48 hpf–control embryos and Fig. E–E’ 48 hpf–Mo1 morphant, and data not shown). This morphological anomaly is known as ocular coloboma. In addition, when the Mo1 morphants were imaged laterally, a lack of pigmentation was observed in the ventral area surrounding the choroid fissure (red arrowhead in Fig. 2E). Normal pigmentation of the eye comes from the melanosomes of the underlying RPE. The defect becomes striking at 5 dpf, when Mo1 morphants almost fully lack coloration through the lens, in the area corresponding to the RPE of the posterior retina (Fig. 2F–F’), whereas 5 dpf controls show a uniform black coloration of the lens (Fig. 2D–D’). Notably, when the same 5 dpf–control embryos and Mo1 morphants were visualized dorsally (Fig. 2D” and F”), the posterior eye cup of Mo1 morphants had irregular edges, incomplete pigmentation and extrusion of non-pigmented tissues toward the forebrain (yellow arrow in Fig. 2F”). To further characterize these defects, histological sections were examined at 5 dpf. While in control eye, a continuous RPE monolayer separates the posterior eye cup from the brain, the RPE of Mo1 morphant eyes was normal in the anterior part but disrupted in the posterior part, resulting in the extrusion of colobomatous neural retina into the brain area (Fig. 3A and B). This likely explains the lack of pigmentation at the posterior part of morphant eyes and an apparent ‘hole’, when visualized at macroscopic level (Fig. 2F). Tissue extrusion in Mo1 morphants showed different degrees of severity that correlate with the different levels of eye alterations detected by macroscopic observations (Supplementary Material, Fig. S3).
Morpholino-mediated knockdown of atxn7 causes coloboma. (A–A’) Representative light microscopy images of control and Mo1 morphant eyes at 24 hpf. Mo1 morphants display large opening of the choroid fissure with different degree of severity. Line shows the gap between the two edges of the neuroectoderm. Red arrow shows areas where one of the neuroectodermal lips of the eye is bent backwards. L, lens. (B) Diagram representing closure of the choroid fissure. Dotted square highlights the ventral area of the developing eye, where the two edges of neuroectodermal layers must get in contact and fuse in order to circumscribe the retina and the RPE in the eye cup, as shown in the control panels. Failure of choroid fissure closure results in a coloboma, as shown in the Mo1 panel. (C–H”) Representative light microscopy images of control embryos (C–D”), Mo1 (E–F”) and Mo2 (G–H”) morphants. Respectively at 48 hpf, lateral views of the eyes (C, E, G) and the corresponding lateral views of whole bodies of the same embryos (C’, E’, G’); at 5 dpf, lateral views of the eyes (D, F, H) and the corresponding lateral views of whole bodies (D’, F’, H’) and dorsal view of heads (D”, F”, H”) of the same larvae. Several pigmentation defects are visible on lateral and dorsal views of morphant eyes. On 48 hpf–eye lateral view, red arrowhead indicates the lack of pigmentation in the area of choroid fissure and developing lens in Mo1 and Mo2 (E and G, respectively) compared to the control (C). On 5 dpf–eye lateral view, a pale coloration through the lens in Mo1 and Mo2 (F and H, respectively) compared to the black coloration in control (D). On the corresponding 5 dpf–eye dorsal view, Mo1 and Mo2 (F” and H”, respectively) have irregular eye edge (red line) and non-pigmented tissue extruding into the area of the forebrain (yellow line and arrow) in comparison with the control (D”).
Interruption of RPE monolayer and retina extrusion into the brain. (A, B) Toluidine blue-stained retinal coronal sections of 5 dpf–control (A) and Mo1 morphant (B) larvae. Histological section of Mo1 retina showing ectopic retinal tissue extruded into the brain area; grey arrow heads indicate interruption of the RPE. Scale bar 50 μm. L, lens.
To confirm the specificity of the Mo1 morphant coloboma phenotype, a second non-overlapping atxn7 morpholino, Mo2, was designed to target the splice donor site of the atxn7 exon 4 (Supplementary Material, Fig. S4A). Similar to Mo1 morphant, embryos injected with Mo2 (1 ng) presented alterations of choroid fissure closure and ventral eye pigmentation at 48 hpf (Fig. 2G), lack of coloration through the lens at 5 dpf (Fig. 2H), as well as irregular edge of the posterior eye cup and extrusion of non-pigmented tissues toward the forebrain (Fig. 2H”). Similarly to Mo1, the frequency of specific eye anomalies increased with the concentration of Mo2 (Supplementary Material, Fig. S4E and F).
Inhibition of splicing by Mo2 should result in exon 4 skipping or intron 4 retention in atxn7 mRNA and create frameshift, leading a premature stop codon (Supplementary Material, Fig. S4A and B). To confirm the efficacy of Mo2 to block intron 4 splicing, RT-PCR was performed with specific primers encompassing atxn7 exons 3 to 5, using 24 hpf embryos injected with an increasing concentration of Mo2. The efficacy of Mo2 was demonstrated by the concentration-dependent decrease of the correctly spliced intron 4 PCR product (471 bp) and increase of the non-spliced product (590 bp) (Supplementary Material, Fig. S4C). Sequence analysis of the 590 bp PCR product confirmed the specificity and efficiency of the Mo2 splice-blocker with intron 4 retention and the presence of a premature stop codon (Supplementary Material, Fig. S4D). RT-PCR analysis also revealed an additional 302 bp product in Mo2-treated samples. Sequence analysis indicated that it originates from the use of a cryptic 5′ donor splice site located in exon 4. This aberrant splicing creates a reading frameshift and a premature stop codon. It is noteworthy that this 302 bp band was never present at any time points through development in control embryos, suggesting that its appearance is a direct consequence of Mo2 injection.
To perform quantitative analysis of the variable penetrance and severity of coloboma and associated pigmentation phenotypes, a four-point grading scale was defined at 72 hpf as follows (Fig. 4A–E): ‘No Phenotype’ corresponds to control eyes where the two edges of the choroid fissure fuse normally, the lens shows uniformly black coloration in lateral view and the posterior eye cup is regular and shows no tissue extrusion into the brain area in dorsal view; ‘Unilateral’ and ‘Bilateral’ phenotypes define severe anomalies, respectively, in one eye or both eyes, characterized by the combined presence of coloboma, lens with pale or no coloration and visible tissue extrusion into the brain; ‘Mild’ are ranging between normal and severe defects. Non-injected and control fish behaved identically with 96.9% ‘No Phenotype’ (Fig. 4A and B). In contrast, 89.3% of Mo1 morphant showed eye anomalies, with 67.8% of ‘Bilateral’ severe phenotype, 14.2 % of ‘Unilateral’ and 7.1% ‘Mild’ (Fig. 4A, C–E). When this point grading scale was used to score embryos in the concentration-effect curve of Mo1, the penetrance and severity of the phenotype increased in a concentration-dependent manner (Supplementary Material, Fig. S2G).
Penetrance and severity of eye phenotype in atxn7-deficient zebrafish. (A) Bar plot showing the frequency and severity of eye anomalies induced by atxn7 gene knockdown. Embryos were injected with 3 ng of Mo1 (n = 112 embryos) or 6 ng of control morpholino (n = 133). The severity of the eye phenotype was scored at 72 hpf, according to categories defined in the Results Section: ‘No Phenotype’, ‘Mild’, ‘Unilateral severe’ and ‘Bilateral severe’. (B–E) Representative images in dorsal view of the eye phenotypes at 72 hpf in Mo1 morphants. (B) ‘No Phenotype’ being the control phenotype where no tissue extruded into the brain area. Categories ‘Mild’, ‘Unilateral’ and ‘Bilateral’ indicate increasing severity of eye phenotype with, respectively, little but present defect (C), severe defect in one eye (D) or both eyes (E) where the posterior eye cup had irregular edge, incomplete pigmentation and extrusion of non-pigmented tissues toward the forebrain. Red arrow indicates affected areas. (F) Bar plot of the four-point grading scale used to quantify the variably penetrant coloboma phenotype induced by atxn7 knockdown (Mo1) compared to Mo1 morphants that also received 150 μg human ATXN7 mRNA (Mo1-N10). With expression of N10, there were significantly fewer 72 hpf larvae with a ‘Bilateral’ phenotype and more larvae with ‘No Phenotype’. N = 100/condition (Chi square P < 0.0001).
Rescue experiments were performed to further support that the coloboma phenotype was caused by the downregulation of atxn7. Human mRNA expressing ATXN7 with 10 glutamines (N10) was injected into Mo1 morphants, and the four-point grading scale was used to quantify the effect on coloboma severity and penetrance. As shown in Figure 4F, injection of ATXN7 N10 mRNA (Mo1-N10) significantly reduced the frequency or the severity of the coloboma phenotype, as compared to embryos injected with Mo1 alone (Chi square P < 0.0001). There was a striking 2.5-fold increase in the proportion of larvae with ‘No Phenotype’ when embryos were coinjected with N10 (Mo1-N10) and a 2-fold decrease of larvae that displayed the most severe phenotype (‘Bilateral’ phenotype). These data indicate that human ATXN7 can compensate for the loss of zebrafish atxn7.
In order to recapitulate the coloboma defect using a different genetic approach, the CRISPR-Cas9 system was utilized as previously reported (34–36). A single guide RNA (sgRNA) was designed to target the coding region in the first exon of atxn7 gene to increase the likelihood of altering the protein reading frame (Fig. 5A). The gRNA was co-injected with Cas9 protein from Streptococcus pyogenes into fertilized one-cell stage embryos. Injected embryos were analyzed at 24 hpf for genome modifications of the target locus or at 24 and 48 hpf for phenotypic analysis. Three controls were used for analysis: non-injected, injected with sgRNA but no Cas9 or injected with Cas9 alone. Using PCR amplification of the target genomic region and heteroduplex mobility assay, 68.62% (35/51) of embryos injected with sgRNA and Cas9 presented indels, while none of the control embryos (n = 15, with five embryos for each of the three control conditions) showed a similar multiple band pattern (Supplementary Material, Fig. S5A). For three randomly selected embryos, the PCR products were subcloned and multiple clones were sequenced. All the 53 sequences confirmed the presence of indels, and all but one mutation led to the introduction of premature stop codon in the atxn7 sequence (Supplementary Material, Fig. S5B–D). Sequence analysis of the control samples did not show any mutation. Once the efficiency of CRISPR/Cas9 system was established, embryos of the strain AB were injected and the resulting F0 animals were scored for ocular phenotypes by light microscopy. At 24 hpf, sgRNA-injected embryos presented first signs of eye defects. By 48 hpf, 24.48 % (12/49) of the sgRNA-injected embryos displayed a clear coloboma, while none of the control larvae (n = 15, with five embryos per condition) showed eye anomalies (Fig. 5B). Taken together, the consistent phenotype observed in embryos injected with either atxn7gRNA + Cas9, Mo1 or Mo2 morpholinos and the phenotypic rescue by human ATXN7 strongly support that alteration of atxn7 expression during zebrafish development leads to coloboma.
Crispr/Cas9-mediated inactivation of atxn7 causes coloboma. (A) Schematic representation of the genomic structure of the atxn7 gene exon 1. The 19 bp target sequence of sgRNA is indicated in the green box, followed by the protospacer adjacent motif (PAM GGG) in the light blue box. Shown at the bottom are the primers used to amplify the genomic region for heteroduplex mobility assay and sequence analyses. (B) Representative lateral views of eye and body of the same embryos at 48 hpf. On the right panels, red arrow indicates the persistence of choroid fissure in the atxn7gRNA+Cas9 P0 founder embryo, which displayed a coloboma defect. On the left panels, none of the control embryos (n = 5 for each of the three control conditions) presented coloboma.
Previous reports identified four ATXN7 paralogs in vertebrates that share distinctive conserved domains (20). The Ensembl zebrafish (Zv9) database identifies single orthologues for human ATXN7, ATXN7L1 and ATXN7L3 genes (atxn7, atxn7l1, atxn7l3, respectively) and duplicated orthologues for ATXN7L2 in the zebrafish genome (atxn7l2a and atxn7l2b) (Supplementary Material, Fig. S6A). ATXN7L3 protein is a subunit of the DUB module and co-exists with ATXN7 in the SAGA complex, while ATXN7L1 and L2 are thought to be commutable with ATXN7 in SAGA. In order to probe if downregulation of atxn7 induced the overexpression of any of the paralogs, which could compensate and modulate the atxn7 morphant phenotypes, semi quantitative RT-PCR was performed on control and Mo1 morphant samples. All paralogs are expressed as atxn7 at 24 and 48 hpf, except for atxn7L1, which showed very low expression at 24 hpf. The results show that there was no significant change in expression of paralogs in Mo1 morphants compared to control embryos (Supplementary Material, Fig. S6B).
Atxn7-deficient zebrafish shows coloboma but not microphthalmia nor anophthalmia
Coloboma is a human ocular malformation, which can be uni- or bilateral and can occur in combination with microphthalmia and anophthalmia (37). In humans, microphthalmia is diagnosed when the eye shows an axial length inferior to two standard deviations compared to the mean of normal individuals of the same age (38). To assess whether atxn7 morphants may have microphthalmic eye in addition to coloboma, larvae were transiently anesthetized and imaged, at three different days through development. Eye diameter (naso-temporal axis) and total body length (anteroposterior body axis) were measured on images for Mo1 morphants, control morpholino- and non-injected embryos (Supplementary Material, Fig. S7A). At 2 dpf, the diameter of Mo1 morphant eyes was significantly smaller, as compared to control eyes [analysis of variance (ANOVA) with post hoc Dunnett test P < 0.0001, n = 30 embryos/condition] (Supplementary Material, Fig. S7B). This eye diameter of Mo1 morphants at 3 and 5 dpf was more than two standard deviations below the means of control morpholino and non-injected embryos (−2.68 at 2 dpf, −2.67 at 3 dpf and −2.55 at 5 dpf Mo1 morphant compared to the controls). While these results indicate the possible combination of microphthalmia with coloboma in Mo1 morphants, further analysis indicated that the total body length of Mo1 morphant larvae was also significantly shorter compared to controls (Supplementary Material, Fig. S7C). As a result, the eye/body length ratio was strikingly similar in the Mo1 morphants and in the controls at all stages, indicating that there is no clear presence of microphthalmia in the Mo1 morphants (Supplementary Material, Fig. S7D). Anophthalmia was not observed in Mo1 morphants.
Proximo-distal patterning of optic vesicle is altered in atxn7 morphant
Multiple studies have shown that genetic perturbations of developmentally important genes and signaling pathways result in the failure of the optic fissure closure in vertebrate eye (39–41). Some of these genes are proposed to form a network of transcriptional factors and signaling molecules that regulate the developmental events for proper eye formation (42). Central in this network is Sonic Hedgehog (shh) that acts as a key regulator of other genes associated with the optic vesicle formation, patterning and morphogenesis (43).
Notably, the Hh signal is required at several developmental stages. Expression happens in the ventral midline of the forebrain where it first regulates the separation of the eye field in two to form bilateral optic vesicles. Subsequently, Hh signaling emanating from the ventral midline controls proximo-distal patterning (P-D axis) of optic vesicle in a concentration-dependent manner; the distal-most region invaginates to form the optic cup and then the retina, while the proximal region gives rise to the optic stalk. Hh signal promotes proximal and represses distal fates by regulating the expression of pax genes: Hh promotes the proximal pax2a in the optic stalk and represses the distal retinal marker pax6 (44,45). Moreover, there is a reciprocal transcriptional repression between pax2 and pax6, which leads to a clear boundary between the optic stalk and vesicle (46).
To test the hypothesis of a possible alteration in Hh signaling pathway, WISH was performed at 18 hpf when the optic vesicle has fully evaginated, using the probes shh, pax2a and pax6 in control and Mo1 morphants. shh labeling in Mo1 morphant was properly localized at the ventral part of the diencephalon in the intermediate region flanked by the optic vesicles but was stronger than in control cases, as seen on dorsal and lateral views (Fig. 6A–C’). High shh expression was consistently visualized in multiple Mo1 injections and WISH experiments. Interestingly, while control pax2a labeling was restrained to the presumptive optic stalk, expression was extended to the optic vesicles in Mo1 morphants (Fig. 6D–E’). Accordingly, we observed a corresponding retraction of pax6 expression in the area of optic vesicles in Mo1 morphants (Fig. 6F–G’). These results indicate an alteration in the proximo-distal patterning of the developing eye in the Mo1 morphants. Importantly, the same alterations of pax2a and pax6 expression have been reported in other mutants and morphants with coloboma phenotype, generated by perturbation of gene expression related to the Hh pathway (47–49). The expansion of pax2a labeling in Mo1 morphant was still visualized at 28 hpf, ruling out the possibility that our observations were caused by a developmental delay (Fig. 6H and I).
atxn7 deficiency alters the expression of Hh and proximo-distal axis gene markers. WISH was used to study the expression dynamics of Hh genes (shh and twhh), markers of the proximo-distal axis (pax2a and pax6) and the shh-regulated gene vax2. Thirty embryos were treated per condition and at least five were imaged. Yolks were removed for better visualization. (A) Schematic diagram with the corresponding bright field image of larvae at 18 hpf stage in dorsal and lateral positions, respectively, highlighting areas of interest—blue optic stalk and green optic vesicle. (B–C’) Representative images of dorsal view and lateral view of stained control and Mo1 morphant embryos with the shh probe. shh probe labels the areas of forebrain and floor plate in control and Mo1 morphants at 18 hpf. However, an increased expression of shh is found in the Mo1 compared to the controls (arrows). Dotted circle indicates area where the optic vesicle is localized in lateral views. (D–G’) The pax2a probe (D–E’) labels the proximal optic stalk in control and Mo1 morphants at 18 hpf, while the pax6 (F–G’) probe labels the distal optic vesicle. The region of pax2a expression is abnormally extended within the optic vesicle in Mo1 region, and conversely, the pax6 expression is substantially contracted within distal optic vesicle. Horizontal lines represent the boundaries of expression in dorsal view, while arrows show in lateral view. (H, I) On the lateral view of head at 28 hpf, the pax2a labels a broader domain of the proximal optic stalk (arrow) Mo1 morphants in comparison with the controls. (J, K) On the frontal view of head at 28 hpf, the vax2 marks the ventral optic cup, and its expression is substantially expanded into the dorsal optic cup in the Mo1 compared to controls. Horizontal lines represent the boundaries of expression in frontal view, while arrow shows expression in the telencephalon (T). (L–M) On the lateral view of the head at 18 hpf, twhh marks the ventral forebrain (black arrow) and the floor plate (red arrow). Compared to the controls, Mo1 morphants present an increased intensity of the staining in both areas. (N, O) On the lateral view of the trunk at 18 hpf, shh marks the notochord and the floor plate. Compared to the controls, Mo1 morphants displayed an increased staining intensity in the notochord area. Y, yolk; N, notochord.
To further confirm the increase of Hh signaling, we tested the expression of vax2 by WISH. vax2 expression in the optic stalk positively correlates with levels of shh and is independent of pax2a expression (50). Consistent with the increased shh signaling, 28 hpf–Mo1 morphants presented an increased labeling of vax2 in the optic stalk and an expansion of the signal into the dorsal retina, compared to controls (Fig. 6J and K). In addition to shh, tiggy-winkle hedgehog (twhh), another member of the Hh family expressed in the zebrafish ventral midline, has also been associated with the development of the retina (44). Similar to shh, twhh appeared to be properly localized in the Mo1 morphants, and its expression was higher compared to controls (Fig. 6L and M).
To determine if the alteration of shh expression level in Mo1 morphants could have other morphological impact beyond coloboma, the formation of somites was analyzed at 72 hpf (51–54). Somites of Mo1 morphants showed a mild but consistent phenotype. The usual chevron shape of somites was maintained in Mo1 morphant embryos; however, their angle was different (Supplementary Material, Fig. S8A and B). When the angle of five first somites anterior to the end of the ventral fin was measured, the mean angle in the Mo1 larvae was significantly more obtuse (117.3° +/− 12.1°) than in non-injected and control embryos (91.7° +/− 4.4°) (ANOVA with post hoc Dunnett test P < 0.0001) (Supplementary Material, Fig. S8C). The increased somite angle of Mo1 morphant did not lead to the flatten phenotype, reported for shh ectopically overexpression in zebrafish (55). This is likely due to the much stronger Hh signaling achieved by ectopic overexpression than in Mo1 morphant. WISH demonstrates that the expression of shh was slightly increased in the area of the neural tube and notochord of Mo1 morphants, consistent with the mild somite phenotype (Fig. 6N and O).
Taken together, these results show that loss of Atxn7 leads to an increased Hh signaling and alteration of the proximo-distal patterning of optic vesicle.
Atxn7 deficiency affects the optic nerve formation
Apart from its role in the formation of the optic cup, Hh signaling is involved at later developmental stages in the regulation of the proliferation and differentiation of progenitor cells (43,56). Multiple retinal cell types, such as RPE, ganglion and amacrine cells (ACs), express Hh genes, and therefore control their own, as well as the differentiation of other retinal cell types. To verify if the eyes of Mo1 morphants presented other alterations than coloboma, retinal sections of 5 dpf larvae were further examined. Lens formation was unaffected, and the overall lamination of the neural retina was normal, as Mo1 morphants presented the characteristic three nuclear layers separated by the two plexiform layers (Fig. 3A and B). The local retinal deformation due to tissue extrusion did not compromise neural retina lamination.
In order to have a closer look at retinal lamination and neuronal differentiation, we have used the transgenic line Spectrum of Fates (SoFa1) (57). SoFa1 expresses a set of fluorescent protein transgenes from promoters specific to one or more cell types. This generates a color code for each class of retinal neurons and allows the simultaneous visualization of the main retinal cell types along the development. Specifically, the red fluorescence protein (RFP) spectrum labels retinal ganglion cells (RGCs) and to a less extent bipolar cells, horizontal cells (HCs) and photoreceptors cells (PhCs); the green fluorescent protein (GFP) shows AC and HC; and the cyan fluorescent protein labels bipolar and photoreceptor cells. The Mo1 morpholino was injected in the SoFa1 line and visualized at 5 dpf, a time point when retinogenesis is normally completed and retinal neurons differentiated. In this genetic background, the lamination and differentiation of retinal neurons were also overall normal in Mo1 morphants (Fig. 7A and B). Interestingly, we could observe the coloboma defect of the Mo1 SoFa1 morphant retina at cellular level and noticed that the nasal and temporal lobes of the retina were not fused and separated by a gap (Fig. 7A’–B’).
atxn7 deficiency affects the optic nerve formation. (A, B’) Confocal 3D lateral view of 5 dpf SoFa1 control (A, A’) and Mo1 morphant (B, B’) retina. In Sofa1, Atoh7:gapRFP transgene expression labels RGC and photoreceptors (PhC) in red, and Ptf1a:Gal4/UAS:gapGFP expression labels AC and HC in green. In SoFa1 control and Mo1 morphant, the layering and differentiation of PhC, AC, HC and RGC were completed and overall normal. Mo1 morphant retina revealed the presence of coloboma defect (B’), as nasal and temporal lobes of the retina are not fused and separated by a gap (white arrowhead). The RGC axon (white asterisk) converged at the posterior part of the retina in control and Mo1 morphants. (C, D) Confocal 3D frontal view of the retina (R) of 5 dpf SoFa1 control (C) and Mo1 morphant (D) in the red spectrum to visualize RGC and their axon. Compared to the control, Mo1 morphant RGC axons fail to bundle and is split in two portions. Scale bar 50 μm. (E–F”) Epifluorescence frontal images of the head of 5 dpf SoFa1 embryos in the red spectrum. In both control (E) and Mo1 morphants (F–F”), the optic chiasma is properly formed (yellow arrow). However, the optic nerve in Mo1 morphants presents atypical routing toward the optic chiasma (green arrow in F–F”). Occasionally, Mo1 morphant presented axons that aberrantly extended toward the anterior commissure (dotted white circle and amplified circle in F’). Scale bar 100 μm. L, lens.
Given that Hh signaling has been shown to play a role in the differentiation of ganglion cells (43), this neuronal cell type was examined more carefully in Mo1 SoFa1 morphants. In control embryos, the ganglion cell axons properly converge in the center of the retina on the lateral view (asterisk in Fig. 7A) and, on the frontal view, extend as a single bundle toward the posterior part of the eye to form the optic nerve (Fig. 7C). However, the ganglion cell axons in Mo1 morphants converge in the retina center but fail to merge properly in a single bundle when extending in the posterior part of the eye (Fig. 7B and D).
Additionally, there has been evidence that the Hh signaling is involved in the regulation of the axon pathfinding. Experiments in chick embryos revealed that the overexpression of shh in the chiasm misrouted the retinal axons in a concentration-dependent manner in an ipsilateral fashion (58). Moreover, analysis of fish mutants of the Hh pathway revealed anomalies in pathfinding (59). Analysis of Mo1 SoFa1 morphants revealed mild changes in the ganglion cell axon pathfinding when observed on the frontal view of the head. In control, the axon showed a typical and regular route by crossing the ventral midline of the diencephalon to form the optic chiasma and, ultimately, to innervate the contralateral tectal lobe of the brain (Fig. 7E). The axons of morphants properly formed the optic chiasma and reached the contralateral tectal lobe. However, 30.76% (n = 4/13) of Mo1 morphant embryos presented atypical routing of the axon (green arrow in Fig. 7F–F”) and aberrant extension toward the anterior commissure (dotted circle in Fig. 7F’), likely due to a more relaxed axon bundle. These results show that retinal layering and neuron differentiation are overall maintained in atxn7-deficient embryos. However, the ganglion cell axon pathfinding and the optic nerve bundling are affected, which is consistent with the ganglion cell alterations upon increased expression of shh previously reported (60).
Photoreceptor terminal differentiation is altered in atxn7-deficient zebrafish
While histological and SoFa1 analyses indicated that the lamination and differentiation of neural retina were overall normal in Mo1 morphant, the monolayer of RPE showed clear interruption due to extrusion of neural retina in the forebrain. The interaction between photoreceptor and RPE is essential for proper visual function (61). Moreover, during development of the retina, the RPE is a source of essential secreted factors, including shh, for the differentiation of photoreceptors (56,62,63). Interestingly, the layering of photoreceptors appeared normal in areas lacking RPE in Mo1 morphant (Fig. 8A and B), and this was confirmed by electron microscopy analysis (Fig. 8C and D). In control embryos at 5 dpf, terminally differentiated photoreceptors showed the typical polarization with elongated nuclei at the basal side, followed by the inner segments that contain mitochondria and then the OSs at the apical side (Fig. 8C). The OSs are the latest built structure of photoreceptors and are normally juxtaposed to the RPE monolayer. Despite the absence of RPE, photoreceptors in Mo1 morphants showed the characteristic polarized morphology but with shorter OSs (Fig. 8D). Surprisingly, in regions where the RPE was not disrupted and the retina not extruded in Mo1 morphants, OSs were strongly altered, at variable degrees along the retina. Some regions showed polarized photoreceptors with OSs as short as in areas lacking RPE (compare Fig. 8F and G), while in other retinal regions, photoreceptors completely lacked OSs (Fig. 8H). Finally, there were more affected regions where photoreceptors presented disorganized inner segments lacking the characteristic mitochondria localization in the most apical area, and their OSs were absent (Fig. 8I). Importantly, no obvious sign of cell death was visualized in these different retinal regions of Mo1 morphants. Together, these observations suggest that the terminal differentiation of photoreceptor leading to the formation of OSs is compromised at different degrees along the retina in Mo1 morphant, independent of the presence or absence of RPE.
atxn7 deficiency compromises the terminal differentiation of photoreceptor. (A–D) Coronal sections of 5 dpf control and Mo1 morphant retina stained with Toluidine blue (A, B) and adjacent sections (area in dotted yellow box from A’ and B’) imaged by transmission electron microscopy (C, D). Photoreceptors (PhC, white bracket) in control (C) and Mo1 morphant (D) were polarized with mitochondria (M) in the apical side of the inner segment, directly below the OS. Control photoreceptor OS (yellow bracket) was elongated and the RPE contained melanin-filled granules. In the Mo1 morphant, photoreceptor OS was small and faced a region devoid of RPE (yellow arrowhead). Scale bar 10 μm in A–D. N, nucleus; INL; inner nuclear layer; OPL, outer plexiform layer; BC, bipolar cell; HC, horizontal cell. (E–I) Electron microscopy images with higher magnification of 5 dpf larvae retinal sections. In control (E), RPE showed melanin-filled granules, while photoreceptors presented long OS, polarized inner segment with the characteristic mitochondria in the most apical area just below the OS and elongated nucleus. (F) In regions where RPE is lacking in Mo1 retina, photoreceptors had smaller OS (yellow arrowhead). (G–I) In some regions where RPE is present, Mo1 photoreceptors had smaller OS (G) or no OS at all (red arrow) (H). (I) In other regions, Mo1 photoreceptors do not have OS (red arrow) and also lack the distinctive inner segment ellipsoid with apical mitochondria. Scale bar 2 μm. (J–M) WISH comparing the expression of crx in control and Mo1 morphant at different hpf. Representative images of dissected eyes from 48 hpf control and Mo1 morphants (J, K), control morphant at 42 hpf (L) and Mo1 morphant at 52 hpf (M). Thirty embryos were treated per condition and at least five were imaged. L, Lens.
Therefore, we hypothesized that, rather than signals coming from the RPE, intrinsic factors of the photoreceptors might be responsible for the alteration of their differentiation in Mo1 morphants. One important transcription factor controlling photoreceptor terminal differentiation in mammalian retina is the cone–rod homeobox protein (CRX) (31). crx was shown to be essential for proper differentiation of zebrafish photoreceptor (64). WISH was performed with a crx probe at 48 hpf. Compared to the controls, Mo1 morphants presented a significant difference in the intensity and localization of the crx labeling. Notably, Mo1 morphants presented areas with decreased or complete lack of labeling, which could explain the different degrees of photoreceptor differentiation (Fig. 8J and K). To determine if this was due to a delay in development, WISH was performed in control embryos at 42 hpf. Compared to the 48 hpf–Mo1 morphants, 42 hpf–control embryos presented homogenous and widespread expression of crx, except for very few areas free of labeling, which did not match the abnormal crx expression pattern seen in Mo1 morphant retinas (compare Fig. 8K and L). Furthermore, compared to Mo1 morphants at 48 hpf, Mo1 morphants at 54 hpf still presented regions lacking crx expression (compare Fig. 8K and M). Together, these results indicate that the downregulation of atxn7 leads to alteration in the expression of crx and compromises the terminal differentiation of the photoreceptors.
Atxn7 deficiency affects SAGA-dependent histone modification
The HAT and DUB modules of SAGA are involved, respectively, in the acetylation of lysine 9 of histone H3 (H3K9ac) and the removal of monoubiquitin from lysine 120 of histone H2B (H2Bub) (22). To determine if SAGA-dependent activities are affected in atxn7 morphants, protein extracts from whole 24 hpf embryos injected with control or with Mo1 were analyzed on western blot using antibodies detecting the modified histones. The data indicate that the global level of H2Bub was decreased compared to control, while the level of H3K9ac was not changed (Supplementary Material, Fig. S9). The data suggest that the reduction of atxn7 expression could impact on SAGA activities.
Discussion
Congenital coloboma is an eye malformation caused by a failure in the closure of the optic fissure during embryonic development and represents an important cause of pediatric visual impairment with the high estimated prevalence of 1/5000 live birth in humans (65–67). Although many genes have been identified to contribute to coloboma, they account for a small fraction of the clinically reported cases (37,39). In the present work, we provide compelling evidence for a role of Atxn7 in regulating the choroid fissure closure in zebrafish. Inactivation of atxn7—through two types of antisense oligonucleotides (a translational blocker and a splice blocker) and F0-derived CRISPR/Cas9 approaches—primarily resulted in an altered proximo-distal patterning of the optic vesicle and coloboma. The role of Atxn7 in the regulation of the choroid fissure closure is consistent with the spatiotemporal expression pattern of atxn7 transcript, which presents maternal contribution and early and persistent zygotic expression, with high expression level in the eye and forebrain. Coloboma phenotype was observed at low concentrations of Mos, suggesting high sensitivity of the ocular development to variation of Atxn7 level. Finally, ectopic expression of human ATXN7 partially rescued the coloboma phenotype, demonstrating that the function of ATXN7 in the closure of the choroid fissure is conserved through vertebrate evolution. Coloboma can occur in combination with microphthalmia and anophthalmia as part of a spectrum of human ocular malformations (37). In atxn7-deficient zebrafish, anophthalmia was not observed, and although the eyes were smaller than in control zebrafish, this was in strict proportion to a smaller body size, suggesting that the structural anomaly of the eye was restricted to coloboma. Coloboma phenotypes could vary in severity depending on the location and the presence of other associated eye defects, which could include alteration of the cornea, iris, ciliary body, lens, choroid, RPE and optic nerve. Our observations in atxn7-deficient embryos indicate that the failure to close the choroid fissure was associated with interruption of RPE monolayer, expansion of retinal tissue along the optic nerve into the forebrain, unbundled optic nerve and misrouting of ganglion cell axon, while the lamination of retinal neurons appeared normal.
The molecular mechanisms of choroid fissure closure have not been completely understood. It is well documented in animal models and in human patients that alteration in Hh signaling can cause proximo-distal patterning defects and coloboma (41). In particular, Hh regulates a number of genes, which have also been directly associated with ocular coloboma (42,43). However, not much is known about how proper levels of Hh activity are regulated during eye development. In this study, we observed an increased level of shh and twhh in the ventral midline of the developing forebrain of atxn7-deficient embryos at 18 hpf. At this precise site and time point, Hh signaling controls the proximo-distal patterning of the optic vesicle by regulating the expression of two genes: pax2a, expressed in the optic stalk, and pax6, expressed in the optic vesicle (44,45). Accordingly, we found that the signal of pax2a was abnormally extended into the area of the presumptive optic vesicle, while the expression of pax6 was retracted in the area of the optic stalk. This abnormal pattern of pax2/pax6 expression is a characteristic consequence of elevated Hh signaling in the ventral midline at 18 hpf as reported earlier (44,45,47,49,68). Therefore, our data suggest that loss of Atxn7 at very early developmental stage in the forebrain accounts for dysregulation of Hh signaling in the ventral midline, which then impacts on the proximo-distal patterning of the optic vesicle. The aberrant expansion of ventral vax2 expression into the dorsal retina of atxn7-deficient embryos is also consistent with elevated shh signaling pathway (50) and could suggest additional alteration of the dorso-ventral axis of the eye. However, at later stages of eye development, atxn7 inactivation did not seem to impact on Hh signaling, since retinal layering and neuronal differentiation were overall maintained in Atxn7 morphant embryos, a situation previously reported for other gene, causing elevated shh and coloboma (68). Additional experiments are necessary to determine how Atxn7 controls the spatio-temporal level of Hh and if restoring the Hh signaling using specific inhibition can rescue the proper proximo-distal patterning of the optic vesicle.
The involvement of Atxn7 in the regulation of Hh signaling is new but consistent with previous work demonstrating the regulation of Shh by SAGA. Interestingly, mouse embryos lacking acetyltransferase activity of the Kat2a (named Gcn5hat/hat mice) showed an increased Shh signaling in the zona limitans intrathalamica (ZLI), leading to diencephalic expansion (69). Kat2a was shown to restrict diencephalic expansion during early forebrain development by promoting retinoic acid (RA) signaling. Failure to propagate RA signaling leads to increased Shh in Gcn5hat/hat mice. It is proposed that Kat2A/SAGA forms a complex with RA receptors and TACC1 (a transcriptional repressor) on the promoter of RA-responsive genes. In the presence of RA, Kat2a acetylates TACC1—a non-histone-substrate—which then dissociates from the complex, allowing transcriptional activation of RA target genes. atxn7-deficient embryos also displayed increased shh staining in the presumptive ZLI at 18 hpf (Fig. 6C’) and malformation of the brain at 5 dpf (data not shown). Additional studies will be necessary to determine the morphological and molecular aspects of brain malformation in atxn7-deficient zebrafish.
The differentiation of photoreceptors is terminated by the formation and elongation of OSs and is completed at 5 dpf in zebrafish embryos. Most photoreceptors in atxn7-deficient embryos at this age were properly polarized and had developed inner segments. However, OSs were either absent or much shorter than normal, suggesting that atxn7 is required for complete photoreceptor terminal differentiation. The incomplete formation of OSs likely explains the decreased expression of rhodopsin in the retina of similar atxn7 morphants previously reported by Yanicostas et al. (30). In the latter study, the underlying mechansim of rhodosin alteration was not explored. Here, we show that malformation of photoreceptors correlates with reduced expression of crx. Crx is one of the major transcription factors regulating rhodopsin expression. The lack of crx expression was not due to developmental delay (Fig. 8J–M) and was likely not a consequence of elevated Hh signaling in the ventral midline of atxn7-deficient embryos, because the expression of crx was previously shown to be independent of shh expression in zebrafish (70).
Interestingly, Crx knockout mice have polarized photoreceptors with normal inner segments. However, OS morphogenesis was found to be completely blocked at the elongation stage in these mutants (71). The similarity between photoreceptor malformation in atxn7-deficient zebrafish and Crx knockout mice further supports that Atxn7 regulates crx expression in zebrafish photoreceptor. It was previously shown that ATXN7 directly interacts with CRX, allowing SAGA to execute epigenetic modifications on CRX-regulated genes (15,72). Therefore, ATXN7 might have a dual role in terminal differentiation of photoreceptor by regulating the expression and transactivation function of CRX.
Mutation of the yeast Sgf73, the ATXN7 homolog, leads to an increased monoubiquitination of H2B (73,74). In contrast, we found that the global level of H2Bub was decreased in atxn7-deficient zebrafish. Our results are consistent with data reported in fly mutated for atxn7 and in HeLa cells knocked down for ATXN7 in which decreased level of H2Bub was also observed (29). To explain the decreased ubiquitination of H2B, it is proposed that the lack of ATXN7 leads to the release of an active DUB module from SAGA and hence the loss of SAGA-related regulation of the free module (29). In contrast to H2Bub, we did not detect modification of the global level of H3K9. However, since our analysis was performed on the whole embryo, it remains possible that the effect on H3K9ac was restricted to affected tissues, such as the ventral midline and neural retina. Nevertheless, our observations suggest that the phenotype of atxn7-deficient zebrafish could be related to dysfunction of Atxn7 within the SAGA complex.
The requirement of Atxn7 in the formation of OSs helps to understand the vulnerability of this structure in SCA7. The OS is a structure filled entirely with discs of folded double membranes in which the light sensitive visual pigments are embedded. In adult retina, OSs are dynamic structures undergoing daily partial shedding at the apical side juxtaposed to RPE and re-synthesis at the base using macromolecules transported through the inner segments. Thus, OSs renewal necessitates high biosynthetic activity and gene expression. In SCA7 mice, photoreceptors progressively lose their OSs due to the progressive decreased expression of photoreceptor-specific genes (13,17–19). Therefore, we propose that one of the primary toxic effects of polyQ expansion is to alter the function of ATXN7 in the morphogenesis of OS, compromising the renewal of OS and eventually leading to photoreceptor degeneration.
In summary, our results demonstrate that Atxn7 plays dual essential roles in vertebrate eye morphogenesis and formation of photoreceptor OSs, by influencing Shh signaling and Crx expression, respectively. Our data further indicate that loss of ATXN7 contributes to the development of coloboma.
Materials and Methods
Zebrafish
Zebrafish were maintained at 28°C in a standard zebrafish facility as described previously (75). Embryos were kept at 28.5°C in water with or without 0.003% 1-phenyl-2-thiourea (PTU, Sigma-Aldrich, Lyon, France) to prevent pigment formation. Embryos were staged according to hpf or dpf as previously described (76). Wild-type embryos were from the AB strain, and transgenic SoFa1 strain was kindly provided by Dr William A Harris (57).
RT-PCR analysis
Total RNA was extracted from 50 embryos at required stage by TRIzol Reagent (Invitrogen, Illkirch, France) and reverse transcribed into cDNA using the SuperScript II Reverse Transcriptase (Invitrogen) according to the manufacturer’s instructions. cDNA fragment (694 bp) encompassing the atxn7 exons 1 to 5 was amplified using the atxn7-specific primers [forward (fw) 5′-CCCTGCCTAGTCCCGAAATA-3′] [reverse (rev) 5′-GACGATTGGTGGCCTTTCC-3′]. Expression of β-actin (fw 5′-CGAGCAGGAGATGGGAACC-3′) (rev 5′-CAACGGAAACGCTCATTGC-3′) was used as an internal control. To determine the efficiency of the atxn7 Mo2, cDNA fragments (471 wild type and 590 Mo2), encompassing exon 3 and exon 5 were amplified using atxn7-specific primers (fw 5′-GGCCTTCCAAGCACATTAC-3′) and (rev 5′-TTGTCTTGGGACGATTGGTG-3′). PCR products were cloned with PCR Cloning Kit Vector pJet1.2 (Thermo Scientific, Illkirch, France), according to the manufacturer’s instructions, and sequenced. Paralogs expression was analyzed using the corresponding primers: atxn7L1 (fw 5′-GTCTACCCACCCAAAGGAGC-3′) (rv 5′-TCCAGAGGAGCCGAGAGAAA-3′), atxn7L2a (fw 5′-AGAGGAACCAAGACGCACAA-3′) (rv 5′-AAAGTGAAGACACCGTGACC-3′), atxn7L2b (fw 5′-GCACGCGCTAAAACGGTAAT-3′) (rv 5′-TCTCGTGGTTGACAGACCCT-3′) and atxn7L3 (fw 5′-TCAGAGTCTTCCTCCTTGGG-3′) (rv 5′-GCCTGCTATTCTGTCTCGCT-3′).
Mo-mediated knockdown of zebrafish atxn7
Mo1 blocking the translation of atxn7 (5′-CGTCATCATCGGCCCTTTCCGACAT-3′), Mo2 blocking the splice site (5′-CAAGCGGAAGGTGGTCTTACCGTAA-3′) and a standard control Mo (5′-CCTCTTACCTCAGTTACAATTTATA-3′) were obtained from Gene Tools (Oregon, USA). Embryos at one- or two-cell stages were injected at indicated Mo concentrations using a Nanoject II micro-injector (Drummond Scientific, France). BLAST search with the Ensembl zebrafish database (Zv9) for the proposed morpholino indicated that there is no presence of similar sequence elsewhere in the genome that could be a potential off-target.
Phenotypic rescue
Capped human mRNA for microinjection was prepared with SP6 mMESSAGE mMACHINE SP6 kit (Ambion, Illkirch, France) using linearized pCS2+ vector containing the full-length coding sequences for the human wild-type ATXN7a with 10Q (N10). For rescue experiments, a mix containing Mo1 (3 ng) and human atxn7 N10 mRNA (150 ng) was injected into embryos at the one- to two-cell stage using a Nanoject II micro-injector (Drummond Scientific), and the phenotypes were analyzed at the indicated stage.
CRISPR sgRNA and Cas9mRNA synthesis and injection
Atxn7 target sites were identified, and the corresponding sgRNA oligos were designed using the Chop-Chop online software (77). Oligos were mixed with the Master mix and tracer fragment T7 included in the GeneArt Precision gRNA Synthesis kit (Invitrogen) for generation of the gRNA template, followed by a PCR amplification (120 bp). In vitro transcription of the template was conducted with the same kit according to the manufacturer’s instructions. A mix containing 2 μg/μL Cas9 protein from S. pyogenes (PnaBIO, California, USA) and 1 μg/μL of gRNA was injected into embryos at the one-cell stage using a Nanoject II micro-injector (Drummond Scientific). To determine the efficiency of the gRNA, genomic DNA of 24 hpf embryos was obtained by alkaline lysis and amplified with atxn7-specific primers (fw-5′CCTCAGAAATTCGCGCACAC 3′) (rv-5′ ATTTCGGGACTAGGCAGGGA 3′). PCR amplification of the target region (357 bp) was performed on genomic DNA of individual embryos using flanking primers, and products were analyzed by heteroduplex mobility assay to determine the presence of indel mutations. For the heteroduplex mobility assay, PCR products were analyzed by electrophoresis on 15% polyacrylamide gels. For sequence analysis, PCR products were subcloned with PCR Cloning Kit Vector pJet1.2 (Thermo Scientific), according to the manufacturer’s instructions, and sequenced. Sequence analysis was performed using Serial Cloner 2.6. Alignment was performed using Clustal Omega (78) and assembled with Jalview 2.10.1 (79).
Phenotype analysis
Different retina phenotypes were analyzed between 24 hpf and 5 dpf, according to the desired experiment. Embryos were anesthetized with 0.02% Tricaine (Sigma-Aldrich) and mounted in 3% methylcellulose. Bright field images were captured using Leica M420 Macroscope with COOLSNAP coupled camera.
Angle measurement and body length
Images captured using Leica M420 Macroscope with COOLSNAP coupled camera were analyzed using ImageJ software (80). Corresponding statistical analysis and figures were performed using GraphPad Prism version 6.00 for Windows.
In situ hybridization
Anti-sense RNA probes were generated by in vitro transcription of the corresponding plasmid containing a portion of the coding sequence of the gene of interest, using SP6, T7 or T3 polymerase and DIG (Roche Applied Science, Indianapolis, USA). The atxn7 was generated by cloning PCR products into the pJET1.2/blunt Cloning Vector (Thermo Scientific), according to the manufacturer’s instructions, and sequenced. Subsequently transcribed using the T7 polymerase, shh, twhh and pax6 were kindly provided by Dr Julien Vermot. The plasmids for pax2.a and Vax2 were kindly provided by Dr Stephen Willson. The plasmid for crx was kindly provided by Dr Pamela Raymond. Embryos were fixed in 4% PFA at 4°C overnight or for 4 h at room temperature, and in situ hybridization on whole-mount embryos was performed principally as described previously (81). Images were captured using Leica M420 Macroscope with COOLSNAP coupled camera.
Western blot analysis
Protein was extracted from pools of 80 to 100 control and atxn7 morphant embryos at 24 hpf. Protein concentrations were measured using the BCA Protein Assay Kit (Protein Assay Dye Reagent, BIORAD, France). 100 μg of total protein per sample was diluted 1× final with Laemmli buffer and DTT 0.1 M, boiled for 5 min and then separated by SDS-PAGE on 10% polyacrylamide gels. Resolved proteins were transferred to nitrocellulose membranes and blocked in 3% milk 1× TBS for 1 h at room temperature prior to incubation with either anti-H2B antibody (1:5000, produced in house), anti-H2Bub (1:2000, MM_0029-P, MediMAbs, Montreal, Canada) or anti-H3K9ac (1:2000, Ab 6661, Abcam, U.K.). Membranes were washed and incubated in goat anti-rabbit or anti-mouse peroxidase secondary antibody (1:10000, Jackson Immuno Research, U.K.). Blots were developed using Immobilion Western (Millipore, France) according to the manufacturer's instructions.
Axon pathfinding analysis
SoFa1-injected embryos were kept at 28.5°C in water with 0.003% PTU to prevent pigment formation. Embryos were anesthetized with 0.02% Tricaine (Sigma-Aldrich). For general optic nerve visualization, embryos were mounted in 3% methylcellulose in a depression slide and oriented for imaging. Images were captured using Leica DM 400B with COOLSNAP coupled camera using the 10× objective. For eye 3D models, SoFa1 embryos were staged, anaesthetized with 0.02% Tricaine (Sigma-Aldrich) and mounted in 0.5% low melting-point agarose (Sigma-Aldrich). Confocal imaging was performed on a Leica SP8 confocal microscope. Images were acquired with a low-magnification water immersion objective.
Tissue sections and electron microscopy
Five dpf larvae were fixed in ice-cold 4% glutaraldehyde in 0.1 M sodium cacodylate + 0.001% CaCl2 (pH 7.4), for 1 h at 25°C with gentle agitation and then with fresh fixative overnight at 4°C. Samples were washed three times for 10 min with 0.1 M sodium cacodylate (pH 7.4) + 0.001% CaCl2, postfixed in 1% osmium tetroxide in 0.1 M sodium cacodylate (pH 7.4) + 0.001% CaCl2. Samples were dehydrated through graded ethanol (50, 70, 90 and 100%) and propylene oxide for 30 min each. Samples were oriented and embedded in Epon 812. Semi-thin (2 μm) sagittal sections were cut with an ultramicrotome (Leica Ultracut UCT) and stained with toluidine blue. Images were captured using Leica DM 400B with COOLSNAP coupled camera using the 20× objective. Selected sections were histologically analyzed by light microscopy. Images were captured with Mega View II (Soft Imaging system), and contrast was adjusted for display purposes. Ultrathin sections were cut at 70 nm and contrasted with uranyl acetate and lead citrate and examined at 70 kv with a Morgagni 268D electron microscope. Images were captured digitally by Mega View III camera (Soft Imaging System).
Sequence analysis of zebrafish atxn7 ortholog and paralogs
The Ensembl zebrafish (Zv9) database identifies a single orthologue of human ATXN7 gene in the zebrafish. The gene named atxn7 (ENSDARG00000074804) is localized on chromosome 11 and comprises 14 exons, 12 of which encode a single protein of 866 amino acids (referred as Atxn7, ENSDART00000110499.3). Human ATXN7 and paralog sequences were used to identify ortholog sequences in zebrafish using ENSEMBL database similarity searches (http://www.ensembl.org/): human ATXN7 (ENST00000295900.10), ATXN7L1 (ENST00000419735.7), ATXN7L2 (ENST00000419735.7), ATXN7L3 (ENST00000454077.6); zebrafish Atxn7 (ENSDART00000110499.3), Atxn7L1 (ENSDART00000122605.2), Atxn7L2a (ENSDART00000090520.4), Atxn7L2b (ENSDART00000143922.1) and Atxn7L3 (ENSDART0000010418.3). Alignment was performed using Clustal Omega (78) and the figure was generated with Jalview 2.10.1 (79). Conserved domains in Atxn7 and paralogs were also reported in (20).
Acknowledgements
We are grateful to J. Vermot and his team for support and helpful discussion, to P. Kessler for confocal imaging, to J. Hergueux for EM sample preparation, to C. Golzio for the development of CRISPR/Cas9 mutant and to S. Geschier and S. Gredler for zebrafish stock care. We warmly thank N. Daigle, F. Klein and D. Oladosu for their critical reading and helpful comments on the manuscript.
Conflict of Interest statement. None declared.
Funding
Agence Nationale de la Recherche Ciliataxia [ANR-13-BSV1-0016-01, ANR-10-LABX-0030-INRT] under the frame programme Investissements d’Avenir labelled ANR-10-IDEX-0002-02; the Fondation pour la Recherche Médicale [DVS20131228917]; Association Rétina-France; Association Connaître les Syndromes Cérébélleux; S.C.R. was supported by a Laboratory of Excellence (LABEX) PhD fellowship [ANR-10-LABX-0030-INRT].
