PAX6 encodes a highly conserved homeodomain-containing transcription factor that plays pivotal roles in normal ocular and neural development (Cvekl and Callaerts, 2017). Dominant PAX6 mutations lead to a spectrum of ocular developmental anomalies (ODAs) depending on the mutation type and gene dosage (van Heyningen and Williamson, 2002). PAX6 haploinsufficiency, which results from loss-of-function variants or 11p13 microdeletions involving this gene or their 5′ regulatory regions, is the major cause of congenital aniridia (MIM# 106210) (Hingorani et al., 2012). By contrast, missense mutations usually exhibit a moderate impact on PAX6 functionality and are often associated with some atypical PAX6-associated phenotypes, such as mild forms of iris coloboma or isolated foveal hypoplasia, or more severe phenotypes of Peter’s anomaly and microphthalmia (Hanson et al., 1999; Azuma et al., 2003; Nallathambi et al., 2006; Jia et al., 2010; Thomas et al., 2014). However, no obvious genotype-phenotype correlations have been established to date (Gupta et al., 1998; Gronskov et al., 1999; Dubey et al., 2015; Sannan et al., 2017).

Aniridia, characterized by the incomplete development of the iris and fovea, is the most frequent PAX6-related condition, with a worldwide incidence of 1:50,000–100,000 births (Hingorani et al., 2012). Patients exhibit photophobia, low visual acuity, and nystagmus. Phenotypic variability is commonly observed (Lee and Colby, 2013). Iris hypoplasia may manifest as a complete absence of the iris (or “aniridia”) but also may present with minor structural iris defects, leading to atypical iris coloboma, iris holes, or stromal hypoplasia (Skeens et al., 2011), symptoms which can be only recognizable on slit-lamp examination. Additionally, a wide range of abnormalities in the cornea, anterior chamber, lens, and optic nerve have been reported conferring a higher risk in patients for secondary glaucoma, cataracts, and aniridia-associated keratopathy (AAK), which further worsen the visual outcome (Lee and Colby, 2013). Inter-familial and intra-familial variability on disease onset and severity of these secondary symptoms manifesting as congenital, childhood, or even adult forms are also observed (Hingorani et al., 2012; Vasilyeva et al., 2017). However, the molecular mechanisms underlying the variable expressivity of the PAX6 mutations have yet to be elucidated. Identifying and understanding the genetic mechanisms that affect the severity of the disease is essential to provide a more accurate diagnosis and better clinical management of PAX6-associated disorders.

Somatic or gonosomal mosaicism might explain part of the phenotypic variability and/or reduced penetrance in aniridia, similarly to that reported in other dominant ocular disorders (Faivre et al., 2006; Beryozkin et al., 2016). Post-zygotic variants (PZV), arising as errors in DNA replication at early embryonic stages, have recently been revealed as a novel source of de novo and somatic mosaic variants thanks to the advent of more sensitive genotyping technologies (Acuna-Hidalgo et al., 2015, 2016). The developmental stage at which PZVs arise has a major influence on their frequency and distribution in affected tissues and thus on phenotypic expressivity and the recurrence risk in offspring (Acuna-Hidalgo et al., 2016). Interestingly, recent studies have showed that PZVs could account for up to 10% of rare neurodevelopmental disorders such as intellectual disability, epilepsy, and autism (Acuna-Hidalgo et al., 2015; Stosser et al., 2017; Myers et al., 2018); therefore, mosaicism may be underestimated in sporadic cases of other developmental diseases.

In aniridia, up to two-thirds of patients are sporadic (Netland et al., 2011; Lee and Colby, 2013) and thought to carry de novo mutations that are generally not detected in parental blood samples. Nevertheless, germline or even low-level somatic mosaicism in one of the parents cannot be ruled out, as multiple tissues are not usually tested during genetic screening. Germline mosaicism has long been postulated in PAX6-related disorders to explain the disease co-occurrence in several affected siblings with no family history of eye disorders (Reed and Falls, 1955; Gronskov et al., 1999; Deml et al., 2016; Riera et al., 2017). However, to our knowledge, PAX6 mosaicism has never been reliably confirmed in a suspected family, mainly due to technical limitations in genetic testing and/or constraints on the availability of germinal or somatic DNA samples other than blood. To date, only a small number of exceptional cases of somatic mosaicism for 11p13 microdeletions have been reported in aniridia (Robinson et al., 2008; Erez et al., 2010; Huynh et al., 2017).

Here, we identify parental mosaicism in three families with PAX6-related ODAs in which gonosomal PZVs have been accurately assessed through quantitative droplet-digital PCR (ddPCR). Detection of mosaicism was very relevant to these families in terms of genetic counseling, which allowed for more accurate determination of the risk of recurrence in future offspring. Remarkably, our findings also highlight the inter-generational variable expressivity in these families, which seems to be clearly explained by the presence of parental mosaic PAX6 mutations as well as by the further transmission of these mutations to the offspring. Thus, our work confirms that mosaicism is an underestimated cause of phenotypic variability and disease recurrence in aniridia and other PAX6-related ODAs.

Phenotypic variability is a well-established phenomenon for PAX6 defects (Hanson et al., 1999; Hingorani et al., 2012) that could be hypothetically attributable to genetic or epigenetic modifiers, pleiotropy, mosaicism, and/or gene dosage (Schedl et al., 1996; Hanson et al., 1999; Vincent et al., 2003; Chou et al., 2015; Cvekl and Callaerts, 2017; Yasue et al., 2017). Here, we present 3 families from our Spanish cohort of ODAs manifesting distinctly inter-generational differences in disease severity and/or ocular phenotype. We confirmed that the variable expressivity observed in all of these families resulted from parental mosaic PAX6 mutations. Both families showed somatic mosaic alleles with a relatively high AF (> 10%) in healthy or mildly affected individuals and their vertical transmission to affected offspring, which suggests the existence of gonosomal PZVs. In one of them, gonosomal mosaicism was able to be assessed thanks to the availability of both somatic and gonadal samples, resulting in better estimation of the transmission risk of the mutation to future offspring. The estimated genetic risk of about one-third found in this family is quite higher than the risk level expected in a sporadic case carrying a de novo variant, which is usually considered to be negligible (Acuna-Hidalgo et al., 2016). In the other two families, the exact risk could not be assessed due to the unavailability of sperm samples. However, both fathers likely carried the pathogenic PAX6 variant in a proportion of their germ line. Remarkably, the identification of mosaicism in an affected individual with aniridia, as we report here, involves modifying the initial genetic counseling, since the likelihood of recurrence is obviously lower than in heterozygous carriers. In view of our findings, parental mosaicism should be considered in sporadic cases with PAX6-associated diseases, even if parents do not show any ocular abnormalities.

In the two mosaic individuals carrying nonsense PAX6 mutations, similar levels of somatic mosaicism led to variable clinical outcomes ranging from congenital aniridia to considerably milder manifestations. Interestingly, we are the first to confirm low-level mosaicism in an affected individual clinically diagnosed with congenital aniridia but who manifested some apparently milder secondary ocular features compared to his fully heterozygous daughter. However, the mosaic father of a second family only showed minor iris defects that went unnoticed until adulthood. Therefore, mosaicism levels in blood are not a reliable indicator of the range and severity of ocular features. The variable phenotypic expression observed in these mosaic individuals could be influenced by different factors, including specific mutational effects, gene dosage, and lineage and timing-dependent PAX6 functionality, not only during ocular development but also in adult tissues (Shaham et al., 2012; Gregory-Evans et al., 2014; Cvekl and Callaerts, 2017). In this sense, previous studies with transgenic and, more recently, genetically edited animal models confirmed that the patterning and differentiation of different ocular structures and cell lineages require different thresholds of PAX6 activity throughout eye embryogenesis (Davis-Silberman et al., 2005; Yasue et al., 2017). Specifically, somatic PAX6 mosaicism in CRISPR-edited mouse embryos caused varying degrees of ODAs, as recently published (Yasue et al., 2017). Therefore, the identification of mosaic individuals for PAX6 mutations both in humans and animal models manifesting variable expressivity strongly reinforces the idea that PAX6 dosage plays a specific role in both eye development and phenotypic modulation. According to our findings and those reported using CRISPR models, somatic PAX6 mosaicism might explain some of the approximately 10% of patients with aniridia who apparently lack PAX6 mutations (Ansari et al., 2016; Bobilev et al., 2016; Vasilyeva et al., 2017) or even mild cases of ODAs (Hanson et al., 1999). The widespread use of more sensitive techniques, such as targeted deep sequencing and/or ddPCR, will allow for the identification of additional cases with somatic or gonosomal PZVs arising at early stages (Acuna-Hidalgo et al., 2016). However, due to limitations on tissue sampling, it will remain challenging to detect PZVs with low AF that could appear later in embryonic development or those restricted to ocular tissues.

It has long been postulated that germline mosaicism in healthy parents could be the most likely explanation for the exceptional disease recurrence in affected siblings with no family history of aniridia (Reed and Falls, 1955; Gronskov et al., 1999) or PAX6-related microphthalmia (Deml et al., 2016; Riera et al., 2017). To date, this possibility has not been fully confirmed due to the limited availability of somatic and gonadal tissues for genetic testing. Here, we report an apparent autosomal dominant 3-generation family with microphthalmia in which we first assumed a reduced penetrance due to both disease recurrence in the first affected siblings and healthy parents. However, NGS analyses only revealed a novel likely pathogenic missense variant in PAX6, a highly penetrant gene. Similar missense mutations in the paired-domain of PAX6 have been occasionally reported to cause microphthalmia (Dansault et al., 2007; Chassaing et al., 2013; Riera et al., 2017; Patel et al., 2018); thus, this variant seemed to be the cause of the phenotype in the family. To delve deeper into the causes of this unusual disease recurrence, we exhaustively investigated the possibility of parental germline mosaicism in the unaffected first generation. In the asymptomatic father of this family, an unexpectedly high mutant AF was detected in several somatic tissues from ectoderm and endoderm lineages indicating that the mosaic PZV had to emerge early in the first embryonic cell divisions. Additionally, the lack of ocular signs in this mosaic individual suggests that the PAX6 activity had to reach an adequate threshold through optic cup morphogenesis. We could not discern whether this fact was related to a gene dosage effect underlying a lower mutant ratio in embryonic eye or if this variant is hypomorphic, similar to that described for other missense PAX6 variants (Tang et al., 1997; Vincent et al., 2003). Despite the rarity of PAX6-associated microphthalmia, with fewer than 10 variants described to date (Deml et al., 2016), it is worth noting that disease recurrence has putatively been associated with mosaicism in a healthy progenitor in two additional families (Deml et al., 2016; Riera et al., 2017). Thus, the possibility of PAX6 mosaicism should also be considered when explaining occurrence in two or more affected siblings with microphthalmia in the absence of family history.

Recent evidence has suggested that transmission of parental mosaicism could explain up to 10% of apparently de novo mutations in some pathologies, and thus mosaicism rates might be higher than previously expected (Acuna-Hidalgo et al., 2016; Stosser et al., 2017). In our cohort of patients with aniridia, mosaic PZVs were identified in at least 3.8% of families, though this rate is likely underestimated. We cannot rule out that some of our patients without PAX6 mutations or some parents of de novo cases might present a low-grade of mosaicism. Therefore, larger systematic studies should be carried out using high-sensitivity ultradeep sequencing, single cell analysis or ddPCR in several somatic tissues to obtain a better estimation of the actual mosaicism rate as well as to elucidate the actual relevance of PZVs in the pathogenesis of PAX6-associated diseases.

In brief, our work confirms the long-held hypothesis of germinal mosaicism in aniridia to explain disease recurrence in the offspring of healthy parents. Additionally, we demonstrate for the first time that mosaic PAX6 variants cause phenotypic variability from congenital aniridia, mild forms of isolated iris coloboma, and even asymptomatic individuals. Consequently, our findings have important clinical implications for genetic counseling and ophthalmic management in congenital aniridia and related developmental eye disorders. Finally, somatic and/or gonosomal mosaicism should be taken into consideration as genetic factors to explain not only extraordinary cases of co-occurrence in siblings but also intra-familial variable expressivity, as well as some of apparently de novo cases and even uncharacterized PAX6-negative cases in aniridia.

MC and MM contributed the conception and design of the study. MT, MM, PR, and CV performed genetic analyses. MT, MM, PR, M-AM-P, and MV participated in the analysis and interpretation of genetic data for the work. MG, FB-K, DR-Z, GR, FM-N, ST-S, BG, and CA participated in the recruitment, clinical evaluation, and interpretation of clinical data. MC wrote the first draft of the manuscript. MT, MM, and MV wrote and/or prepared sections of the manuscript. All authors contributed to manuscript revision, reading and approving the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.