Molecular insights into foveal hypoplasia: development, genetics, mechanisms, and models

Study participants and clinical assessment

This study was approved by the clinical research ethics board of the University of British Columbia and adhered to the tenets of the Declaration of Helsinki. Patients with foveal hypoplasia were examined by retinal specialists at the Eye Care Centre in Vancouver, Canada. Informed consent was obtained from all participants included in the study. Fundus imaging, autofluorescence, and spectral-domain optical coherence tomography (OCT; Heidelberg Engineering, Heidelberg, Germany) were performed.

Functional testing in anole lizards

This study was approved by the Animal Care Committee at the University of British Columbia in accordance with the guidelines established by the Canadian Council on Animal Care. Adult female anoles were purchased from Charles Sullivan (Nashville, TN). Animals were maintained at an ambient temperature of 28 °C, with 70% humidity on a 14:10 light/dark cycle. A diet of live crickets (Acheta domesticus), which had been dusted with a 2:1 mix of calcium powder/vitamin supplement (Rep-Cal Research Laboratories, Los Gatos, CA), was provided every 2 days.

For OCT imaging, the anoles were fasted for 24 h before induction of general anesthesia with 30 mg/kg alfaxalone plus 0.1 mg/kg dexmedetomidine by subcutaneous injection into the cervical region. Topical local anesthetic (amethocaine drops) and 2.5% phenylephrine and 0.5% tropicamide were used to dilate the pupil. A drop of coupling reagent (Viscotears lubricating gel, Alcon, Fort Worth TX) was placed on the eye, and images were acquired using an image-guided OCT2 system (Phoenix-Micron, Bend, OR). For inducing injury to the choroid, anoles were anesthetized and dilating drops added as described above. After adding lubricating drops (Alcon Laboratories, Fort Worth, TX), the laser burns were induced (532-nm laser set at 500 mW for 5-s bursts) using the image-guided laser system (Micron IV, Phoenix-Micron, Bend, OR). After laser photocoagulation, the eyes were rinsed in saline, and antibiotic ointment was added. For the bait-capture test, anoles were fasted for 1 to 5 days in groups of eight before being offered a cricket to capture and eat. The number of anoles that caught the bait was recorded. For western blotting the proteins were extracted, separated by SDS polyacrylamide gel electrophoresis, and then transferred to Immobilon-FL membrane (Millipore Sigma, Etobicoke, ON, Canada) by standard methods. Membranes were incubated with a 1:100 dilution of a polyclonal PEDF antibody (PA5-75359, ThermoFisher Scientific, Toronto. ON, Canada) and visualized with a goat anti-rabbit secondary antibody (IRDy800 cw) using a Li-COR Odyssey detector (Mandel Scientific, Guelph, ON, Canada). ImageJ software was used to analyze band intensities relative to loading control (GAPDH).

For Western blot densitometry and quantitation of the bait capture test an unpaired two-tailed Student’s t test was used for comparisons between treated and untreated groups. P values less than 0.05 were considered significant. All statistics were performed in either the GraphPad Prism 5.0

Foveal anatomy

The human fovea (derived from the Latin fovea, meaning “pit”) is seen on OCT as an indentation in the macular region of the retina that measures approximately 1,500 μm in diameter [21] (Figure 1A). The inner nuclear layer (INL, containing amacrine and horizontal cells), ganglion cell layer (GCL), and blood vessels slope away, so only the photoreceptor layer (ONL) and a few Müller cell processes are directly exposed to incoming light [22]. The density of cone photoreceptors in the fovea is higher than anywhere else in the retina and accounts for 8% of the activity in the primary visual cortex, even though it only represents approximately 0.01% of the retinal area [23]. At the center of the fovea is a rod photoreceptor-free area known as the foveola (cone density 200,000/mm²) [24]. Here, the ONL has widened due to increased cone packing, and there is outer segment (OS) lengthening, seen as a small upward bend in the ellipsoid zone hyperreflective band (Figure 1A). At the center of the foveola, there is a 100-μm diameter region that is S-cone photoreceptor free [25], which corresponds to the phenomenon of foveal tritanopia (inability to distinguish blue from yellow) [26].

Clinical evaluation of fovea health

Noninvasive OCT has revolutionized the visualization of retinal pathologies, including foveal hypoplasia [27-30]. Furthermore, the use of high-resolution spectral-domain OCT has led to the development of the Leicester grading system by Irene Gottlob and colleagues, which is now used in routine practice to characterize the different severities of FH [31]. In the grading system, a normal foveal structure is composed of four elements, including complete excavation of the inner retinal layers, a foveal pit, widening of the ONL at the fovea, and OS lengthening (Figure 1A). The only difference to these features seen in grade 1 FH is a shallow pit due to incomplete excavation of the inner retinal layers. In grade 2 FH, there is no pit formation, but the ONL widening and OS lengthening are still present, showing that pit formation is not a prerequisite for cone packing. In grade 3, the only feature observed is ONL widening (Figure 1B), and in grade 4, there are no features of foveal specialization (Figure 1C), also known as foveal plana [31]. An atypical form of foveal hypoplasia is characterized on OCT imaging by disruption of the ellipsoid zone hyperreflective band (representing the photoreceptor inner segment ellipsoid highly packed with mitochondria), with a foveal pit or retention of the inner retinal layers.

However, some examples do not fit this classification, and perhaps a more nuanced grading system would be instructive when also considering the underlying genotype, the timing of foveal development, and the point at which development was arrested. One report examining FH in genotyped albinism cases revealed that some grade 1 patients had normal foveal pit parameters (OCA1 and OCA2 genotypes), suggesting the need to split the classification into grade 1a (normal pit) and grade 1b (partial pit) [32]. This has been recently added to the Leicester grading classification, although it has yet to be adopted more broadly [14]. Another report examined FH in PAX6-genotyped aniridia syndrome cases [8]. A patient was reported with a partial pit and ONL widening, but there was no OS lengthening (Figure 1D); thus, the presence of a pit and OS lengthening is not mutually coincident. Considering the timing of foveal development, the pit starts to form at 24 to 26 weeks’ gestation, whereas OS lengthening does not start until after birth, reaching maturity in the fourth year of life. Therefore, it seems likely that other factors contribute to the specification of the four anatomic foveal elements, not directly related to genotype or the presence of a foveal pit.

Studies in families have revealed that the grade of FH in a parent does not necessarily reflect the outcome in offspring with the same genetic mutation [8]. For instance, the patient with grade 4 FH (shown in Figure 1C) had a child with grade 1 FH (Figure 1D). This is important because best-corrected visual acuities (BCVAs) in grade 1 and 2 FH tend to be better than those in grades 3 and 4, in particular correlating with OS length [14,33]. In the family described above, the BCVA logarithm of the minimum angle of resolution (logMAR) of the parent (aged 33 years) and child (aged 17 years) was 0.4/hand movements (HM) and 0.4/0.1, respectively. In another family, the parent (aged 52 years, BCVA 0.6/0.4) had grade 3 FH, and the child (aged 13 years, BCVA 1.3/1.3) had grade 4 FH, representing a worse outcome for the offspring [8]. Therefore, it is essential to visualize the health of the fovea as soon as possible to help predict likely visual outcomes. The introduction of handheld OCT devices [34,35] for use in preverbal children with nystagmus has been an important advance for monitoring and evaluating foveal maturation over time.

A key feature of FH is the absence of the FAZ, and this is not included in the OCT grading system. The addition of angiography to OCT (OCTA) allows a more detailed analysis of retinal vasculature stratified by layers, giving much improved visualization of the FAZ than is possible with traditional fluorescein angiography [36]. In normal retina, the FAZ can be clearly seen with OCTA at the level of the superficial plexus (Figure 1E), whereas in FH, the FAZ is absent with blood vessels crossing the retina (Figure 1F) [37]. OCT has also been used to evaluate foveal health in diabetic retinopathy. Foveal neovascularization has been seen in 8% of cases with proliferative disease [38]. In addition, the size of the FAZ is increased before the onset of clinical signs of diabetic retinopathy [39,40]. Some studies in patients with age-related macular degeneration have also revealed a larger FAZ, foveal atrophy, and changes to the foveal microvascular circulation [41,42], alongside a reduction in the thickness of the inner retinal layers [43,44]. Finally, patients with retinopathy of prematurity have either a small or absent FAZ [45,46].

Foveal developmental timeline

Most knowledge relating to foveal development comes from extensive comparative morphological studies of human and primate eyes pioneered by Anita Hendrickson and colleagues [21,22,25,47-50]. In brief review, during ocular development, the presumptive fovea is first recognizable as a thickening of the GCL layer, beginning at around 11 to 12 weeks’ gestation [51] and continues until 22 weeks’ gestation, when it reaches five to seven cells deep (Figure 2A), with a single layer of cone photoreceptors in the ONL that lack inner and outer segments [21,52]. During this time, blood vessels have begun to grow along the horizontal meridian along the GCL/inner plexiform layer interface but do not extend into the location of the presumptive foveal pit. Stellate astrocytes and the ganglion cell plexus blood vessels form a perifoveal ring that is coincident with the rim of the incipient fovea [53]. This suggests that antiangiogenic factors repel blood vessel growth, allowing formation of the FAZ [53,54]. As development proceeds, the GCL and INL start to thin out by radial displacement and migration of cells, resulting in the start of foveal pit formation between 24 and 26 weeks’ gestation (Figure 2B). By 28 to 29 weeks, the foveal pit is more prominent (GCL/INL each three to four cells thick), and the inner segments of photoreceptors have formed, along with the fibers of Henle (Figure 2C). At about 36 weeks, there is still a single layer of cone cells, but they now have short outer segments, and the GCL/INL has each reduced to two to three cells thick (Figure 2D). Shortly after birth, the foveal pit has further deepened, with the GCL/INL each reduced to one to two cells thick (Figure 2E). During the first postnatal year, cone photoreceptors migrate toward the foveal region, and the fibers of Henle are angled away from the central fovea. By 15 months of age, only a few cells in the GCL remain, allowing unimpeded access of incident light to the photoreceptors. The foveal cones have elongated and apical process from the retinal pigment epithelium (RPE) have intercalated between the outer segments. It is not until the fourth year of life that extensive cone-packing with elongation of the foveola cone outer segments is observed (Figure 2F). The ganglion cells in the foveal region receive input from single cone photoreceptors, which maximizes spatial resolution. In comparison, in the peripheral retina, a single ganglion cell receives input from multiple photoreceptors, maximizing sensitivity [55]. The postnatal foveal morphological developmental time frame corresponds with the age at which 20/20 visual acuity is achieved in children [56]. Based on comparative studies in monkeys, the human fovea is predicted to further mature, reaching its full potential by 5 to 8 years of age [24].

Foveal development and the FAZ

Work from Jan Provis and colleagues in primates has proposed that the presence of an FAZ is a critical element contributing to pit formation [22,53,57]. Human clinical imaging using OCT or OCTA (Figure 1E,F) provides the strongest evidence that without the FAZ, a foveal pit does not form and vice versa [28,37,58]. From 22 to 25 weeks, the developing blood vessels are excluded from the presumptive pit region where the FAZ will be established (Figure 2A,B), before the formation of the foveal depression [53,59]. One model proposes that pit formation requires the presence of an FAZ on which first intraocular pressure and then retinal stretch drive pit formation [60-63]. However, more recent molecular studies have revealed several antiangiogenic and axon guidance factors expressed by ganglion cells that inhibit astrocyte migration and blood vessel formation [53,64,65]. It is most likely that a combination of both molecular and mechanical stresses defines the avascular region of the retina and subsequent pit formation.

Molecular factors contributing to foveal development

In regions of high tissue specialization, such as the fovea, along with gene defects associated with the absence of a fovea, it is expected that during development, there will be differential patterns of gene expression (higher or lower levels of expression of specific genes driving foveal formation compared to nonfoveal tissue). Several tissue and single-cell studies have identified differentially expressed genes to gain further insight into the formation of the fovea and the contribution this may have to diseases affecting the foveo-macular region of the retina [64,66-76]. Most of these studies took tissue from adult donor eyes (human or primate), making the relevance of the data for identifying genes involved in the regulation of foveal development unclear. Comprehensive information was derived regarding candidate genes for adult-onset macular diseases and gene expression profiles in individual foveal cones. Across the data sets, there were commonalities in differentially expressed genes, adding rigor to the studies [71-73]. However, two studies analyzed gene expression changes at the time when the fovea was undergoing development [64,76], which have yielded some interesting candidate genes.

In the first study, macular versus nonmacular tissue was isolated from 19- to 20-week gestation eyes, the time when thickening of the ganglion cell layer marks the location of the incipient fovea. Complementary DNA generated from the tissue was then hybridized to human genome expression microarrays [64]. Differential gene expression was identified for 25 axon guidance genes and two antiangiogenic factors. The axon guidance genes included regulators of vascular development, such as ephrin, semaphorin, slit, and netrin gene families [77]. In a follow-up study, the authors found that differentially expressed ephrin-A6 repelled vessel growth in the FAZ region [65]. The two antiangiogenic factors that were upregulated in the macular samples were pigment epithelial-derived factor (PEDF) and natriuretic peptide precursor B. PEDF is a neurotrophic factor that decreases the survival of new endothelial cells in blood vessels through an apoptotic mechanism [78]. In addition, PEDF inhibits the activity of vascular endothelial growth factor (VEGF), a proangiogenic factor expressed in ganglion cells [79]. Localization of PEDF to the incipient foveal region [64] suggests it has an important role in preventing blood vessel formation in the developing FAZ. Natriuretic peptide precursor B is localized to ganglion cells, Müller glia, amacrine cells, and the RPE [80,81]. Its function is less well understood, but some natriuretic factors can inhibit VEGF-induced angiogenesis [82]. These studies suggest that expression of antiangiogenic factors and axon guidance factors plays a role in repelling blood vessel growth in the FAZ and could be the focus of future studies.

The second study used single-cell RNA sequencing and single-cell ATAC sequencing from the marmoset retina, and notable gene expression differences were observed between P0 neonatal foveal cones and peripheral cones compared to adult foveal and peripheral cones [76]. RDH12, BACH1, AHR, and SOX6 were upregulated in neonatal foveal cones versus peripheral neonatal cones, but by adulthood, the differential expression was not maintained. Mutations in RDH12 cause Leber congenital amaurosis, where an early feature is damage to the macula [83]. The BACH1 transcription factor represses Wnt signaling and angiogenesis [84], so it may be involved in the formation of the FAZ. Mutations in AHR have been found to cause isolated foveal hypoplasia with nystagmus [85], and Ahr is expressed in developing retinal ganglion cells in mice [86]. Thus, investigation of AHR expression in human or primate foveal development is warranted. SOX4 was found to be upregulated in neonatal peripheral cones, whereas SOX6 was elevated in foveal cones. While SOX4 is known to be involved in retinal ganglion cell differentiation and axon guidance [87], the role of SOX6 in the retina is unknown. Loss of Sox6 in mice leads to dopaminergic neuronal degeneration [88], and reduced levels of SOX6 are seen in patients with Parkinson’s disease [89]. Furthermore, Sox6 is expressed in differentiating and migrating cortical neurons in the developing brain but not the adult brain [90]. Thus, the role of SOX6 in neonatal foveal cones could be associated with the migration and packing of cones into the fovea. Further studies in model systems on RDH12, BACH1, AHR, and SOX6 genes and their downstream targets could have a considerable impact on our understanding of foveal specification and their role in FH.

The candidate gene approach to target morphological observations in foveal development has also identified some interesting molecules. For example, the fibers of Henle (cone axons) undergo thinning and lengthening from mid-gestation to early childhood [21]. Similarly, cone photoreceptors undergo migration and outer segment lengthening. Thus, it could be hypothesized that growth factors may play a role in this process. It has been shown that fibroblast growth factor (FGF) 2 is expressed by cone photoreceptors during the postnatal period of outer segment elongation [91]. Additionally, a receptor for FGF signaling (FGF receptor 4) is also upregulated in cones postnatally and specifically in the fibers of Henle [92]. Growth factor studies in the FAZ have highlighted several candidates of interest. Transforming growth factor β has been shown to inhibit endothelial cell and Müller cell proliferation, which could be an important factor maintaining the developing avascular fovea [93]. In another study, VEGF messenger RNA was found to be expressed at high levels in astrocytes in advance of new blood vessel growth and in ganglion cells at the incipient fovea [94]. VEGF is usually upregulated in response to local developmental hypoxia [95]. There could be developmental hypoxia at the incipient foveal region due to increased neuronal maturation requirements of the developing foveal retina [96], yet VEGF does not induce blood vessel formation there. This could be for several reasons: (1) ganglion cells are supplied oxygen by diffusion from the choriocapillaris, so perhaps VEGF is having nonangiogenic effects on ganglion cells, such as neurotrophic/neuroprotective properties [97] or neuronal outgrowth of cell axons [98], and (2) antiangiogenic factors, such as PEDF, expressed at the FAZ overwhelm VEGF and thus inhibit astrocyte and endothelial cell proliferation [99]. Testing of these growth factors during foveal development will provide more concrete evidence of their roles in this process.

Genetic diseases exhibiting foveal hypoplasia

To date, 42 disease loci have been reported to include foveal hypoplasia as a phenotypic feature, and all but one of the genes have been identified (Table 1). More than half of the genes are associated with a syndromic form of albinism, including oculocutaneous albinism (8 genes), Hermansky-Pudlak syndrome (11 genes), and Chediak-Higashi syndrome (1 gene). Foveal hypoplasia is most frequently inherited in an autosomal recessive manner (31 diseases), whereas there are five cases of autosomal dominant transmission, five cases of X-linked inheritance, and one case of digenic inheritance.

Ocular albinism

Albinism occurs due to defects in the production of melanin or its metabolism [5]. Ocular albinism (OA), with a prevalence of 1:50,000, is phenotypically characterized by poor visual acuity due to foveal hypoplasia and is frequently associated with horizontal nystagmus, photophobia, and strabismus caused by optic chiasm misrouting [100]. OA is caused by a mutation in the GPR143 gene, which encodes a G protein-coupled receptor in the melanosome membrane of RPE cells [101,102]. GPR143 binds tyrosinase and is involved in melanosome maturation [103]. Female carriers of GPR143 variants often show a mild phenotype with reduced visual acuity. In one report, “mud-splatter” pigmentation of the fundus was observed due to random patches of X-inactivation in the RPE, iris transillumination defects, and retained inner retina layers at the fovea [104]. We have also seen phenotypic changes in a patient with confirmed OA disease (GPR143 c.885+748G>A, a known pathogenic variant [105]). This patient had reduced visual acuity (20/25 in both eyes), streaky RPE pigmentation on fundus and autofluorescence imaging, and a thinner-than-normal central fovea (left, 182 μm; right, 185 μm; normal thickness, ≥210 μm) due to reduced widening of the ONL (Figure 3). This corresponds to grade 1b FH.

Oculocutaneous albinism

Eight oculocutaneous albinism (OCA) loci cause hypopigmentation in the eyes, hair, and skin with an estimated global prevalence of 1:12,000 to 15,000 [106], although this does vary depending on subtype and geographical location. OCA1 (1:40,000) is caused by mutations in the tyrosinase gene (TYR), which catalyzes several rate-limiting steps in the production of melanin [103]. OCA1 is subdivided into OCA1A (complete lack of tyrosinase activity) and OCA1B (reduced tyrosinase activity) [100]. OCA2 (1:40,000 worldwide and 1:1,500-3,900 in sub-Saharan Africa) has a highly variable phenotype and is caused by mutations in the OCA2/P gene that regulate melanosome pH [107,108]. OCA3 (1:8,500 in Africa, but rarer worldwide) is caused by mutations in the tyrosinase-related protein 1 (TYRP1) gene. TYRP1 is a catalase involved in maintaining the stability of the TYR protein and is involved in the formation of melanosome structure [109,110]. OCA4 (1:100,000), although rare, is the most common form of syndromic albinism (24% of cases) in Japan [111]. OCA4 is caused by mutations in the solute carrier family 45, member 2 (SLC45A2), resulting in retention of the protein in the endoplasmic reticulum [112]. SLC45A2 functions as a proton/glucose exporter that increases the luminal pH of melanosomes, which inhibits melanin biosynthesis [113]. OCA5 was described in a single Pakistani family that was mapped to chromosome 4q24, but the causative gene has not yet been identified [114]. OCA6 is rare and caused by mutations in the solute carrier family 24, member 5 (SLC24A5) [115]. SLC24A5, a potassium-dependent sodium/calcium exchanger located in the melanosome membrane [116], is involved in melanosome maturation and melanin synthesis [117]. OCA7 was described in a family from the Faroe Islands and mapped to chromosome 10q22.2-q22.3 [118]. The causative gene was identified as C10orf11, now renamed LRMDA (leucine-rich melanocyte differentiation-associated) protein, which is a melanocyte differentiation gene involved in melanosome biogenesis [119]. Subsequent mutations were identified in several Dutch, Kurdish, and Lithuanian patients [118,120]. OCA8 is another rare disease caused by mutations in the dopachrome tautomerase (DCT) gene [121,122], which is also known as the tyrosinase-related protein 2 (TYRP2) gene [123]. DCT/TYRP2 catalyzes the tautomerization of L-DOPAchrome (the red melanin precursor protein) into 5,6-dihydroxyindole-2-carboxylic acid, a colorless building block of eumelanin [124,125].

Hermansky-Pudlak syndrome

Hermansky-Pudlak syndrome (HPS) is a rare autosomal recessive disorder with a prevalence of approximately 1 to 9 per 1,000,000 (Orphanet). HPS is characterized by oculocutaneous albinism with bleeding diathesis (prolonged bleeding times and/or easy bruising) [126] and can be associated with pulmonary fibrosis [127], granulomatous colitis [128], or immunodeficiency [129]. The disease is genetically heterogeneous, with 11 known loci, and all the genes have been identified (Table 1). The protein products of these genes are involved in membrane trafficking required by lysosome-related organelles such as melanosomes in melanocytes and delta granules in platelets [130,131]. Based on genotype-phenotype correlations, the individual genes can be grouped into one of four multi-subunit complexes known as BLOC-1, BLOC-2, BLOC-3, and AP-3 [131]. This functional grouping aids in clinical management and enhances understanding of the pathobiology underlying lysosome-related organelle damage. BLOC-1 deficiency: there are eight subunits in this complex, of which four (DTNBP1, BLOC1S3, PLDN, and BLOC1S5) are known to be defective in HPS types 7, 8, 9, and 11, respectively [132]. BLOC-1 patients exhibit typical ocular albinism with bleeding diathesis and colitis. BLOC-2 deficiency: this complex consists of three subunit proteins (HPS3, HPS5, and HSP6) that cause HSP types 3, 5, and 6, respectively [133]. Patients with BLOC-2 deficiency experience the mildest symptoms, including decreased visual acuity with age, nystagmus, minimal skin hypopigmentation, and mild bleeding complications. BLOC-3 deficiency: this complex consists of two subunit proteins, HSP1 and HSP4, which are defective in HPS types 1 and 4, respectively [134,135]. Patients with this type of BLOC deficiency exhibit the most severe form of OCA and bleeding diathesis. In addition, patients experience a progressive form of lethal pulmonary fibrosis and often receive a lung transplant [136]. AP-3 deficiency: this complex consists of two subunit proteins, AP3B1 and AP3D1, that are defective in HPS types 2 and 10, respectively [137,138]. In addition to OCA, patients exhibit immunodeficiency with increased susceptibility to infection caused by neutropenia. Furthermore, in HSP type 10, additional sensorineural hearing loss and neurodevelopmental delay are present. Based on this genotype-phenotype correlation, early genetic diagnosis is recommended to determine the most beneficial clinical management.

Chediak-Higashi syndrome

A single locus has been reported for Chediak-Higashi syndrome (CHS), which is a rare autosomal recessive condition characterized by variable OCA, severe immune deficiency, mild bleeding tendency, and neurologic manifestations, including cerebellar ataxia and peripheral neuropathy [139]. The defective gene is a lysosomal trafficking regulator gene (LYST) that not only regulates lysosomal trafficking [140] but also is responsible for the synthesis and transport of cytoplasmic granules [141]. Approximately 500 independent cases have been reported that can be classified into three phenotypes: (1) nonsense and frameshift variants are associated with classical, severe CHS; (2) missense variants can be associated with classical or atypical CHS, with milder, late-onset neurologic symptoms; and (3) large deletions and insertions can range from mild to severe CHS, depending on how much of the gene is disrupted [142]. Abnormalities in LYST result in abnormally large melanosomes, which impair melanin transfer in melanocytes, causing the variable OCA phenotype with grade 1 FH.

Achromatopsia

Six autosomal recessive achromatopsia loci exhibit an atypical form of FH (Table 1). All of the genes are photoreceptor-specific genes (CNGA3, CNGB3, GNAT2, PDE6C, PDE6H), except ATF6, which encodes a transcription factor regulating the unfolded protein response during endoplasmic reticulum stress [143]. The OCT imaging results in these patients show retention of inner retinal layers with disruption of the ellipsoid hyperreflective layer, which is associated with cone photoreceptor degeneration [10,14]. From a clinical perspective, the visual acuity in this group of patients is significantly worse than any of the other grades of FH [14].

Isolated foveal hypoplasia

Two genes are associated with cases of isolated FH (Table 1). Thus far, 60 disease-causing variants in the solute carrier family 38, member 8 gene (SLC38A8) have been found in patients with South Asian, East Asian, Northern European, and Arab ancestry [144]. Although the phenotype typically includes nystagmus, many cases of optic chiasm misrouting have now been recorded, but without the pigmentary deficits of albinism (FHONDA). SLC38A8 is a sodium-dependent glutamine transporter expressed in the brain and retina. Expression studies in cone-enriched retinal organoids suggest SLC38A8 could be involved in foveal development [145]. Since there are no pigmentary abnormalities, it has been suggested that SLC38A8 is downstream of the albinism genes [146].

The second gene causing isolated FH is the aryl hydrocarbon receptor gene (AHR), where only three variants have been described in patients of Middle Eastern and North African ancestry [3,147]. FH is grade 3 or 4 in these cases. AHR is a ligand-activated transcription factor involved in immunity and neurogenesis. Targeted deletion of the Ahr gene in mice affects optic nerve myelin sheath formation and horizontal eye movement abnormalities [148]. During retinal development in mice, Ahr is expressed in developing ganglion cells, so it may be involved in determining the location of the incipient fovea in humans. The exact role of AHR in foveal development warrants testing in a model system.

Developmental abnormalities with foveal hypoplasia

Mutations in 11 genes (paired box 6 [PAX6], solute carrier family 25 member 9 [SLC25A19], ferm domain-containing protein 7 [FRMD7], calcium channel voltage-dependent alpha-1F subunit [CACNA1F], NF-kappa-B essential modulator [NEMO], frizzle class receptor 4 [FZD4], kinesin family member 11 [KIF11], tetraspanin 12 [TSPAN12], low density lipoprotein receptor-related protein 5 [LRP5], norrin cystine knot growth factor [NDP], growth differentiation factor 3 [GDF3]) are associated with FH present at birth or in the first few months of life (Table 1). The most common of these is infantile nystagmus (prevalence 2.4 per 1,000), which is either idiopathic or accompanying other diseases such as aniridia, albinism, and achromatopsia. Approximately 25% of patients with infantile nystagmus who have disease-causing variants in the FRMD7 gene have a decrease in foveal pit depth, retained inner nuclear layers, and decreased cone outer segment length [12]. In addition to this grade 1 FH, patients have mild visual acuity deficits and optic nerve hypoplasia. It is not clear why nystagmus occurs, but it could be associated with chiasmal axonal misrouting, a defect in cone photoreceptor cell development, or smooth pursuit system abnormalities [149].

In approximately 80% of cases of aniridia syndrome (prevalence 1:70,000), varying grades of FH are reported [8]. The disease is caused by variants in the PAX6 gene, which result in haploinsufficiency of the PAX6 transcription factor [150]. To date, 197 unique variants have been reported (LOVD), although this does not include unique mutations in regulatory elements or chromosomal defects [151].

Aland Island eye disease (AIED) [152] and incontinentia pigmenti [153] both exhibit hypopigmentation defects, but without the axonal misrouting deficits seen in typical albinism. AIED is a calcium transport channelopathy due to mutations in the CACNAIF gene, and it is allelic with congenital stationary night blindness (CSNB2A) and X-linked cone-rod dystrophy (CORDX3). Mutations in CACNAIF cause loss of neurotransmission between photoreceptor cells and bipolar cells and cone-rod dystrophy. The few patients with AIED who have been examined exhibit grade 3 FH, which is not a feature of CSNB2A, and occurs rarely in CORDX3. Incontinentia pigmenti is an X-linked ectodermal dysplasia with foveal deficits in approximately 35% of patients, which includes the absence of a foveal pit and blood vessels crossing into the FAZ [11,154]. With a prevalence of approximately 1:100,000, disease-causing variants in IKBKG/NEMO result in a decrease in activity of the nuclear factor–κB pathway involved in cell death. How CACNA1F and IKBKG/NEMO genes are involved in foveal development awaits further testing.

FEVR is caused by mutations in five genes (FZD4, KIF11, TSPAN12, LRP5, NDP). In a comprehensive study, 25% of FEVR cases exhibited grade 1 bilateral FH and 47% unilateral FH [155]. The incidence of FH for each genotype was different, occurring in 100% of KIF11 cases, 59.3% of LRP5 cases, and 42.9%, 40%, and 38.2% of NDP, TSPAN12, and FZD4 cases, respectively. FEVR affects normal retinal angiogenesis, with NDP, FZD5, and LRP5 forming a complex that activates the β-catenin signaling pathway during development. Therefore, these genes could play a role in the developing FAZ.

Recently, atypical FH has been described in Best’s disease [13]. Approximately 27% of patients exhibited FH with a shallow foveal pit, persistence of the inner retinal layers, and blood vessels starting to cross the FAZ. The BEST1 protein oligomerizes to form a BEST1 chloride channel, and a mutant protein may lead to protein mislocalization in the endoplasmic reticulum of RPE cells. The expression of BEST1 and its role during foveal development require further investigation. Mutations in LAMB2 [156], SLC25A19 [157], and GDF3/GDF6 [158] are all associated with rare cases of FH as a feature of other congenital diseases. Whether FH is a consistent feature of these diseases is unclear.

Genetic pathways leading to foveal hypoplasia

Based on the genes involved in FH, several disease mechanisms can be proposed. For instance, mutation of many different genes involved in pigment biogenesis leads to FH, but how these defects occurring in the RPE cause foveal hypoplasia in the adjacent retina has remained an open question.

It has been long established that the RPE plays a critical role in retinal homeostasis, providing growth, trophic, and survival factors during early development; morphogenesis; and a maintenance role in the adult retina. It has been shown that the Pax6 protein cooperates with Mitf to transactivate Tyr and Tyrp1 in RPE cells [159], suggesting that the RPE may have a role in foveal morphogenesis through downstream effectors (Figure 4). TYR is required to produce L-DOPA, leading to eumelanin production through several intermediate steps [146]. TYRP1 stabilizes TYR during this process and, in addition, is required to convert 5,6-dihydroxyindole-2-carboxylic acid to eumelanin. Through a feedback loop, secreted L-DOPA binds to GPR143, resulting in the apical secretion of PEDF [160] and the inhibition of basal VEGF release (a known angiogenesis stimulant). Thus, either the lack of L-DOPA (due to mutations in TYR/TYRP1) or the inability of L-DOPA to activate mutant GPR143 results in reduced PEDF secretion into the interphotoreceptor matrix. PEDF plays a role in photoreceptor neurogenesis, ganglion cell survival, cone opsin synthesis, and the inhibition of angiogenesis [161], which are all key components of foveal morphogenesis. Importantly, in a Xenopus laevis model of retinal dysmorphogenesis, PEDF was effective in preserving the development, spatial organization, and morphology of photoreceptors and maintained steady-state levels of opsin after RPE detachment, supporting the notion that RPE-derived PEDF is actively involved in retinal homeostasis [162]. PEDF is a 50-kDa glycoprotein that is synthesized in RPE early in human development (~17 weeks’ gestation) [163], positioning its expression at around the time of foveal development. Localization of PEDF to the incipient foveal region in fetal macaque retina [64] suggests it may have an important role in preventing blood vessel formation in the developing FAZ. Based on structure-function relationships, a specific 17–amino acid region defines its neurotrophic action, which interacts with a PEDF receptor (encoded by the PNPLA2 gene) that localizes to photoreceptor inner segments [164]. Thus, we can propose that PEDF is a crucial signal downstream of the pigment biogenesis pathway involved in foveal morphogenesis.

A second well-characterized pathway involves the genes causing Hermansky-Pudlak syndrome. As described above, four multi-subunit protein complexes (BLOC-1, BLOC-2, BLOC-3, and AP-3) are responsible for the formation and maturation of melanosomes (Figure 5). Defects in melanosome formation or abnormal trafficking contribute to the disease features of HSP. Specifically, how these defects cause foveal hypoplasia remains to be determined. Based on evidence from GPR143 mutation data in ocular albinism, it is suggested that GPR143 is a biosensor, providing regulatory feedback that controls the size and number of melanosomes [165] by regulating the expression levels of MITF [166]. Thus, it is possible that the genes affected in HPS are connected to the expression of pigment biosynthesis genes downstream of MITF, thereby connecting the two pathways.

Animal models

Surprisingly little is known about how each of the FH disease genes contributes to normal foveal development due to a lack of human tissue analysis and suitable model system availability. Primate models have been instructive, but they are expensive to use and do not provide enough embryos to be used as a functional model system. Routinely used laboratory models, such as rodents, pigs, dogs, and zebrafish, do not have a fovea structure, but some models (e.g., cats and chickens) have an area centralis, where there is an accumulation of cones, photoreceptors, and retinal ganglion cells [167,168]. High visual acuity is critical to the survival of many vertebrate species that are predators and capture live prey primarily through visual guidance. Animal species with a foveal pit include albatross, raptors, pigeons [169-171], seahorses and pipefish [172], and some reptiles [173,174]. However, most of these do not have a fully curated genome, are more difficult to breed, or are considered endangered species. Of the reptile species, the anole lizard is suitable for laboratory study because lizards are easy to maintain and breed in captivity [175,176], are not an endangered species, and lay eggs before the fovea develops. In addition, the initial genome sequence [177] has now been reannotated to a deep transcriptome level [178].

Anole lizards—anatomy

The diurnal green anole (Anolis carolinensis) is an arboreal lizard that lays eggs before foveal formation. The eggs develop rapidly over a period of 22 to 27 days with 19 defined developmental stages [179]. Therefore, to determine the validity of this model system for studying foveal biology, we previously tracked foveal morphogenesis in this species [180]. The anole retina is bifoveated [181] with a central convexiclivate fovea that is thought to provide high visual acuity function [182], while the temporal, shallow fovea plays an important role in depth perception for prey-capture roles [183]. Similar to the human fovea, there is GCL thickening at the position of the presumptive pit at embryonic stage (ES) 10 before pit formation (Figure 6). The foveal pit starts to form at ES17, with thickening of the ONL, and continues to develop after hatching. A yellow pigment underlying the central foveal retinal region, similar to the macular pigment in humans (lutein), is present, which blocks blue light from causing oxidative damage [184]. Structurally, the anole fovea differs from the human fovea in that it has a thicker nerve fiber layer, and there are no nuclei in the ONL at the center of the pit (foveola)—a functional improvement that allows more light to directly access the photoreceptors. The brown anole lizard (Anolis sagrei) has also been recently characterized in detail as a model to study foveal development [176,185].

Anole lizards—gene expression studies

For the lizard model to be valuable to study molecular events during foveal development, some of the same genes would need to be expressed in both human and anole tissues. Previous literature reviews of candidate genes that might be involved in foveal/macular development identified a list of 98 genes, of which 86% had orthologous protein sequence matches to the Anolis genome [180]. These candidate genes were identified by several differential expression analysis studies or were known disease genes, suggesting similar molecular pathways between humans and anoles may be active. The expression of two candidate genes (Pax6 and Pedf) has been examined in the developing lizard fovea to test this notion. Pax6 protein expression was first detected at ES14 throughout the retina, but the highest levels of Pax6 expression were observed in the GCL at ES17 (Figure 6). Since PEDF is a downstream target of the pigment pathway genes involved in FH, its expression was tested in anole tissue, revealing Pedf protein was threefold higher in foveal versus nonfoveal tissue at ES10 (Figure 6). This suggests that Pedf may be a crucial signaling molecule that could, for instance, mark the future location of the incipient fovea. Now, many more candidate genes can be tested in this model system by either gene expression analysis or protein localization, assuming that suitable antibodies are available.

Anole lizards—functional testing

For a model system to be valuable, functional testing modalities need to be available to, for instance, standardize the evaluation of genetic manipulations or test potential treatment options. For example, to study the function of the central fovea, we used laser burns to damage the pigmented choroid of the anole eye that provides support to the retina. This resulted in foveal tearing, seen 1 month after laser injury (Figure 7A,B). Lizards were then tested using two functional modalities. Using optokinetic tracking of head movements, a behavioral response mediated through retina-brain circuitry that approximates a measurement of visual acuity [186,187], a 60% reduction in spatial threshold frequency was observed (Figure 7C). Second, a standardized bait-capture test (Figure 7D) was used, where anoles are fasted for specific lengths of time before being offered a live cricket to eat [188]. After 5 days of fasting, eight of eight wild-type controls caught the bait, whereas only three of eight laser-injured anoles were able to catch the bait (Figure 7E). These studies show that the central fovea is required for high-acuity vision.

As seen in human studies, noninvasive OCT has been particularly important in the visualization of retinal pathologies. Therefore, we wondered whether this could be possible in anoles. We used a Pheonix-Micron image-guided OCT2 system to take images in an anesthetized anole. The segmentation of retinal layers we observed matched well with the histology (Figure 8). This OCT imaging could therefore be used in testing genetic manipulation of the anole genome.

Anole lizards—genetic manipulation

One aspect of the anole model that has been challenging is genetic manipulation. Female anoles retain sperm, and the oocytes are fertilized in the oviduct, making injection into single-cell embryos problematic. However, work from two other groups recently showed that CRISPR/Cas9 gene editing in the anole and gecko lizard was achieved, removing this barrier to success [189,190]. Here, a surgical approach was pioneered by Doug Menke and colleagues to inject CRISPR reagents into unfertilized oocytes in the ovaries. After 7 days of recovery, the females were reintroduced to males, and the injected oocytes were fertilized and then laid 3 to 4 weeks postinjection. The Tyr gene was targeted to measure the efficiency of genome editing, as the resultant phenotype would be the loss of pigmentation. Interestingly, using three guide RNAs, biallelic knockout lizards were identified in the F0 generation. Subsequent analysis of the eyes from Tyr-targeted anole mutants revealed that the central fovea was present, but they had lost the temporal fovea, which was an unexpected finding [191]. More recent observations in knocking down the Oca2 gene in lizards also caused a loss of the temporal fovea [192]. To complement this work, wild-type lizards were exposed to N-phenylthiourea, an inhibitor of tyrosinase, which caused loss of the central fovea. Most recently, this group has targeted the ATF6 gene, which causes achromatopsia type 7 [143] by affecting outer segment development and consequent defects in the ellipsoid zone on OCT imaging of the fovea. Targeting of the Atf6 gene in lizards resulted in the absence of cone outer segments [193]. These gene targeting experiments have provided proof-of-principle that the anole lizard could be an excellent model to study the foveal development mechanism, but functional characterization of these knockouts needs to be validated.

Treatment for foveal hypoplasia

Due to the timeline for normal foveal development, diseases exhibiting foveal hypoplasia are currently untreatable. However, since foveal development continues after birth in humans, there is a potential therapeutic window during the first few years of life. An interesting approach toward treating albinism was proposed by Glen Jeffrey and colleagues in mice with targeted deletion of the Tyr gene, resulting in an underdeveloped retina and chiasmal misrouting [194]. Transgenic expression of tyrosine hydroxylase (which can catalyze the conversion of L-tyrosine to L-DOPA) in the RPE on a Tyr-deficient mouse background rescues the retinal abnormalities and visual function, despite the absence of melanin synthesis [195]. This suggests that a lack of L‐DOPA or one of its metabolic derivatives is key to the abnormal retinal development and function seen in albinism. Therefore, one possible intervention that has been tested in a clinical trial is oral L-DOPA supplementation [196]. However, albinism patients treated with L-DOPA did not have any improvement in visual acuity. This may be related to the patients aged 3 to 60 years who were treated, which falls outside the critical window of postnatal foveal maturation and retinal plasticity [21,48,197]. To further test the validity of whether postnatal retinal development and visual function can be modulated, Tyr-deficient mice have been administered oral L-DOPA in the early stages of life by Helena Lee and colleagues [198,199]. Compared to untreated mice, those treated from birth or from weaning for 28 days with L-DOPA had improved packing density of photoreceptors in the ONL, normalized a- and b-wave amplitudes on electroretinography testing, and improved spatial frequency thresholds on optokinetic tracking (OKT) testing measuring visual acuity. Interestingly, treatment with L-DOPA showed a dose-dependent increase in PEDF levels in Tyr-deficient mouse eyes. These data suggest that a clinical trial of L-DOPA in albinism infants is warranted. Although mice do not have a fovea, the rescue of the morphological and function defects during postnatal treatment suggests that this could be tested in the Tyr anole CRISPR model to see if the absence of the temporal fovea can be rescued.

A second approach to treat GPR143-related albinism has been to develop antisense oligonucleotides (AONs) to block aberrant retention of an intron caused by the mutation, creating a new splice site acceptor [105]. Using melanocytes cultured from patient biopsy specimens, the AON was able to restore GPR143 expression that underwent posttranslational glycosylation. AONs are being tested in a wide range of ophthalmic conditions [200], as well as to inhibit the growth of aberrant blood vessels that may be applicable to FAZ defects [201]. These types of approaches may be relevant to FH if they can be administered during the window of neurologic plasticity in the first few years of life [197].

Based on knowledge gained from understanding how L-DOPA and PEDF inhibit angiogenesis to support normal development of the FAZ, clinical trials of L-DOPA for inhibiting foveal angiogenesis in AMD have been performed. One study showed that patients being treated with L-DOPA had a 21% to 35% reduction of conversion to neovascular AMD [202], and in a second study, L-DOPA treatment was correlated with a 32% reduction in new-onset geographic AMD [203]. Similarly, L-DOPA has been tested in mouse models of diabetic retinopathy and found to have a neuroprotective role on ganglion cells [204]. These examples exemplify how knowledge of foveal developmental mechanisms can have an impact on diseases that affect the adult fovea.

Conclusion

The fovea is critically important to human vision. Now that most genes that cause FH have been identified, a suitable animal model system has been validated, and functional testing is available, this should stimulate fundamental advances in understanding the molecular and cellular mechanisms of foveal maturation and how gene mutation leads to foveal damage. This information is needed to direct interventions to improve foveal maturation and thereby prevent poor vision.