Identification of missense MAB21L1 variants in microphthalmia and aniridia

Abstract Microphthalmia, coloboma, and aniridia are congenital ocular phenotypes with a strong genetic component but often unknown cause. We present a likely causative novel variant in MAB21L1, c.152G>T p.(Arg51Leu), in two family members with microphthalmia and aniridia, as well as novel or rare compound heterozygous variants of uncertain significance, c.184C>T p.(Arg62Cys)/c.‐68T>C, and c.658G>C p.(Gly220Arg)/c.*529A>G, in two additional probands with microphthalmia, coloboma and/or cataracts. All variants were predicted as damaging by in silico programs. In vitro studies of coding variants revealed normal subcellular localization but variable stability for the corresponding mutant proteins. In vivo complementation assays using the zebrafish mab21l2 Q48Sfs*5 loss‐of‐function line demonstrated that though overexpression of wild‐type MAB21L1 messenger RNA (mRNA) compensated for the loss of mab21l2, none of the coding variant mRNAs produced a statistically significant rescue, with p.(Arg51Leu) showing the highest degree of functional deficiency. Dominant variants in a close homolog of MAB21L1, MAB21L2, have been associated with microphthalmia and/or coloboma and repeatedly involved the same Arg51 residue, further supporting its pathogenicity. The possible role of p.(Arg62Cys) and p.(Gly220Arg) in microphthalmia is similarly supported by the observed functional defects, with or without an additional impact from noncoding MAB21L1 variants identified in each patient. This study suggests a broader spectrum of MAB21L1‐associated disease.


| INTRODUCTION
Developmental ocular disorders have complex genetic etiologies due to the intricate and tightly controlled genetic networks involved in eye development (Skalicky et al., 2013). Microphthalmia, anophthalmia, and coloboma (MAC) are rare congenital malformations of the eye involving a small eye, absence of an eye, and gap in ocular structures, respectively (Gregory-Evans et al., 2004;Verma & Fitzpatrick, 2007). Over 80 genes have been published in association with MAC phenotypes . However, about 50% of patients lack a confirmed genetic diagnosis, suggesting novel genes have yet to be discovered (Plaisancie et al., 2016).
Aniridia is a panocular disorder with its primary feature being a partial or complete absence of the iris, but also including lens opacities, glaucoma, keratopathy, foveal and optic nerve hypoplasia, strabismus, ptosis, and fibrosis syndrome (Hall et al., 2019;Hingorani et al., 2012;Lim et al., 2017). Up to 90% of cases with aniridia can be explained by loss-of-function mutations in the PAX6 gene (Hingorani et al., 2012); rarely, disruption of FOXC1, PITX2, and other genes have been identified as causative (Hall et al., 2019;Hingorani et al., 2012). However, there is a small portion of the aniridia population that remains genetically unexplained.
The Male-Abnormal 21-Like gene MAB21L2 is a recently identified factor involved in human MAC-spectrum disorders, where both dominant and recessive missense alleles have been recognized as causative in eight unrelated families (Aubert-Mucca et al., 2020;Deml et al., 2015;Horn et al., 2015;Patel et al., 2018;Rainger et al., 2014). In four out of seven dominant families, the pathogenic variant is a missense allele affecting residue 51 of the resulting protein. Similarly, a mouse Mab21l2 model with a heterozygous p.(Arg51Cys) mutant allele (identical to two of the affected human patients (Horn et al., 2015;Rainger et al., 2014), further demonstrated the importance of this residue, resulting in defects in early ocular development including rudimentary and mispositioned optic cup, undetectable optic stalk, abnormalities of the retinal pigment epithelium, and failure to induce a lens placode (Tsang et al., 2018).
MAB21L1, a closely related family member to MAB21L2, has also been implicated in human disease. Homozygous MAB21L1 variants have been reported in six unrelated families exhibiting cerebellooculo-facio-genital syndrome, with five of these variants resulting in premature truncation of the protein (Bruel et al., 2017;Rad et al., 2019). Ocular abnormalities included corneal dystrophy/opacities, nystagmus, strabismus, dry eye, pigment granularity, retinal degeneration, optic atrophy, buphthalmos, and cataracts (Bruel et al., 2017;Rad et al., 2019). A null mouse model for Mab21l1 likewise exhibits embryonic ocular defects, though more severe than the published human phenotype. Abnormalities include microphthalmia, malformed retina, and retinal pigment epithelium, along with aphakia, thickened cornea, and absent iris (Yamada et al., 2003).
The precise protein function(s) of the MAB21L family is unknown. A possible role in transcriptional regulation has been suggested (Baldessari et al., 2004) with nuclear localization (Mariani et al., 1999) and a mild affinity for nucleic acid shown in vitro (de Oliveira Mann et al., 2016;Rainger et al., 2014).
Here, we report three families with unique MAB21L1 variants exhibiting ocular phenotypes including MAC-spectrum and aniridia.
Functional studies of the proteins associated with coding variants revealed differences from wild-type MAB21L1. Thus, this study suggests a phenotypic expansion for MAB21L1-associated human disease.

| Editorial policies and ethical considerations
Human studies conformed to the US Federal Policy for the Protection of Human Subjects and were approved by the Children's Hospital of Wisconsin Institutional Review Board and the Sydney Children's Hospitals Network Research Ethics Committee, with written informed consent obtained from all participating individuals and/or their legal representatives.

| Human DNA screening and in silico variant analyses
The MAB21L1 variant in Individual 1 was initially identified through clinical exome sequencing using the previously published protocol (Guillen Sacoto et al., 2020) and matched to the study through the Matchmaker Exchange GeneMatcher node (Philippakis et al., 2015;Sobreira et al., 2015) followed by enrollment and research exome sequencing and analysis; the MAB21L1 variants in Individuals 2 and 3 were identified through research exome sequencing and analysis using previously described methods (Deml et al., 2016;Ma et al., 2020). Variants in known ocular genes were ruled out in all three families. All variants were confirmed via Sanger sequencing of the MAB21L1 coding region by amplification of a 1405-base pair (bp) product using the flanking primers F-5ʹ-CCGAAAGGCATTTTT GATCC-3ʹ, R-5ʹ-TCCGCTTCCCCTACTTTTTC-3ʹ, and also internal primers F-5ʹ-AGATCACGCCGGCCTTTA-3ʹ, R-5ʹ-ACCCAGGCGTCG CTCTC-3ʹ. Polymerase chain reaction (PCR) amplicons were sequenced as previously described (Deml et al., 2015) or through Functional Biosciences™ DNA Sequencing Services. Parental samples were analyzed using the same protocol. To determine relative positions of variants identified in Individual 2, the amplified 1405-bp product was cloned into a pCR®II-TOPO® plasmid, which was followed by sequencing of 32 independent clones; 18 clones contained c.184C>T p.(Arg62Cys) allele and wild-type 5ʹ-UTR (untranslated region) sequence and 14 clones contained c.-68T>C 5ʹ-UTR variant and wild-type coding region sequence indicating their trans configuration.
Additional in silico analyses were performed for the MAB21L1 UTR variants (c.-68T>C and c.*529A>G) to assess possible functional impacts. To determine changes to minimum free energy and RNA secondary structure, both 5ʹand 3ʹ-UTR DNA sequences carrying wild-type or variant alleles were submitted to the RNAfold 2.4.17 Webserver, accessed through the Vienna RNA Websuite 2.0 (Lorenz et al., 2011). To assess changes to microRNA target prediction, wildtype, and the variant 3ʹ-UTR sequence was submitted to Micro-SNiPer (Barenboim et al., 2010) with a minimum seed length constraint of seven base pairs. In addition, the PolymiRTS Database 3.0, a database of human SNPs affecting predicted microRNA (miRNA) target sites (Bhattacharya et al., 2014), was searched for the 3ʹ-UTR variant, rs1775984, to determine overlapping predicted miRNA sites.
PolymiRTS calculated the strength of the predicted miRNA site and provided a conservation score, which was determined based on the number of vertebrate genomes in which the miRNA site is present, and context+ score change (a ranking of miRNA target predictions) (Garcia et al., 2011). To determine variant effects on RNA binding protein (RBP) binding sites, wild-type and the variant 5ʹ-and 3ʹ-UTR sequences were submitted to RBPmap version 1.1 (Paz et al., 2014) to identify RBP motifs. Predicted motifs containing the affected nucleotide were assessed for lost/gained interactions between wildtype or mutant sequence and RBPs.
In silico protein modeling was executed for the MAB21L1 wild- Immunofluorescence experiments were used to determine protein localization. The transfection protocol was as described above with 7.5 μg wild-type and mutant N-terminally tagged MAB21L1 constructs into HLE-B3 cells. Cells were fixed with 1:1 methanol/ acetone permeabilized with 1% Triton X-100 and blocked with 10% donkey serum in 1× PBS, followed by overnight incubation with 1:100 mouse anti-FLAG primary antibody at 4°C. The next day, cells were incubated with 1:1000 donkey anti-mouse Alexa Fluor 488 secondary antibody (Thermo Fisher Scientific) and stained with In-vitrogen™ 4ʹ,6-diamidino-2-phenylindole, dihydrochloride (DAPI) (Thermo Fisher Scientific).

| Animal care and use
The care and use of zebrafish have been approved by the Institu- . Developmental stages were determined as previously described by hours post fertilization (hpf) and morphology (Kimmel et al., 1995). p.(Gln48Serfs*5) heterozygous cross (Deml et al., 2015). At 24 hpf, when the mutant phenotype is clearly observable in homozygous mab21l2 Q48Sfs*5 embryos, injected offspring were examined for the presence or absence of the previously described ocular phenotype, and the proportion of normal embryos was determined. Graphs were created using GraphPad Prism 9. Statistical significance was determined using an unpaired samples t-test and a p < .05. ( Figure 1 and Table 1). No systemic abnormalities were noted. The amino acid substitution was predicted to be likely damaging by 4/5 prediction programs and had REVEL (0.585) and CADD PHRED (31) scores suggesting pathogenicity (Table 1). Again, the arginine residue at position 62 was found to be highly conserved with a GERP++RS score of 5.66 and a PhyloP score of 9.87 (Table 1) (Figure 1 and Table 1). The amino acid substitution was predicted to be likely damaging by 4/5 programs and had a CADD PHRED score of 26.2, suggesting pathogenicity (Table 1); the REVEL score was 0.45 (~25% of pathogenic variants have a score below 0.5 (Ioannidis et al., 2016)). The glycine residue at position 220 was found to be highly conserved with a GERP++RS score of 5.76 and a PhyloP score of 7.80. The variant was found to be present in 1/112,548 alleles in the ethnically matched European

| In silico analysis of coding and noncoding alleles
For in silico modeling of the wild-type and mutant MAB21L1 proteins, iTASSER and Pymol software were utilized ( Figure S1). The wild-type protein structure is described as two-lobed, containing an N-terminal and C-terminal lobe, with an α-helix spine (α1) spanning T A B L E 1 (Continued) kcal/mol for c.*529A>G) in comparison to corresponding wild-type sequences. As miRNAs typically bind to sites within the 3ʹ-UTR, potential target sites were assessed in wild-type and the variant 3ʹ-UTR sequence (Grimson et al., 2007). This revealed several target sites that were either disrupted or created by the c.*529A>G variant (Table S1). In addition, several RBP motifs in both the 5ʹ-and 3ʹ-UTR were predicted to be affected: c.-68T>C disrupted a potential SRSF3 motif and created an RBM6 motif; c.*529A>G disrupted predicted motifs for HNRNPL, IGF2BP2, RBM41, and SRSF3 RNA binding factors and generated a new sequence expected to bind SRSF5 (Table S2). These RNA-binding proteins (or their closely related family members) have demonstrated important roles in RNAregulation (including splicing, export, stability, polyadenylation, and translation) (Cao et al., 2018;Jain et al., 2012;Oberdoerffer et al., 2008;Rothrock et al., 2005;Sutherland et al., 2005;Twyffels et al., 2011;Zhong et al., 2009;Zhou et al., 2020); in addition, SRSF3 and HNRNPL showed associations with ocular disease, mainly glaucoma (Jain et al., 2012;Schmitt et al., 2020). Finally, in silico evaluations using FATHMM-MKL predicted deleterious effects, and GERP++RS and PhyloP indicated nucleotide conservation for both 5ʹ-and 3ʹ-UTRs (Table 1 and Figure S2). Therefore, though the mechanisms remain unclear, it is possible the noncoding variants in Individuals 2 and 3 could be contributing to disease through disruption of miRNA and RBP target sites, and thus, regulatory activities upon the MAB21L1 mRNA.

| Expression and localization of MAB21L1 wild-type and mutant proteins
The stability and localization of wild-type and mutant proteins were In accordance with Mendel's principles, heterozygous mab21l2 Q48Sfs*5 parents are expected to produce 75% phenotypically normal embryos comprising a mix of heterozygous (p.Gln48Serfs*5/+) (50%) and wild-type (+/+) (25%) genotypes and 25% homozygous (p.Gln48Serfs*5/p.Gln48Serfs*5) fish that exhibit the ocular phenotype (Deml et al., 2015). Consistent with this, 77.56% ± 0.27% ( (Bruel et al., 2017;Rad et al., 2019). The data presented in this report suggest that heterozygous missense variants in MAB21L1 may also be disease-causing and extend the associated disease spectrum. which is involved in the recognition of cytosolic nucleic acid and subsequent production of 2ʹ,3ʹ-cGAMP (Ablasser et al., 2013;Gao et al., 2013). Notably, all three affected residues, Arg51, Arg62, and Further evidence via identification of other similar families will be needed.
The possible connection of MAB21L1 with MAC-spectrum in humans is plausible given the broad expression of this factor in the developing ocular structures in animal models. A Mab21l1 null mouse displays a severe developmental ocular phenotype consistent with MAC-spectrum (Yamada et al., 2003). Homozygous embryos fail to form a lens vesicle and exhibit aphakia, malformed retina, and abnormally thick cornea, culminating in severe microphthalmia, disorganized retinal lamination, and highly abnormal anterior structures including absent lens and iris in adults (Yamada et al., 2003). Expression of Mab21l1 overlaps the developmental pattern of its close homolog, Mab21l2 (Yamada et al., 2003;Yamada et al., 2004), whose deficiency is also associated with severe ocular defects in mouse (Tsang et al., 2018;Yamada et al., 2004) and zebrafish models ( (Rainger et al., 2014).
The link between MAB21L1 and aniridia is also consistent with prior knowledge. The most common cause of aniridia is pathogenic variants in the PAX6 gene, accounting for up to 90% of cases (Hingorani et al., 2012), with FOXC1, PITX2, and a few other genes explaining some of the remaining cases (Hall et al., 2019;Hingorani et al., 2012) but still leaving about 5%-10% of aniridia genetically unexplained. Interestingly, in Caenorhabditis elegans, a mab-18 mutant (a PAX6 orthologue) was found to have a very similar phenotype to a mab-21 mutant (a MAB21L orthologue), both affecting sensory ray formation of the tail (Baird et al., 1991). Furthermore, both Mab21l1 and Mab21l2 have previously been suggested as downstream targets of Pax6 in mice. In small-eye (sey) homozygous embryos, in situ hybridization identified a significant reduction in Mab21l1 expression in both the surface ectoderm and optic vesicles during ocular development, whereas expression of Mab21l2 in the same tissues was unchanged (Yamada et al., 2003); at the same time, in heterozygous Pax6 lacZ/+ mice (St-Onge et al., 1997), Mab21l2 was found to be upregulated in lens tissue implying that Mab21l2 expression may be normally repressed via Pax6 in the lens (Wolf et al., 2009). Pax6 binding sites were identified in the regulatory regions of Mab21l2 (Wolf et al., 2009), which further corroborates this interaction and suggests a direct effect. Furthermore, though variants in PAX6 are typically connected with aniridia, a small number of bilateral and unilateral microphthalmia/coloboma cases have been identified, with or without aniridia . The missense variants reported in this manuscript suggest a role for MAB21L1 in microphthalmia, aniridia, and coloboma in humans, similar to the PAX6 spectrum. Thus, this study provides additional support for the likely involvement of MAB21L1 and PAX6 in the same pathway. Further genetic screening for MAB21L1 variants in a wide spectrum of ocular disorders will help to define its role in human eye development and disease.