frizzled 5 mutant zebrafish are genetically sensitised to developing microphthalmia and coloboma Open Access

Microphthalmia, anophthalmia and coloboma (MAC) is a spectrum of ocular malformations arising from defects during early stages of eye formation that constitutes the most severe cause of childhood blindness (Plaisancié et al., 2019; Williamson and FitzPatrick, 2014; Yoon et al., 2020). Anophthalmia and microphthalmia describe, respectively, the absence of the eye or the presence of a reduced ocular globe, while coloboma typically refers to the persistence of an opening in the ventral portion of the eye. More than 100 MAC-associated genes have been identified in the past two decades. However, it is estimated that these genes explain only ∼30% of the cases described in human disease (Plaisancié et al., 2019; Williamson and FitzPatrick, 2014). One possible reason for this is that genetic variants may predispose to MAC, but defects are manifested only when combined with other risk factors or additional pathogenic gene variants. Therefore, such variants may potentially be overlooked by conventional searches, despite their relevance to understanding the aetiology of MAC.

Eye formation is critically dependent on the Wnt signalling pathway (Fuhrmann, 2008; Shah et al., 2023; van Amerongen and Nusse, 2009; Wiese et al., 2018). Wnt ligands bind to the extracellular domain of Frizzled (Fzd) receptors and a variety of co-receptors. Fzd receptors are seven-pass transmembrane proteins with an extracellular N-terminal domain that interacts with Wnt ligands. Upon Wnt/Fzd interaction, a conformational change promotes activation of Dishevelled (Dvl) (Bowin et al., 2023), inducing different responses in a highly context-dependent manner. Some of these involve transcriptional regulators, such as β-catenin, Yap or Jun, while others lead to modulation of cytoskeletal dynamics or calcium signalling (Shah et al., 2023; van Amerongen and Nusse, 2009).

Several of the components of the Wnt signalling network lead to eye malformations when mutated in animal models. For example, eye-specific disruption of β-catenin results in eye malformations in mouse (Hägglund et al., 2013; Westenskow et al., 2009), and functional abrogation of the Wnt co-receptor LDL-receptor-related protein 6 (LRP6) or the Wnt antagonist Dickkopf 1 (DKK1) result in microphthalmia and coloboma (Lieven and Rüther, 2011; Zhou et al., 2008). A role for the Wnt pathway in eye formation is further highlighted by the essential role of the Wnt/β-catenin pathway effector, Tcf7l1, during eye field specification (Kim et al., 2000; Andoniadou et al., 2011; Young et al., 2019), and by temporally dynamic expression of several Fzd genes in the eye primordium during embryogenesis (Fuhrmann et al., 2003; Nikaido et al., 2013). Among them, fzd5 is the only Fzd receptor showing an eye-specific expression pattern (Cavodeassi et al., 2005; Fuhrmann et al., 2003; Sumanas and Ekker, 2001; Burns et al., 2008; Van Raay et al., 2005; Liu and Nathans, 2008; Nikaido et al., 2013).

Recent studies have uncovered a series of variants of FZD5 associated with microphthalmia and/or coloboma in humans (Aubert-Mucca et al., 2021; Holt et al., 2022; Jiang et al., 2021; Liu et al., 2016). A total of 21 cases with missense, nonsense or frameshift variants have been described (Cortés-González et al., 2024; Holt et al., 2022). Only two variants have been functionally analysed in vitro and in vivo (Cortés-González et al., 2024; Liu et al., 2016). The first (Liu et al., 2016) is a frameshift variant leading to a premature termination codon, which generates a truncated form of the protein missing the transmembrane and the C-terminal domain. The truncated form of FZD5 was shown to work as a dominant negative, interfering with Wnt activity, and it has been proposed that this would explain the autosomal dominant inheritance pattern leading to ocular phenotypes described in these patients. The second (Cortés-González et al., 2024) is a missense variant leading to the replacement of a highly conserved proline by a leucine at the junction between the first intracellular domain and the second transmembrane domain (Pro267Leu). This variant has been shown to behave as a hypomorph, unable to efficiently transduce the Wnt signal (Cortés-González et al., 2024). In this case, the pattern of inheritance is recessive, and the variant in heterozygosis does not display any defective phenotype.

Conditional loss of function of Fzd5 in the mouse eye leads to microphthalmia, coloboma and persistent foetal vasculature (Liu and Nathans, 2008), phenotypes that are exacerbated when the function of Fzd8, co-expressed with Fzd5 in the eye primordium, is downregulated (Liu et al., 2012). fzd5 anti-sense RNA approaches have also been described in Xenopus and zebrafish, which result in smaller eye primordia and reduced proliferation in the eye anlage (Cavodeassi et al., 2005; Liu et al., 2016; Van Raay et al., 2005).

Here, we describe the generation of four fzd5 mutant alleles in the zebrafish, two reproducing a complete loss of function condition, and two reproducing a dominant-negative condition. Homozygote fzd5 mutant embryos show no major ocular phenotype. However, quantification of eye size indicates that the eyes of fzd5 homozygotes are smaller than those of heterozygotes or wild-type embryos. We show that further downregulating Wnt activity in these mutants exacerbates the ocular defects, suggesting that the mutants have compromised Wnt activity. In addition, interfering with the activity of other genes potentially involved in eye formation in these fzd5 mutant lines results in eye malformations in an additive or synergistic way. Our results add novel insight into the mechanisms leading to eye phenotypes in fzd5 mutants and highlight the importance of considering multiple genetic defects when searching for the molecular aetiology of ocular malformations in humans.

In zebrafish, fzd5 expression starts at 10 h post-fertilisation (hpf) in the eye field (Cavodeassi et al., 2005) and continues at optic vesicle (Fig. 1A,B; 12-16 hpf) and optic cup stages (>24 hpf) up until the start of retinal differentiation (Fig. 1C). As the eye matures, fzd5 expression is downregulated in the differentiated retina and only maintained in the ciliary marginal zone (CMZ) (Fig. 1D). Expression can also be detected throughout embryonic development in the ventral telencephalon and hypothalamus (Fig. 1A,D). To assess the requirement for Fzd5 in eye formation in zebrafish, we used genome editing to generate four mutant alleles (Fig. 1E): two predicted loss-of-function alleles (sgu1 and sgu2) and two (sgu3 and sgu4) mimicking the dominant-negative variant described in humans (Liu et al., 2016).

Two short deletions of eight (fzd5^sgu1) or five (fzd5^sgu2) nucleotides were generated by transcription activator-like effector nuclease (TALEN) injection (see Materials and Methods; Fig. 1E) and resulted in frameshifts and premature termination codons, giving rise to truncated products of 36 [p.(His18Leufs*37)] and 37 [p.(His18Leufs*38)] amino acids, respectively (Fig. 1E). These truncated forms lack an intact Wnt binding functional domain and are thus predicted to behave as complete loss-of-function mutations.

The fzd5^sgu3 and fzd5^sgu4 alleles were generated by CRISPR-Cas9, injecting RNA guides designed to edit a region of zebrafish fzd5 that is homologous to that affected in the dominant human FZD5 variant described by Liu et al. (2016) (Fig. 1E). The fzd5^sgu3 allele harbours a 15-nucleotide insertion, generating a premature termination codon at position 231 [p.(Cys232*); Fig. 1E]. This form encodes an intact extracellular domain and no transmembrane domains (the first transmembrane domain starts at position 235 of the wild-type protein), and thus the truncated product of this allele is predicted to have dominant-negative effect. fzd5^sgu4 bears a 19-nucleotide deletion leading to a frameshift and a stretch of 13 non-sense amino acids before a premature termination codon at position 242 [p.(Ala228Argfs*243)]. The extracellular domain of fzd5^sgu4 is intact up to amino acid 228, and, similar to fzd5^sgu3, no transmembrane domains are present (Fig. 1E). Consistent with a predicted truncation of Fzd5 produced by fzd5^sgu3, a fusion form of fzd5^sgu3 tagged to red fluorescent protein (fzd5^sgu3-RFP) fails to localise to the cell membrane (Fig. 2D-F,J,K, magenta or grey; arrowheads in Fig. 2F,K), in contrast to the cell-membrane localisation of a wild-type fusion form of fzd5 (wt-fzd5-RFP) (Fig. 2A-C,G-I, magenta or grey; arrowheads in Fig. 2C,G,I).

A luciferase reporter assay (TOPFlash; Hua et al., 2018) in HEK293 cells revealed that the ability of fzd5^sgu3 to promote Wnt signalling was severely compromised (Fig. 2L). Cells transfected with 50 ng M50 8X TOPFlash only showed minimal luciferase activity. Robust activation of the Wnt pathway can be achieved by co-transfection of Fzd receptors with lrp6 (Hua et al., 2018). Indeed, co-transfection of TOPFlash with wild-type fzd5-myc and lrp6 resulted in a 6.6-fold increase in luciferase activity compared to co-transfection of TOPFlash with lrp6 alone (Fig. 2L). However, co-transfection of lrp6 with fzd5^sgu3-myc (Fig. 2L) failed to activate above the levels seen with lrp6 alone (Fig. 2L).

A dominant-negative effect of fzd5^sgu3 upon wild-type fzd5-induced TOPFlash activity was revealed by co-transfecting a constant level of wild-type fzd5-myc DNA with increasing amounts of fzd5^sgu3-myc and normalising to luciferase activity for wild-type fzd5-myc alone (Fig. 2M). Equimolar amounts of wt-fzd5 and fzd5^sgu3 resulted in ∼75% luciferase activity, a twofold excess of fzd5^sgu3 reduced this to 50% and a fivefold excess reduced this further to 25% (Fig. 2M). We then selected the fzd5^sgu1 and fzd5^sgu3 alleles for further analysis of the effect of loss-of-function and dominant-negative forms of Fzd5, respectively, on eye formation.

No gross morphological defects were observed in homozygotes for both fzd5^sgu1 and fzd5^sgu3 alleles, but embryonic eyes were generally smaller. These differences, although subtle, were consistent and statistically significant between different genetic backgrounds when the projected eye area was measured (Fig. 3A,B and Fig. 7D). To determine whether there was a subtle/mild coloboma phenotype or a delay in choroid fissure fusion, we performed in situ hybridisation with pax2.1, a gene expressed in the lips of the choroid fissure and optic nerve, at the time of choroid fissure fusion. No difference in pax2.1 expression was observed in either zygotic or maternal zygotic (MZ) fzd5^sgu1 mutants or zygotic fzd5^sgu3 mutants (Fig. 3E-J; an example of altered pax2.1 expression associated to coloboma can be seen in Fig. 5F and Fig. S2B,C). No differences were observed in fzd5 expression in the proliferating CMZ of zygotic fzd5^sgu1 mutants (Fig. 3C,D). Retinae of homozygous fzd5^sgu1 animals that survived to adulthood showed normal lamination and no obvious structural defects (Fig. S1).

The findings that zygotic and MZfzd5 homozygotes displayed only a subtle reduction in eye size during embryonic stages raised the possibility that other Fzds or Wnt pathway components compensate for the loss of Fzd5 to maintain pathway activity during eye development. For instance, several other Fzd genes are expressed throughout the anterior neural plate and eye primordia during embryogenesis (Nikaido et al., 2013). If so, we hypothesised that the mutants may be sensitised to the effect of other manipulations that disrupt Wnt signalling. To assess whether this was the case, we treated clutches of embryos derived from mating fzd5^+/− parents with threshold levels of XAV-939, a selective antagonist of tankyrase activity that reduces Wnt/β-catenin activity (Kulak et al., 2015).

XAV-939 treatments of embryos derived from fzd5^sgu1/+ parents led to a subset of embryos displaying coloboma. From 135 treated embryos in two independent experiments, 25 (19%) showed coloboma (Fig. 4B, compare with DMSO-treated control in Fig. 4A). Subsequent genotyping confirmed that homozygote fzd5^sgu1 embryos were more sensitive to the treatment than wild types (Fig. 4A-C,E; Table S1). Indeed, 22/25 embryos in the coloboma group were fzd5^sgu1 homozygotes and none were wild type (Fig. 4C). These genotypic distributions (0 wild type: 3 heterozygote: 22 homozygote) markedly deviated from the typical Mendelian distributions of genotypes (1 wild type: 2 heterozygote: 1 homozygote; Table S1). No fzd5^sgu1 homozygotes were identified in a group of over 40 embryos without a coloboma phenotype (Fig. 4C; Table S1).

XAV-939 treatment on MZfzd5^sgu1 embryos resulted in 62% of them (33/53) showing coloboma (arrow in Fig. 4F, compare with DMSO-treated MZfzd5^sgu1 control embryos in Fig. 4D; see quantifications in Fig. 4G). A reduction in eye size also occurred in all MZfzd5^sgu1-treated embryos. Notably, the coloboma/small eye phenotype observed in XAV-939-treated MZfzd5^sgu1 embryos was more severe than that observed in XAV-939-treated zygotic fzd5^sgu1 embryos (compare Fig. 4F and B). Consequently, although eye size phenotypes are similar in zygotic and MZfzd5^sgu1 embryos (Fig. 4A,B; see quantifications in Fig. 3A,B and Fig. 7D), the loss of maternal Fzd5 further compromises the ability of the forming eye to cope with additional modulations to Wnt pathway activity.

As for the fzd5^sgu1 allele, XAV-939 treatment of embryos derived from the mating of heterozygous fzd5^sgu3 parents led to a subset of embryos displaying coloboma (Fig. 4I, arrow; compare with DMSO-treated control embryo in Fig. 4H), but in this case the percentage of colobomatous embryos was much higher (20/32 total embryos, 62%; Fig. 4J) than that recovered from identical treatments in the fzd5^sgu1 background (19%; see Fig. 4C). A defective eye phenotype was recovered not only in homozygous embryos (7/8 embryos show coloboma) but also in most heterozygotes (12/16 heterozygotes show coloboma; Fig. 4J; Table S1). Thus, the distribution of genotypes in the coloboma group (1:12:7) and the group with wild-type eye morphology (7:4:1) markedly deviated from a Mendelian 1:2:1 distribution (Table S1). Collectively, these results suggest that fzd5^sgu3/+ heterozygote embryos are more sensitive to further abrogation of Wnt activity than fzd5^sgu1/+ embryos.

The experiments above suggest that a truncated form of Fzd5 expressed from its endogenous locus may act as a dominant-negative form to compromise the ability of the forming eye to maintain Wnt signalling. However, as this effect is only evident when the pathway is additionally compromised by drug treatment, perhaps the levels of the dominant-negative form present in vivo in the fzd5^sgu3 mutants do not significantly affect pathway activity on their own. To test this possibility, we used a transgenic line that expresses GAL4 under the control of the eye-specific rx3 promoter (tg{rx3:Gal4}; Hernández-Bejarano et al., 2015; Weiss et al., 2012; Hernández-Bejarano et al., 2022) to express high levels of exogenous dominant-negative Fzd5 within the forming eye (UAS:fzd5DN; see Materials and Methods).

tg{rx3:Gal4};UAS:fzd5DN embryos showed defective optic vesicle evagination resulting in optic disc and optic nerve defects. Control tg{rx3:Gal4} embryos expressed the pan retinal marker mab21l2 in the optic vesicles, and no expression was observed in the forebrain midline (Fig. 5A). In contrast in tg{rx3:Gal4};UAS:fzd5DN embryos, mab21l2-expressing cells were present in forebrain tissue between the eyes, suggesting incomplete optic vesicle evagination (Fig. 5B,C, asterisk; 11/11 embryos showed phenotype). By 72 hpf, optic disc coloboma was observed in a subset of the tg{rx3:Gal4};UAS:fzd5DN embryos (Fig. 5E, asterisk; Fig. 5F, arrows; 7/17; compare with control in Fig. 5D).

These results suggest that interfering with Wnt/Fzd function in the retinal primordia leads to eyes with coloboma and optic nerve defects. In support of this conclusion, we observed optic nerve phenotypes upon overexpression of a dominant-negative form of the Wnt co-receptor Lrp6 [LRP6-DC, predicted to interfere with Wnt activity (Cavodeassi et al., 2005; Mao et al., 2001; Tamai et al., 2000)], either in the whole embryo (by LRP6-DC mRNA injection into one-cell-stage zebrafish embryos) or only in the eyes (by injection of a UAS:LRP6-DC construct into tg{rx3:Gal4} embryos; Fig. S2B,C,E, asterisk; compare with wild type in Fig. S2A,D).

The results above show that further downregulating Wnt activity in homozygous fzd5^−/− fish leads to exacerbated ocular phenotypes and suggest that these animals may be compromised in their ability to cope with additional genetic or environmental factors that challenge the robustness of Wnt signalling. To explore this in more detail, we next used a multiplex CRISPR-Cas9 approach to abrogate the function of lrp6 and fzd4 in fzd5^sgu1 mutant embryos (F0 approach; Fig. 6A,B; Table S1; Kroll et al., 2021). These two genes are expressed in the eyes, and while Lrp6 functions together with Fzd5 to activate the Wnt pathway, Fzd4 is a close paralogue of Fzd5 potentially showing functional redundancy with Fzd5.

Whereas abrogation of lrp6 activity by F0 injection did not cause observable phenotypes in wild-type embryos (Fig. 6F), 33% of embryos (21/64) developing visibly smaller eyes were observed upon identical injections in embryos derived from fzd5^sgu1/+ parents (Fig. 6D, compare to wild type in Fig. 6C). Genotyping revealed that the small-eye group contained most of the homozygous fzd5^sgu1 embryos in the batch (17/19 homozygous embryos have small eyes; Fig. 6F), and the phenotypically wild-type group contained most of the heterozygotes and wild types (Fig. 6F). The genotypic distributions in both cases, as well as in two additional independent experiments, markedly deviate from typical Mendelian proportions (Fig. 6F; Table S1). Quantifying the projected eye area confirmed that the homozygous fzd5^sgu1 embryos had smaller eyes than the wild-type, phenotypically normal embryos (Fig. 6G; fzd5^+/+ versus fzd5^−/−, P=1.10×10⁻⁷; fzd5^+/− versus fzd5 ^−/−, P=5.02×10⁻²²). However, the percentage of eye area reduction between lrp6 F0-injected fzd5^sgu1 homozygotes and wild types (∼14%) was not different from that seen in uninjected fzd5^sgu1 and fzd5^sgu3 homozygote embryos compared to wild types (see Figs 2 and 7). This result suggests that removing lrp6 function does not exacerbate the small eye phenotype detected in fzd5^sgu1 homozygote animals. A lack of genetic interaction is also suggested by the observation that lrp6 F0 injection did not lead to a reduction of projected eye area in fzd5^sgu1 heterozygotes (Fig. 6G).

fzd4 abrogation by F0 injection led to small eyes in embryos derived from the mating of fzd5^sgu1/+ parents in two independent experiments [32% (12/38) or 19% (8/43) of small eyes, respectively; Fig. 6E]. Genotyping showed that the small-eye group contained all the fzd5^sgu1 homozygotes and a proportion of heterozygotes; no phenotype was observed abrogating fzd4 in wild-type embryos (Fig. 6F; Table S1). Quantification of projected eye areas confirmed a statistically significant reduction in eye size in both fzd5^sgu1 homozygotes and small-eye fzd5^sgu1/+ heterozygotes compared to the normal-eye size groups (Fig. 6H; fzd5^+/+ versus fzd5^−/−, P=0.005402; fzd5^+/+ versus fzd5^+/−-small eyes, P=0.000992; fzd5^+/−-normal eyes versus fzd5^−/−, P=0.000015; fzd5^+/−-normal eyes versus fzd5^+/−-small eyes, P=0.000004). Thus, we concluded that embryos homozygous or heterozygous for the fzd5^sgu1 mutation were more prone than wild types to developing small eyes when fzd4 activity was abrogated.

The genetic interactions presented above confirmed the sensitised nature of the fzd5^sgu1 genetic background and suggested that these mutants could be exploited to identify novel genes that, when disrupted, enhance susceptibility to eye malformations. To explore this, we selected six genes never associated with eye defects before, based on their co-expression with fzd5 in the eye field as identified from single-cell RNA-sequencing datasets (Farrell et al., 2018). We then used the F0 approach to abrogate the function of these genes in fish heterozygous or homozygous for the fzd5^sgu1 allele. Only one of the six genes tested [angio-associated migratory cell protein (aamp)] showed a genetic interaction with the mutant fzd5 allele. All the others [starmaker (stm), Cbp/p300-interacting transactivator, with Glu/Asp-rich carboxy-terminal domain, 2 (cited2), PPARG-related coactivator 1 (pprc1), switching B cell complex subunit a/b (swap70a/b)] did not show any phenotype in wild-type embryos and did not exacerbate the small-eye phenotype present in fzd5 mutants (Fig. S3).

aamp encodes a poorly studied, but highly conserved, WD40 repeat domain-containing protein. In vitro studies suggest a role for Aamp in blood vessel development and smooth muscle migration, but no roles in eye formation have been reported (Beckner and Liotta, 1996; Vogt et al., 2008). aamp expression is highly enriched in the early optic vesicle and maintained in the eye primordium. As development progresses, its expression becomes restricted to the ciliary marginal zone, as well as other proliferative domains in the forebrain and tectum (Liu et al., 2016; Thisse and Thisse, 2004).

Abrogation of aamp generated by F0 injections resulted in smaller eyes at 4 days post-fertilisation (dpf), with an average reduction in projected eye area of 25% (Fig. 7A,C). The small-eye phenotype was reproduced and fully penetrant in a fish homozygous for a stable predicted loss-of-function allele (aamp^u601) containing an 8 bp insertion in exon 2 (Fig. 7A,B; Fig. S4). In addition to small eyes, mutant aamp^u601embryos showed a variable lower jaw reduction (Fig. 7A, arrow), and failed to inflate the swim bladder (Fig. 7A).

Abrogation of aamp function by F0 injection in embryos from a cross of fzd5^sgu1/+ parents resulted in an eye area reduction of 42% (Fig. 7D). This constituted a significant reduction in eye size compared to the 25% reduction in aamp F0 injections in a wild-type background and the 16% reduction observed in uninjected fzd5^sgu1 homozygote embryos (Fig. 7C,D). Thus, this approach uncovered a novel additive genetic interaction between loss of aamp and fzd5 in the optic primordium.

Variants of FZD5 in humans have been shown to lead predominantly to isolated coloboma and less frequently to coloboma and microphthalmia (Holt et al., 2022; Jiang et al., 2021); but, to date, there is no clear understanding of the cause of this phenotypic variability. Here, we analysed two fzd5 mutants in the zebrafish, one reproducing a complete loss-of-function condition (fzd5^sgu1) and a second reproducing a predicted dominant-negative condition (fzd5^sgu3). Eye formation seemed largely unaffected in homozygote embryos for both alleles, with only subtle reduction in eye size revealed upon quantification. However, coloboma was additionally present upon further downregulation of Wnt activity in the homozygote mutants.

Our results indicate that the fzd5^sgu3 allele compromises Wnt signalling more than fzd5^sgu1. Downregulation of Wnt activity by XAV-939 treatment of embryos derived from a fzd5^sgu3/+ incross led to coloboma not only in fzd5^sgu3 homozygotes, but also in most of the heterozygote embryos. This is markedly different from the result obtained from XAV-939 treatments of embryos derived from a fzd5^sgu1/+ incross, in which coloboma was only observed in homozygote embryos. Collectively, our results suggest that zebrafish embryos can compensate for loss of Fzd5 activity, but fzd5 mutants are predisposed to developing microphthalmia and coloboma in conditions in which Wnt activity is further compromised.

Our results do not entirely fit with the phenotypes described in humans. Human variants have been predominantly associated with isolated coloboma and, on rare occasions, to microphthalmia/coloboma, but only one case showing microphthalmia in the absence of coloboma has been reported (Holt et al., 2022; Jiang et al., 2021). Moreover, the mutants in zebrafish described in this study have a much milder phenotype than that observed in humans. Indeed, most of the human cases described in the literature show dominant inheritance of the mutant variants (Holt et al., 2022), and, to date, only one case has been described in humans carrying a variant with a recessive inheritance pattern (Cortés-González et al., 2024). Our interpretation is that the variants described in humans predominantly lead to more severe abrogation of Wnt activity than the conditions we described in this study. However, this does not explain why most of the cases described in humans display coloboma in the absence of microphthalmia.

The sensitised nature of the zebrafish fzd5 mutants described in this study may provide an explanation for the dominant inheritance and variable expressivity of many of the alleles described in humans. Indeed, by simultaneously abrogating other genes potentially involved in eye formation in the fzd5^sgu1 background, we could identify novel genetic interactions on eye size with fzd5 (fzd4 and aamp). A proportion of the FZD5 cases described in humans have been shown to carry variants also in other loci, and, at least in one case, compound variants of FZD5 and another gene (DDX3X) required for eye morphogenesis and leading by itself to coloboma has been described (Holt et al., 2022). Thus, other patients may carry still unidentified variants of other genes contributing to the eye phenotypes described. Notably, most of the FZD5 cases in the literature have been identified by whole-exome sequencing or by using customised next-generation sequencing panels of genes involved in ocular development, and thus potential variants in regulatory sequences of other interacting genes may have been overlooked. The identification of additional variants could be facilitated by assessing genetic interactions of candidate genes in the zebrafish mutants described in this study.

Although loss-of-function mutants for lrp6 and fzd4 in isolation have no effect on eye morphogenesis and eye size, aamp loss of function results in fully penetrant reduction of eye size. Aamp encodes a highly conserved WD40 repeat domain-containing protein (Beckner and Liotta, 1996; Vogt et al., 2008). In humans, WD40 repeat domains (WDRs) make up one of the most abundant protein–protein interaction domains, and WDR-containing proteins play important roles in nearly all major cellular signalling pathways (Haag et al., 2021). Variants of WDR proteins have been associated with various human pathologies including neurological disorders and holoprosencephaly (Haag et al., 2021), but the in vivo requirements for aamp are not known. Homozygous aamp^U601 mutants are not viable up to adulthood, suggesting that the mutant phenotype is not exclusive for the eye and corroborating in vitro studies that suggest a role for aamp in blood vessel development and smooth muscle migration (Beckner and Liotta, 1996; Vogt et al., 2008). Given the expression of aamp in proliferative niches of both the retina and the forebrain (Thisse and Thisse, 2004), we hypothesise that the small-eye phenotype is the result of a reduction in retina cell proliferation. Only seven AAMP loss-of-function cases were identified in the Genomics England 100,000 Genomes Project rare disease cohort. A subset of these patients show intellectual disabilities; however, there is no association with ocular defects, although subtle eye phenotypes may have been overlooked. Alternatively, eye phenotypes may only manifest in humans when AAMP is mutated in combination with other genes relevant for eye morphogenesis and associated with eye malformations, such as FZD5.

In summary, the animal models for fzd5 presented in this study constitute valuable novel tools to unravel the genetic network cooperating with fzd5 during eye formation, opening the exciting opportunity to exploit them to identify novel genetic interactions relevant to understand the aetiology of coloboma and microphthalmia.

AB and tupl wild-type zebrafish strains, the transgenic line Tg{rx3::Gal4-VP16}^vu271Tg (Weiss et al., 2012), and mutant lines fzd5^sgu1, fzd5^sgu2, fzd5^sgu3, fzd5^sgu4 and aamp^U601 were maintained and bred according to standard procedures (Aleström et al., 2020; Westerfield, 1993). Mutant lines were maintained in heterozygosis. With attentive husbandry, homozygous fzd5^sgu1 fish could be grown to adulthood and bred to obtain maternal zygotic mutant embryos. All experiments conform to the guidelines from the European Community Directive and the British (Animal Scientific Procedures Act 1986) legislation for the experimental use of animals.

The fzd5^sgu1 and fzd5^sgu2 loss-of-function alleles were generated with the TALEN approach. TALEN arms were generated by Keith Joung's team and acquired from Addgene (TAL6232 and TAL3263; target sequence TCGGATTTTGGCTGCATGtcctgctgctgtttcaACTGTCTGGGCTCGGAGA; Sander et al., 2011). The TALEN arms target the 5′ region in the fzd5 locus, between nucleotides 137 and 144 in the open reading frame. TALEN mRNAs were synthesised (mMessage mMachine SP6 kit, Ambion) and purified (NEB RNA cleanup kit) following manufacturer instructions. F0 founders were generated by co-injection of the two TALEN arm mRNAs into one-cell-stage AB/tupl embryos. Genotyping of F0 founders and their progeny was performed by CRISPR-STAT (Carrington et al., 2015) or high-resolution melt (HRM) analysis (Parant et al., 2009) from genomic DNA samples obtained from tail fin biopsies. The primers used are detailed in Table S2.

The target sequence to generate the fzd5^sgu3 and fzd5^sgu4 dominant-negative alleles was selected using the Ensembl Genome Browser. The optimal target DNA sequence on the fzd5 open reading frame (ENSDARG00000025420) was identified following the recommendations from Hwang et al. (2013). To generate the guide RNAs, oligonucleotides (Table S2) were ordered from Integrated DNA Technologies (IDT), annealed and cloned in PDR274 or pCD039, linearised with DraI, transcribed in vitro (NEB Hiscribe RNA synthesis kit) and purified (NEB RNA cleanup kit) following manufacturer instructions. Cas9 mRNA was synthesised (mMessage mMachine SP6 kit, Ambion) and purified (NEB RNA cleanup kit) following manufacturer instructions. F0 founders were generated by co-injection of guide RNA (25 ng/µl) and Cas9 mRNA (300 ng/µl) into one-cell-stage AB/tupl embryos. Cleavage efficiency was assessed in pools of injected embryos by HRM analysis. Genotyping of F0 founders and their progeny was performed by HRM analysis from genomic DNA samples obtained from tail fin biopsies. The primers used are detailed in Table S2.

The aamp^u601 allele was generated by CRISPR/Cas9 (IDT) using a guide RNA targeting exon 2 of aamp (Table S2), just downstream of the start codon. The guide RNA was annealed to tracrRNA oligonucleotide (IDT #1072532), assembled with Cas9 protein (IDT #1081058) and injected into one-cell-stage AB/tupl embryos.

Guide RNAs to target lrp6, fzd4, stm, cited2, pprc1, swap70a and swap70b were generated as described above, and co-injected with Cas9 mRNA into AB/tupl embryos or embryos derived from the incross of fzd5^sgu1 heterozygous parents at one-cell stage. Guides were assessed for efficiency by genotyping injected wild-type embryos, and the minimum concentration of guides necessary to lead to efficient gene editing was subsequently selected to test genetic interactions. Best combinations often involved guides in adjacent exons to attempt to generate large deletions, in addition to small indels. Injected embryos did not show any non-specific phenotypes or developmental delay and were analysed to assess eye morphology, categorised according to phenotype and subsequently genotyped for fzd5. Identification of indels in lrp6 and fzd4, and fzd5 genotypic status was determined by HRM analysis on DNA samples obtained from individual embryonic tails. Identification of indels in stm, cited2, pprc1, swap70a and swap70b was determined by MiSeq analysis. At least two rounds of injection were done for each tested gene, and one of them was genotyped in full.

aamp loss of function was phenocopied by F0 injection as previously described (Kroll et al., 2021). Three synthetic RNA guides were designed (Table S2) and ordered from IDT. Guides were annealed to tracrRNA oligonucleotide (IDT #1072532), assembled with Cas9 protein (IDT #1081058) and injected into one-cell-stage wild-type embryos (Kroll et al., 2021). A subset of the injected embryos was genotyped by HRM analysis (primers described in Table S2) to confirm the presence of gene editing.

lrp6-DC-pCS2+ was Addgene plasmid #27258; RRID: Addgene_27258). The fzd5-DN fragment was generated by amplifying the first 687 bp [encoding for the first 229 amino acids and lacking all transmembrane and cytoplasmic domains) from the fzd5-pCS2+ plasmid (Cavodeassi et al., 2005); primers detailed in Table S2]. The lrp6-DC and fzd5-DN fragments were subcloned into a UAS/Tol2 bidirectional plasmid (Distel et al., 2010; Hernández-Bejarano et al., 2015; Kajita et al., 2014).

pCS2-wt-fzd5-RFP (Cortés-González et al., 2024) was used as a template to generate fzd5^sgu3-RFP, fzd5-myc and fzd5^sgu3-myc fusion constructs using the NEB Q5 site-directed mutagenesis kit (E0554; primers detailed in Table S2).

Suboptimal concentrations of XAV-939 (Sigma-Aldrich) were determined by treating wild-type embryos with a series between 100 µM and 15 µM. A concentration of 30 µM was selected for subsequent treatment. Embryos were collected and treated at shield stage. Eye phenotypes were scored at 3 dpf, and embryos were genotyped by HRM analysis as described above.

Overexpression of fzd5-DN and lrp6-DC in the optic vesicle was performed with the Gal4/UAS system (Halpern et al., 2008). 20-30 ng of GFP:UAS:fzd5-DN or GFP:UAS:lrp6-DC plasmid DNA was injected into the cell of one-cell-stage Tg{rx3::Gal4-VP16}^vu271Tg (Weiss et al., 2012) embryos. Only embryos with homogeneous GFP expression in the optic vesicles were selected and processed for analysis (as described in Hernández-Bejarano et al., 2015). An average of 15 embryos were processed for each marker/stage analysed.

Fzd5 protein localisation was examined in zebrafish embryos using a fzd5-RFP C-terminal fusion (pCS2-wt-fzd5-RFP; Cortés-González et al., 2024). pCS2-wt-fzd5-RFP and pCS2-fzd5^sgu3-RFP were used as templates to synthesise mRNA for injection (mMessage mMachine SP6 kit, Ambion). 200 pg of mRNA was injected into one-cell-stage fertilised embryos, allowed to develop at 30°C until dome stage (4.5 hpf) and fixed overnight in 4% paraformaldehyde. Embryos were briefly washed in PBS+0.3% Triton X-100 and incubated with phalloidin-FITC (at 0.5 µM, to detect subcortical actomyosin) and Hoechst (at 1 µg/ml, to detect DNA) in PBS+1% Triton X-100+1% DMSO for 4 h at room temperature, briefly washed in PBS and mounted in 1% low-melt-point agarose for imaging.

HEK293 cells were maintained in Dulbecco's modified Eagle medium with 10% foetal calf serum at 37°C in a humidified atmosphere of 5% CO₂. Transfections were carried out using polyethylenimine as described (Longo et al., 2013).

For immunofluorescence analysis, cells were plated on poly-L-lysine-coated glass coverslips and fixed in 4% paraformaldehyde at 24-48 h after transfection then stained with 10 µM Hoechst 33342. Coverslips were mounted on glass microscope slides using ProLong Diamond Antifade (Thermo Fisher Scientific).

Cells for TOPFlash assays were plated in quadruplicate on 24-well plates and transfected with M50 Super 8X TOPFlash (Addgene plasmid #12456; RRID: Addgene_12456) along with pRLTK (Promega) for normalisation. Activation of the Wnt pathway was achieved by co-transfection of lrp6 and Fzd constructs, as described (Hua et al., 2018). Cells were processed 48 h after transfection with the Dual-Luciferase^® Reporter Assay System (Promega). Luminescence readings were made with a GloMax^® Discover Microplate Reader. Data were analysed by pairwise multiple Student’s t-test using Excel and JASP software (Version 0.17.3) on macOS 10.15.7.

Preparation of RNA antisense probes and whole-mount in situ hybridisation was performed as previously described (Hernández-Bejarano et al., 2015). In situ hybridised clutches derived from fzd5^+/− incrosses were genotyped in full for fzd5, and all homozygote embryos were imaged and compared to wild-type controls.

Adult tissue was collected and fixed as previously described (Moore et al., 2002), decalcified by incubation in 0.35 in EDTA pH 7.8, dehydrated in an ethanol series and cleared in methyl salicylate before embedding in wax. Sections were cut in a Leica RM2255 rotary microtome at 7 µm thickness. Sections were collected on slides coated with aminopropyl triethoxysilane, baked overnight at 37%, and stained with Haematoxylin and Eosin following standard protocols. High-resolution images of the slides were acquired on a Hamamatsu Nanozoomer 2.0-RS slidescanner, and images were selected and exported using NDP-View2 software.

Embryos were staged according to the staging tables from Kimmel et al. (1995). All embryos looked morphologically similar, and no evidence of developmental delay was detected. Embryos were fixed in 4% paraformaldehyde and embedded in 3% methylcellulose for imaging. Images were acquired using a Nikon SMZ1270 microscope with a Plan Apo 1× WF WD 70 mm objective and a Leica MC190 HD camera, operated by Leica LAS EZ imaging software. Larvae were imaged lying on their sides such that only one eye was visible. Eye size measurement was performed in Fiji (Schindelin et al., 2012). Images were thresholded, a region of interest (ROI) was drawn to isolate the eye, and the thresholded area within the ROI was measured. Data were plotted using the python library Seaborn (Waskom, 2021), and statistical analysis was performed with the Pingouin package (Vallat, 2018).

In situ hybridised embryos and dissected eyes were mounted flat in a drop of glycerol, and dorsal or lateral images were acquired with a 20× (0.70 NA) dry lens using a Nikon Eclipse microscope connected to a digital camera (DS-Fi3) and operated by Nikon software (NIS-Elements).

Imaging of dome-stage embryos and transfected mammalian cells expressing wt-Fzd5-RFP and Fzd5^sgu3-RFP fusions was performed in a Nikon A1R inverted confocal microscope with a 40× dry lens or a 100× oil immersion lens, respectively. Images were acquired with Nikon NIS Elements C software. Raw confocal images were processed with ImageJ.

Processed images were exported as TIFF files, and all figures were composed using Photoshop.

We thank Sergio Salgüero and Ilyas Ali for their contribution to preliminary analysis of the fzd5^sgu1 line; Yvette Bland for her help in the preparation of adult tissue for histology sections; members of our laboratories and our colleagues Lucas Fares-Taie and Kenzo Ivanovitch for critically reading the manuscript. We acknowledge the Image Resource Facility and the Biological Research Facility at City St George's University of London, and the Zebrafish Facility at University College London, for their support with image acquisition and animal care.