The transcription factor Maz is essential for normal eye development

Formation of the eye is an evolutionarily conserved developmental process that leads to morphogenesis of the retina, lens and associated eye structures. Disturbances within the eye's developmental cascade are responsible for a spectrum of ocular abnormalities, often leading to partial or complete loss of vision. The Wnt pathway is an established, crucial and conserved signaling pathway in eye morphogenesis. Wnt pathway signaling is essential for multiple developmental events during embryogenesis, regulating cell proliferation and fate decisions (Liu et al., 2007; Livesey and Cepko, 2001). The pathway also has a fundamental role in human diseases (Wang et al., 2019). Wnt proteins activate target cells by binding to the Frizzled (Fzd)/low-density lipoprotein receptor related protein (LRP) complex at the surface of the cell (MacDonald et al., 2009; Nusse and Varmus, 1992). The activated Fzd/LRP complex transduces Wnt signaling into the cell through canonical and non-canonical signaling pathways. The canonical Wnt/β-catenin pathway acts through β-catenin as a transcriptional cofactor, whereas the non-canonical pathway [Wnt/planar cell polarity (PCP) and Wnt/Ca²⁺ pathway] through FZD receptors and/or ROR1/ROR2/RYK co-receptors activates the PCP, receptor tyrosine kinase (RTK) or Ca²⁺ signaling cascades (Katoh, 2017).

Antagonistic regulation of extracellular Wnt signaling is achieved through two broad classes of secreted proteins: one of these is collectively referred to as Dickkopf (Dkk) and the other is the secreted frizzled-related protein (Sfrp) class (Kawano and Kypta, 2003). Members of the Dkk class inhibit the binding of Wnt to the co-receptors LRP5/6 to antagonize Wnt signaling. Members of the Sfrp class can bind directly to Wnt ligands through an extracellular cysteine-rich domain with homology to the Fzd receptors, and thus alter their ability to bind and activate Fzd.

In the developing eye, Wnt signaling controls multiple developmental and morphogenic patterning processes. These processes include the dorsoventral patterning of the optic cup, the lens, the retinal pigment epithelium (RPE), the vascular system and the ciliary margin (CM) (Carpenter et al., 2015; Drenser, 2016; Fuhrmann, 2008; Fujimura, 2016; Hagglund et al., 2013; Liu et al., 2007; Wang et al., 2019). The peripheral rim of the optic cup is the point where the non-pigmented inner layer and the pigmented outer layer meet, and is a unique region of the eye cup that forms two peripheral tissues, specifically the ciliary body (CB) and the iris (Davis-Silberman and Ashery-Padan, 2008). Gain- and loss-of-function experiments, both in vitro and in vivo, demonstrate that Wnt/β-catenin signaling is required for the proper formation of these structures and for the proliferation of progenitors in the CM (Cho and Cepko, 2006; Liu et al., 2003, 2007). Wnt2b, expressed in the dorsal RPE, and Fzd4 are both bona fide candidates that regulate CB and iris formation, as well as control the expansion of stem cells (Kubo et al., 2003, 2005). Wnt/β-catenin activity is tightly controlled in the peripheral retina by Foxg1, Sox2, Axin2 and Sfrp1/2 (Alldredge and Fuhrmann, 2016; Esteve et al., 2011; Fotaki et al., 2013; Heavner et al., 2014; Matsushima et al., 2011). Although Sfrps antagonize Wnt signaling by sequestering Wnt proteins, they might have alternative roles as the Sfrp1/2 compound null mutant showed inactive Wnt signaling and phenotypic alteration of the peripheral retina displaying abnormal neural retina characteristics (Esteve et al., 2011).

Maz, a gene encoding Myc-associated zinc-finger transcription factor, is expressed ubiquitously during development and is required for normal genitourinary (GU) development (Haller et al., 2018). The MAZ protein binds to purine-rich promoters that contain a consensus sequence similar to the one found in promoters regulated by Wilms’ tumor 1 (WT1) protein. WT1 modulates mitogen-activated protein kinase (MAPK) signaling and the Wnt pathway (Kim et al., 2009). In vitro knockdown of MAZ in HEK293 cells results in differential expression of several WNT morphogens required for normal GU development (Haller et al., 2018). Much less is known about the role of Maz in the developing eye. Here, we show that reduction or elimination of Maz expression in mice leads to a spectrum of eye phenotypes. We provide evidence that Maz exerts its role modulating Wnt pathway activity during eye development by suppressing the expression of Sfrp2 in the developing CM. Using a combined approach of exome sequencing (ES) and copy-number variant (CNV) analysis, we identified mutations in the human homolog MAZ underlying ocular disorders. Our results reveal that Maz is an important component of the molecular pathways controlling eye development.

In situ hybridization (ISH) of Maz at different stages of mouse development revealed ubiquitous expression in developing tissues, including the eye (Haller et al., 2018). At embryonic day (E)10.5, Maz is found in the embryonic eye with higher levels in the dorsal retina. At E14.5, Maz expression is in the neuroretina and in the CM, whereas in the lens, expression is restricted to the anterior lens epithelium (Fig. S1). At later stages [E16.5 to postnatal day (P)10], Maz transcripts are found in the outer neuroblastic layer with some scattered cells showing higher expression, whereas in the adult eye, Maz was detected in Müller glia cells (Blackshaw et al., 2004).

To investigate whether Maz deletion results in eye abnormalities, mice harboring a recently described targeted deletion of Maz were examined (Haller et al., 2018) (Fig. 1A-C). Mice heterozygous for the deletion (Maz^+/−) were relatively normal, with a slight deviation in survival from Mendelian expectations. However, ∼90% of homozygous null mutants (Maz^−/−) died before weaning, demonstrating that Maz activity was essential for postnatal viability. Analysis of Maz-deficient embryos at different stages of development showed them to die at the perinatal stage with developmental eye defects. Homozygous inactivation of Maz caused a variable eye phenotype in 80% of the mutants, ranging from unilateral microphthalmia, sometimes with coloboma, to bilateral anophthalmia (Fig. 1D,E). Some C57 BL/6J wild-type (WT) animals (5%) exhibit ocular abnormalities, which appear phenotypically different from the previously identified eye phenotype associated with the C57BL/6N genetic background (Mattapallil et al., 2012); however, the penetrance of these abnormalities is significantly increased in a gene-dose-dependent fashion in Maz^+/− (21%) and Maz^−/− (82%) mice (Table S1 and Fig. 1D). Histological analysis of E18.5 Maz^−/− eyes, the latest stage at which Maz-deficient embryos can be consistently recovered, revealed a spectrum of mutant eye phenotypes ranging from grossly normal eyes of slightly smaller size to mutants with severely affected eyes with significantly reduced size (Fig. 2B-D) in comparison to the WT eye (Fig. 2A). Mildly affected mutant eyes (Fig. 2D) displayed anterior segment dysgenesis (ASD), abnormal keratolenticular connections with absence of anterior segment (Peter's anomaly) and occasionally abnormal lens fibers. In contrast, the WT eye (Fig. 2E) exhibited the lens and cornea separated by the anterior segment, and the lens presenting both the epithelial cells and organized fibers. In severely affected mutants (Fig. 2D), the microphthalmic eyes were accompanied by different anomalies including ventral coloboma. In some cases, an ectopic expansion of the peripheral dorsal CM was observed (Fig. 2C,J) compared to the dorsal CM in the WT eye (Fig. 2A,I). In extreme cases, both the dorsal and ventral pigmented epithelium were thickened to resemble a mirror image of the subjacent neuroretina in comparison to the normal dorsal (Fig. 2I) and ventral (Fig. 2K) pigmented epithelium in the WT, and the lens was entirely absent (Fig. 2L). In a few extreme cases, the ocular tissue was absent (data not shown).

Phenotypic analysis showed that ∼30% of Maz^+/− adult mice present eye abnormalities with lens opacities observed consistently. Detailed histological comparison of eyes in WT and Maz^+/− adult mice showed that haploinsufficiency of Maz causes a significant decrease in retinal thickness and a change in the open angle of the ciliary body (Fig. S2). Homozygous adult mice that survived for a few weeks included two males with no obvious eye phenotype and two runted females with very microphthalmic eyes. Eye tissue was recovered from only one of the affected females and consisted only of RPE (data not shown).

To determine the onset of these abnormalities in eye development, we investigated eye-specific gene expression and morphogenesis during early eye development of Maz mutants. During development, the eye field at the anterior neural plate gives rise to the optic vesicles, which then turn into the optic cups. The eye field expresses several genes that are crucial for eye formation, including Rax, Pax6, Six3, Lhx2, Otx2 and Six6 (Mathers and Jamrich, 2000). When we examined morphogenesis and expression of Rax, Pax6 and Lhx2 at E9.5 and E10.5, we found that at E9.5 their expression domains were similar in Maz^−/− and WT embryos (Fig. 3A,B,I,J). At this stage of embryonic development, there were no obvious genetic or morphological differences between the developing WT and mutant eyes, as exemplified by Rax (Fig. 3A,B) and Pax6 (Fig. 3I,J) expression.

At E10.5, expression of these genes was still very similar (Fig. 3C-H), although in some cases, as shown by Pax6 expression, the eye field was very small and located abnormally close to the brain (Fig. 3E,F). At this stage, the optic cups are formed on schedule in both WT and mutant mice, but the first morphological differences between Maz^−/− and WT embryos appear at the periphery of the optic cup. In the Maz mutant, the dorsal fold-back hinge-point is overgrown and displaced to the outside of the optic cup (Fig. 3K,L). Beginning at E12.5, major differences appear between WT (Fig. 3Q,S,U) and mutant (Fig. 3R,T,V) eyes. The mutant eyes are smaller and grossly altered with defects in both the dorsal and ventral optic cup (Fig. 3N-P). The expressivity of the phenotype in the mutant is variable, ranging from a smaller eye with ectopic expansion of the peripheral optic cup and the lens abnormally rotated (Fig. 3N) to the complete absence of the ventral region of the eye (Fig. 3O). In a few cases, embryos appear anophthalmic, but sections reveal that they have a small optic cup with RPE tissue attached to the brain (Fig. 3P). By E14.5, the dorsal retina of Maz mutant eyes is frequently overgrown and the lenses are either abnormally positioned or absent (Fig. 3Q,R). In some mutant embryos, a rudimentary eye structure can be found close to the base of the brain (Fig. 3S,T). At E16.5, in some mutants the retina is abnormally thick and defects were observed in the extraretinal hyaloid vasculature (Fig. 3U,V).

To better understand the functional consequences of the inactivation of Maz in eye formation, we investigated if the morphology and proliferation of the CM was altered. We queried whether this region had lost its defining characteristics. Gene expression and proliferation in the CM of Maz mutants was compared with that observed in WT eyes.

The peripheral CM, from which the ciliary body and iris are derived, is characterized by a high level of Pax6 expression (Matsushima et al., 2011), expression of Otx1 (Martinez-Morales et al., 2001) and Msx1 (Monaghan et al., 1991), as well as an absence of expression of the cell-cycle regulator Ccnd1. This region also displays a low proliferation rate, as diagnosed by low 5-bromo-2'-deoxyuridine (BrdU) incorporation (Trimarchi et al., 2009) and by the relative absence of phosphohistone H3 (PPH3)-positive cells. At E16.5, Pax6 expression is highest in the distal tips of the retina of both the WT and Maz mutant eyes (Fig. 4C,D). However, Pax6-expressing cells are irregularly positioned in the CM of Maz mutants with an abnormal thickness (Fig. 4D) compared to the WT retina (Fig. 4C). This abnormal morphology is also observed in H&E-stained CM coronal sections (Fig. 4A,B). Expression of the retinal progenitor cell marker Rax revealed that this gene is normally expressed in the prospective neuroretina with lower levels in the CM (Fig. 4E,F), but the area of low Rax expression is enlarged in mutant embryos (Fig. 4E′,F′).

Although there is a wide range of ocular defects in Maz mutants, comparison of expression of specific markers of the CM showed that molecularly there are two different phenotypes. In both WT and in mildly affected embryos (type I), Otx1 is expressed in the entire CM (Fig. 4G,H). In contrast, in the most hypoplastic Maz^−/− eyes (type II), there is a clear reduction of Otx1 expression in the CM (Fig. 4I). Msx1, a gene that defines the posterior domain of the CM (Belanger et al., 2017; Marcucci et al., 2016), is expressed in the proximal region of the CM but is absent in the distal region of the peripheral retinas in WT embryos at E16.5 (Fig. 4J) and E18.5 (Fig. 4M). In the CM of Maz mutants, Msx1 expression is expanded to the distal domain of mildly affected mutants (Fig. 4K,N) and is absent in the most hypoplastic retinas (Fig. 4L). In the WT retina (Fig. 4O) and in the type I phenotype (Fig. 4P), Ccnd1 expression is strongest in the neural retina and absent in the CM. In the mutant type II embryos, the CM displays an abnormally high expression of Ccnd1 (Fig. 4Q).

PPH3 immunostaining in retinal sections of WT E18.5 embryos showed few or no proliferating PPH3-positive cells in the thinner CM, although there were many PPH3-positive cells in the neuroretina (Fig. 4R). In Maz mutants, no PPH3-positive cells were observed in the long ectopic expansion of the peripheral cup (Fig. 4S), whereas in severely affected mutants the characteristic thin region of the CM was absent and the periphery displayed the characteristic PPH3-positive cells of the neural retina (Fig. 4T).

We also evaluated the incorporation of 5-ethynyl-2′-deoxyuridine (EdU) in the CM of Maz-deficient retinas. We observed a significant decrease in the number of EdU-positive cells in the ciliary marginal zone (CMZ) of Maz^−/− retinas (Fig. 4U,V,Y). One explanation for the loss of EdU-positive cells in the CM might be selective cell death. To determine if mutant CM cells exhibit increased apoptosis, we performed TUNEL reaction on sectioned WT and mutant eyes. No accumulation of apoptotic TUNEL-positive cells was observed in the CM of E14.5 (data not shown) or E12.5 embryos. In contrast, there was a significant increase in TUNEL-positive cells observed in the ventral region of the Maz-deficient eye cups compared with those in the control (Fig. 4W,X,Z).

These results show that in the absence of Maz activity the patterning of the CM is affected. These findings also raised the possibility that during eye development Maz is regulating Wnt signaling, as both the Msx and Otx1 genes are known targets of this cascade (Liu et al., 2007; Willert et al., 2002).

Canonical Wnt signaling (Wnt/β-catenin) is essential for eye development. It is not only required for the transition of the optic vesicle to the optic cup, but also to promote the differentiation of the RPE and to maintain dorsal retinal identity (Fujimura, 2016; Hagglund et al., 2013). Additionally, canonical Wnt signaling is implicated in the development of the CM in several species, including mice and chickens (Cho and Cepko, 2006; Liu et al., 2003, 2007). Altered Wnt/β-catenin signaling results in numerous ocular malformations (Fuhrmann, 2008; Fujimura, 2016). Several Wnt signaling family members are active in the developing peripheral retina, such as Wnt2b, Fzd4 and Lef1 (Liu et al., 2003). Maz is a transcription factor that binds G-rich consensus sequences similar to those bound by WT1, a known Wnt pathway regulator (Kim et al., 2009). On the basis of this similarity, we hypothesized that Maz regulates Wnt pathway genes that are required for the correct patterning of the eye. Wnt/β-catenin activity is present in the neuroblastic layer, the RPE, the distal portion of the optic cup and the peripheral retina, where it is controlled by Sfrp1 and Sfrp2 (Esteve et al., 2011). To evaluate how the Wnt pathway is affected in the embryonic eyes of Maz mutants, a quantitative real-time PCR (qPCR) analysis was performed in Maz mutant and WT eyes using specific primers for several members of the Wnt cascade and its regulators, Sfrp1/2 (Cho and Cepko, 2006; Liu et al., 2003). Three genes, Wnt2b, Fzd4 and Sfrp2, were significantly upregulated in Maz mutants (Fig. 5A). Wnt2b, possibly using Fzd4 as a receptor, functions to establish the formation of the CM and also to keep a pool of CM progenitor cells in an undifferentiated state (Kubo et al., 2003, 2005). Sfrp2 belongs to a type of secreted regulatory proteins for Wnt signaling and has homology with the Wnt receptor Frizzled; Sfrp2 is generally accepted as an antagonist that binds and sequesters Wnt ligands to prevent signal activation. However, Sfrp1 and Sfrp2 are also required for the activation of Wnt/β-catenin signaling in the peripheral retina (Esteve et al., 2011). Therefore, our results suggest that Maz is a regulator of the Wnt pathway through the regulation of Wnt signaling members, including the modulator Sfrp2.

To confirm that Sfrp2 expression is increased in the Maz mutants, Sfrp2 expression was monitored by ISH (Fig. 5B-G). At E12.5, the Sfrp2 expression domain was predominantly located in the dorsal retina of WT embryos (Fig. 5B) and expanded in the Maz mutants (Fig. 5C,D). In E16.5 WT embryos, Sfrp2 expression is confined to the central retina. Although there is a gradient of expression, with higher expression close to the CM, the CM itself does not express the Sfrp2 gene (Fig. 5E). In some of the Maz mutants Sfrp2 expression is relatively normal compared with that in WT (Fig. 5F), but in some embryos Sfrp2 expression is expanded into the CM (Fig. 5G). These findings suggest that Maz possibly acts through inhibition of Sfrp2 expression in the CM.

To test the hypothesis that Maz normally suppresses Sfrp2 to promote Wnt activity in the CM, compound embryos with T-cell factor/lymphoid enhancer factor-β-galactosidase (TCF-Lef1-LacZ), a canonical Wnt/β-catenin reporter, were analyzed. These mice carry LacZ under the control of a minimal heat shock protein (HSP) promoter and six copies of the TCF-Lef1 responsive element (Liu et al., 2003). E10.5-E12.5 Maz^−/−/TCF-Lef1-LacZ+ and control littermates Maz^+/+/TCF-Lef1-LacZ+ were evaluated for LacZ activity. In control embryos at E10.5, TCF-Lef1 activation was detectable in the dorsal optic vesicle (Fig. 5H,H′), whereas at E12.5 activation was found in the neuroblast layer of the neuroretina, the CM and the RPE (Fig. 5J). In contrast, in E10.5 Maz mutant eyes (n=6), LacZ was almost absent in the dorsal part of the optic vesicle (Fig. 5I,I′), whereas in E12.5 Maz mutants (11 of 13) Wnt reporter activity was reduced in the whole retina, including the CM and RPE (Fig. 5K). In a few cases (2 of 13), Wnt activity in the eye was observed only in the RPE (Fig. 5L).

To further confirm loss of Wnt/β-catenin activity in the embryonic eye of Maz mutant eyes, we evaluated if the dorsoventral patterning of the retina was affected. Tbx5 (Fig. 5M) and Raldh1 (Fig. 5Q) are normally expressed in the dorsal domain of the optic vesicle at E10.5. The expression of both genes is preserved in Maz mutants, but while the expression of Raldh1 is relatively normal, the expression domain of Tbx5 appears thinner and elongated (Fig. 5N,R). At E12.5, the expression domain of Tbx5 is significantly reduced and thinner in comparison to the control embryos (Fig. 5O,P). This observation suggests that the induction of dorsal identity is not affected, but the activity of Maz is important for its maintenance. Furthermore, it was observed that expression of the ventral-specific gene Vax2 is extended into the dorsal domain, where normally it is absent (Fig. 5S-V). At E10.5, essentially all the cells in the mutant neuroretina adopt a ventral identity (Fig. 5V). These results suggest that in the Maz mutants there is a partial ventralization of the retina, as expression of the dorsal genes Tbx5 and Raldh1 is preserved. Therefore, the diminished Wnt/β-catenin reporter activity, together with the partial loss of dorsal identity in the retina of Maz mutants, supports the conclusion that Maz is a regulator of Wnt signaling and is essential for the robust formation of the D-V patterning and the correct specification of the CM. Our findings suggest that this regulation, at least in the peripheral retina, is primarily through suppression of Sfrp2.

In humans, ocular diseases have not been directly associated with MAZ mutations. Screening CNVs in patients with autism and cognitive disorders led to identification of the deletions and duplications of chromosome 16p11.2 as one of the most frequent genetic causes for autism spectrum disorders, schizophrenia and other neurodevelopmental disorders (Miller et al., 1993; Weiss et al., 2008). More recently, CNVs in this region were associated with urogenital (Haller et al., 2018) and microphthalmia, anophthalmia and coloboma (MAC) ocular malformations (Bardakjian et al., 2010; Hernando et al., 2002).

To understand the role of MAZ in human ocular morbidities, we searched the literature and cases reported in the Database of Chromosomal Imbalance and Phenotype in Humans Using Ensembl Resources (DECIPHER). This database catalogs potentially pathogenic genomic changes in patients. There were 580 patients harboring deletions and duplications in the region encompassing MAZ, with only 391 providing data on clinical features. From those cases with available clinical data, 44 patients exhibit various ophthalmic malformations. Excluding patients with refractive errors (hypermetropia, myopia and astigmatism) and/or strabismus, only 22 of 391 (5.6%) CNV carriers (eight duplications; 14 deletions) exhibit ocular disorders (Fig. S3). The frequency of ocular morbidities in this analysis is similar to that previously reported (2.1%) from a cohort of 357 carriers of the 16p11.2 BP4-BP5 deletion and 68 intrafamilial, non-carrier controls (Zufferey et al., 2012).

Owing to the fact that MAZ is a transcription factor that regulates expression of other transcription factors and signaling molecules, MAZ is a candidate contributor to ocular malformations associated with 16p11.2 CNVs. Interestingly, knockdown of the 16p11.2 homolog genes coro1a, maz or fam57b in zebrafish results in small eyes with protruding lenses (Blaker-Lee et al., 2012; Schmitt and Dowling, 1994). Thus, MAZ was considered a strong candidate gene for both its location and function.

We performed a retrospective analysis of ∼11,500 consecutive individuals undergoing clinical ES to determine if MAZ variants are associated with eye abnormalities. Importantly, this cohort is enriched for structural congenital defects. A total of 75 rare (with allelic frequency <0.01) or novel [never reported in The Genome Aggregation Database (gnomAD)] MAZ variants were identified. Some 38 patients with novel variants (six of them with eye phenotype) were excluded, as they had variants in another gene that probably caused their phenotype. A total of 37 non-synonymous coding single-nucleotide variants (SNVs) were identified and their frequencies were retrieved from gnomAD (Lek et al., 2016). As summarized in Table 1, seven of the 37 individuals with MAZ variants (19%) were affected with different eye disorders (including cone-rod dystrophy, rapid progressive vision loss, retinitis pigmentosa, congenital cataracts and optic nerve hypoplasia). The dominant phenotype observed in these MAZ variants is consistent with the ocular defects observed in the heterozygous Maz loss-of-function mice. We selected three MAZ variants predicted to be damaging and/or not tolerated, using the Polyphen2 and SIFT programs for structural analysis. We noted that variant A446V is located in the C-terminal polyalanine domain of the protein (Fig. 6A,B). Several studies have suggested that this domain might have a role in transcriptional repression in proteins like kruppel and engrailed in Drosophila and in the glucocorticoid receptor in humans (Han and Manley, 1993; Lavoie et al., 2003; Licht et al., 1990). Two additional variants, R276L and A404V, affect amino acids highly conserved from zebrafish to humans (Fig. 6C). The three-dimensional structure of the human WT, R276L and A404V variants of MAZ were modeled by the comparative protein method (Fig. 6D-F). Both variants are predicted to perturb local interactions and affect the function of the protein. The non-conservative substitution of R276L, involving a basic/aliphatic amino acid replacement, was predicted to have a deleterious effect on the protein structure (Fig. 6E). The semiconservative A404V sequence change involves exchanging an alanine, which is considered one of the best helix-forming residues, for a weaker valine within the zinc-finger domain and is therefore predicted to have a potential functional effect (Fig. 6F). The results suggest that these SNVs represent potentially pathogenic MAZ alleles, although functional studies will be necessary to discard the possibility of an occurrence by chance. In support, MAZ presented a probability of loss-of-function intolerance (pLI) of 0.94 and an o/e score of 0.07. These scores are indicative of a gene with strong intolerance to loss-of-function, where only 7% of the expected loss-of-function variants were observed. Together, these findings implicate rare and novel variants of MAZ as potential causative variants associated with eye abnormalities. Moreover, the functional analysis of the Maz mutant mouse model and the evidence that Maz regulates the Wnt/β-catenin pathway support the role for MAZ variants in human eye disease.

In this study, we have demonstrated that Maz function is essential for correct eye patterning during development. This was achieved in both a mouse model with a targeted deletion of Maz (Haller et al., 2018) and by genomic studies of human patients. We show that disruption of Maz results in microphthalmia, coloboma and a range of eye abnormalities with varying expressivity and incomplete penetrance in mice, and outline a molecular mechanism whereby Maz regulates Wnt activity in the developing eye. First, Maz was found to be necessary for the activation of the Wnt TCF-Lef1 reporter. Second, we demonstrated that genetic deletion of Maz leads to defective dorsoventral patterning in the optic vesicle and abnormal differentiation of the CM, which involved the Wnt pathway. Our observations suggest that the canonical Wnt/β-catenin pathway is normally upregulated by Maz in the eye. However, disruption of Wnt/β-catenin activity in Maz mutants cannot be explained by attenuation of Wnt signaling members, as expression of the key factor establishing the dorsal retina Lrp6 is not changed and Wnt2b and Fzd4 (the candidates to promote CM formation and maintenance of the progenitors) are significantly upregulated in the CM of Maz mutants. Of note, increased Wnt2/Fzd4 expression was insufficient to activate the Wnt reporter and induce a correct CM in the Maz mutant. This might reflect an absent reinforcing synergistic signal or, alternatively, the presence of an inhibitor of the inducing effect of Wnt2b/Fzd4 overexpression. Our data implicated the latter, as Sfrp2 overexpression can effectively inhibit Wnt signaling both in vivo and in vitro, its overexpression in the Maz mutant might be sufficient to abrogate Wnt activity. The enhanced expression of Wnt2b and Fzd4 in the mutant indicates that Maz function is required for the correct regulation of these genes and might involve a negative feedback, potentially explaining the increased Wnt2b/Fzd4 expression of these genes in the Maz mutants.

Sfrps modulate the Wnt cascade by their interaction with both Wnt ligands and their receptors (Esteve and Bovolenta, 2010). Sfrp2 is not expressed in the CM, but it is expressed in the undifferentiated retinal neuroepithelium: its expression is progressively downregulated in the central retina forming a central (low) to peripheral (high) gradient. Sfrp1 is mainly expressed in the RPE and the CM (Marcos et al., 2015). In the CM, Sfrp2 cooperates with Sfrp1 to enhance activity of Wnt/β-catenin signaling, possibly through the promotion of interactions of Wnt ligands with their Fzd receptors (Esteve et al., 2011; Lopez-Rios et al., 2008).

Loss of Maz function results in an upregulation of Sfrp2 and a variable CM phenotype. Some mutants showed a reduction of the CM markers, whereas others had an abnormal distoproximal patterning of the CM. This observation indicated that Maz, possibly through its influence on Wnt signaling, was essential for correct establishment of the neuroretina-CM boundary as well as for patterning of the CM. Several studies indicate that Wnt/β-catenin signaling activity in the peripheral retina is controlled by Sfrp1/Sfrp2 to establish the border between the peripheral and central neural retina. Eyes of Sfrp1^−/− Sfrp2^−/− compound mutants lack the retinal periphery domain and show an increased number of retinal ganglion cells (Esteve et al., 2011).

Based on the proposed role of Sfrps in the CM and increased expression of Sfrp2 in the Maz mutants, we expected Sfrp2 expression to upregulate the TCF-Lef1-LacZ reporter in the CM. Paradoxically, the activity of the Wnt reporter was impaired in the CM of Maz mutants. One possible explanation for this discrepancy is a biphasic mode of regulation in which an increase of Sfrp2 at the border of the CM initially favors Wnt/β-catenin signaling by bringing ligand and receptor together, but once Sfrp2 reaches a specific threshold it inhibits this activity. This idea is supported by the observation that forced expression of Sfrp1 in the wing imaginal disc of Drosophila impairs spreading of the Wingless (Drosophila Wnt homolog) gradient. This impairment results in an inhibition of expression of Wingless target genes that require a high level of Wingless and an activation of expression of those genes that require low levels of Wingless (Esteve et al., 2011).

Alternatively, the variations in the CM phenotype (including both expansion and reduction of the CM) might be explained by Maz regulation of other unidentified spatio-temporal factors regulating Wnt signaling to control multiple aspects of peripheral eye differentiation. For example, tsukushi (TSK; also known as TSKU), a group of soluble molecules belonging to the small leucine-rich proteoglycan (SLRP) family, inhibit Wnt2b activity and repress the induction of peripheral eye character by quenching Fzd4 activation, thereby regulating the size of peripheral structures and especially the CB (Ohta et al., 2011). Another possibility is that Maz, which is also expressed in the periocular surface ectoderm and lens epithelium, might affect the Wnt signaling from the surface ectoderm to the retina that promotes the morphogenesis of the CM (Carpenter et al., 2015).

Reciprocal interactions to establish eye tissue boundaries occur in vitro, showing a remarkable plasticity between neuroretina-RPE fates during eye development (Kuwahara et al., 2015; Rowan et al., 2004). In optic cups derived from mouse embryonic stems cells, aggregate interactions between the neuroretina and RPE promote self-organization of the CM. This self-organization is induced by Wnt agonists and is reversed by Wnt inhibitors. However, some heterogeneity in the responsiveness to the reversal of this trigger was also observed. In some aggregates, RPE and the neuroretina were co-generated and CM was self-organized between them. In some, the aggregates became entirely neuroretina, and yet others were resistant to the reversal trigger and remained as RPE. These studies suggested that neuroretina-RPE self-organization follows the ‘bi-stability’ mechanism in the regulatory network (Sasai, 2013). This occurs when state A and state B inhibit each other, whereas each state activates itself. Under these circumstances, the system tends to give a black or white, but not a gray, outcome. This mechanism produces a sharp boundary of domains similar to a toggle switch. In this context, Maz as a regulator of Wnt activity might have an important role in the specification of the CM by excluding the expression of Sfrp2 in the CM, which might involve activation, autoregulatory loops and mutual inhibition of Wnt2b and Fzd4 in the peripheral retina. Small differences in the expression levels of these key genes are likely to have a pivotal role in the specification of either the neuroretina or CM in the periphery of retinal epithelium.

Additionally, the expansion of the proximal Msx1 expression domain in the CM observed in some of the Maz mutants suggested that the spatial patterning of the mouse CM in the distal and proximal domains is also altered (Belanger et al., 2017; Kuwahara et al., 2015; Marcucci et al., 2016). It is therefore possible that Maz also regulates the proximodistal patterning of CM. As Msx genes are targets of Wnt signals (Willert et al., 2002), and β-catenin induces expression of both Msx1 and Otx1, it is possible that this also takes place through the regulation of the canonical Wnt cascade.

On the basis of our observations, we propose that Maz has an important role as a regulator of Wnt activity in the promotion of CM fate. In this working model (Fig. 7), Maz regulates Wnt/β-catenin activity in the CM by limiting expression of Sfrp2 to the central retina. Modulation of Sfrp2 expression in the CM might be crucial, not only for the determination of cell fate in the border between peripheral and central retina, but also for the correct proximodistal patterning of the CM. However, the involvement of other Maz mechanisms in the context of eye development, in addition to the canonical Wnt signal, cannot be ruled out. Indeed, many studies of Maz in other settings have illustrated other potential roles, such as competing or interacting with other transcription factors (Sohl et al., 2010; Morii et al., 2002; Lee et al., 2016). Of particular note in the eye, developmental vascular regression is regulated by Wnt/β-catenin and MYC-CDKN1A (Nayak et al., 2018; Rao et al., 2013). Interestingly, Myc is transcriptionally repressed by Maz, suggesting that it might have additional roles in the regulation of the Wnt cascade (Izzo et al., 1999).

On the basis of our studies showing a gene-dose-dependent regulatory role of Maz in mouse eye morphogenesis, we evaluated MAZ as a candidate gene for human eye disease. Several cases of CNVs associated with anophthalmia/microphthalmia were reported in the literature in the 16p11.2 chromosomal region, where MAZ is located (Bardakjian et al., 2010; Bardakjian and Schneider, 2011; Hernando et al., 2002). We found that heterozygous CNV deletions and duplications of chromosomal region 16p11.2 are associated with ophthalmic comorbidities, including microphthalmia, coloboma, cataracts and others, in nearly 6% of symptomatic individuals carrying 16p11.2 CNVs in DECIPHER. These results are consistent with previously reported associations in the literature (D'Angelo et al., 2016; Zufferey et al., 2012).

Although the 16p11.2 CNV locus contains more than 25 genes and a simple phenotype-genotype correlation could not be inferred, we identified seven genetic MAZ variants present in patients with variable eye anomalies. Two of these, R276L and A404V, are predicted to produce deleterious structural and functional modifications of the MAZ protein. Given that non-penetrance has been observed for the variant P76Q in the mother of the affected individual, it is possible that this individual is non-penetrant or the phenotype is milder. The non-penetrance of such a variant is similar to the incomplete penetrance observed in Maz mutants and might reflect effects of the background or compensation by paralogs. Similar non-penetrance has been observed in families with pathogenic variants of FZD5, a gene in the WNT pathway that causes coloboma (Liu et al., 2016). However, additional functional studies are necessary to understand the significance of these variants.

In summary, our study highlights MAZ as an important dose-sensitive gene responsible for abnormal ocular development in 16p11.2 syndrome. This is supported by an in vivo loss-of-function mouse model, showing that Maz is required for eye morphogenesis. The identification of MAZ/Maz function on eye development in both human and mouse broadens opportunities to elucidate the disease mechanisms and treatment for ocular malformations.

All animal protocols were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) at Baylor College of Medicine, and the experiments were performed in adherence to the National Institutes of Health Guidelines on the Use of Laboratory Animals. Generation of Maz knockout mice (Maz^−/−) was previously described (Haller et al., 2018) and produced using the CRISPR-Cas9 methodology. Two single-guide RNAs flanking Maz exon 2 to 4 were used. Cas9 recombination resulted in a deletion of the zinc-finger C2H2 domains, producing a truncated protein unable to bind DNA. Animals were maintained in a C57BL/6J background. Maz^+/− crosses were used to generate Maz^−/− embryos, as the deletion of Maz resulted in perinatal lethality.

TCF-Lef1-LacZ reporter mice were obtained from The Jackson Laboratory. Heterozygous TCF-Lef1-LacZ mice with a CD1 background were crossed back two generations to Maz^+/− mice to generate Maz^+/−/TCF-Lef1-LacZ mice with a C57BL/6J and CD1 mixed background. Heterozygous Maz^+/−/TCF-Lef1-LacZ mice were crossed to obtain Maz^−/−/TCF-Lef1-LacZ mice and control Maz^+/+/TCF-Lef1-LacZ mice, respectively.

Embryos were fixed in 10% formalin, dehydrated in a progressive graded series of ethanol and embedded in paraffin. Sectioned heads were deparaffinized, rehydrated in PBS and treated with antigen retrieval solution (10 mM sodium citrate, pH 6) in a microwave for 20 min. After a rinse with PBS, sections were circled with a PAP pen and incubated with blocking buffer (5% bovine serum albumin/0.5% Tween-20 in PBS) for 1 h at room temperature in a humidifier chamber. Next, sections were incubated overnight with one of two antibodies: phosphohistone H3 (1:500, Upstate) or Pax6 (1:200, DSHB). The next day, slides were incubated with the corresponding rabbit Alexa Fluor 594 or mouse Alexa Fluor 488 secondary antibodies for 1 h and followed by three washes. Sections were mounted with anti-fade mounting medium and visualized by fluorescence microscopy.

Incorporation of EdU was analyzed in six embryos by genotype and each eye analyzed separately (n=12), as previously reported (Mead and Lefebvre, 2014) with some modifications. The Click-iT^® EdU Alexa Fluor^® 488 Imaging Kit (Invitrogen) was used, which contains all components needed to label DNA-synthesizing cells and to detect EdU incorporated into DNA. A pregnant mouse was injected intraperitoneally with 10 µl EdU solution/g body weight. After 2 h of EdU labeling, heads of embryos were fixed with 10% buffered formalin for 2 h at room temperature. The heads were then dehydrated and embedded in paraffin. Sections (7-10 µm) were used to perform all the instructed steps of the assay in the dark. For quantification of EdU-positive cells in the CM, cells were counted on three sections through the medial plane of the lens for a total of six animals per genotype.

Cell death was detected using the In situ Cell Death Detection Fluorescein Kit (Sigma-Aldrich) in four embryos. Manufacturer's instructions were followed for the labeling reaction on eye sections. After permeabilization and labeling with TUNEL reaction mixture, sections were analyzed by fluorescence microscopy.

Mouse embryos (E9.5-E12.5) were dissected and fixed overnight with 4% paraformaldehyde. Embryos were then dehydrated and re-hydrated in methanol series and treated with 10 µg/mL of proteinase K for 10 min. ISH was performed overnight with digoxigenin (DIG) antisense or sense probes. Post-hybridization washes and DIG antibody incubation were performed according to a standard protocol (Wilkinson, 1992). Signal was visualized with BM purple (Roche).

Embryos (E9.5-E16.5) were fixed in 4% paraformaldehyde phosphate buffer overnight. Tissue was rinsed in PBS and embedded directly in Optical Cutting Temperature (OCT) compound (VWR Chemicals) or progressively dehydrated in ethanol and embedded in paraffin. Paraffin sections from embryos (E14.5-E18.5) were hybridized overnight with RNA DIG-labeled probes at 65°C. The next day, sections were incubated in 5× SSC with RNase (1 µg/mL) at 37°C, covered with blocking buffer (BMB, Roche) and incubated with alkaline-phosphatase-coupled DIG antibody (1:5000) overnight. Sections were washed and signal was visualized with BM purple (Roche). Cryostat sections were processed for ISH using a standard protocol (Blackshaw, 2013).

RNA from E16.5 eyes was extracted using Qiagen RNeasy mini kit. Complementary DNA (cDNA) was obtained using High-Capacity RNA to cDNA Kit (Applied Biosystems). qPCR reactions were performed in 96-well plates with the SYBR Green PCR Master Mix. Reactions were performed in triplicate, using three independent samples. Samples were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Relative expression software tool (REST) analysis (Pfaffl et al., 2002) was performed to determine level of significance.

To detect X-gal staining, embryonic heads (E12.5-E16.5) or eyes (E17.5) were fixed in a 4% buffered paraformaldehyde (PFA) solution for 10 min on ice and rinsed twice in PBS for 5 min. Fixed samples were incubated in a staining solution (5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 2 mM MgCl₂, 0.01% sodium deoxycholate and 0.02% nonident P-40 plus 1 mg/mL X-gal in PBS) overnight. The next morning, samples were rinsed in PBS and post-fixed in 4% PFA before paraffin embedding and sectioning.

DECIPHER database (https://decipher.sanger.ac.uk/) was searched to identify patients with 16p11.2 CNVs harboring eye phenotypes.

For the ES cohort, data from ∼11,500 individuals undergoing clinical ES at Baylor Genetics Laboratory were examined for all called variants in MAZ (Yang et al., 2014).

Results comparing two groups were analyzed either by Student's t-test or by Mann–Whitney U-test. P-values are indicated in the figure legends.

We thank Drs Valerie Wallace, Paola Bovolenta and E. Levin for the gift of plasmids. The Pax6 and Islet 1 antibodies were obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242, USA. Our thanks are extended to Drs Valeria Canto-Soler, Adlai Nelson, Daniel Galvan and Paul Overbeek for critical reading of the manuscript. We also thank the BCM in situ Hybridization Core for their expertise and services. The authors would like to thank the Genome Aggregation Database (gnomAD) and the groups that provided exome and genome variant data to this resource. A full list of contributing groups can be found at https://gnomad.broadinstitute.org/about.