The retinal pigmentation pathway in human albinism: Not so black and white

Albinism is a pigment disorder affecting eye, skin and/or hair. Patients usually have decreased melanin in affected tissues and suffer from severe visual abnormalities, including foveal hypoplasia and chiasmal misrouting. Combining our data with those of the literature, we propose a single functional genetic retinal signalling pathway that includes all 22 currently known human albinism disease genes. We hypothesise that defects affecting the genesis or function of different intra-cellular organelles, including melanosomes, cause syndromic forms of albinism (Hermansky-Pudlak (HPS) and Chediak-Higashi syndrome (CHS)). We put forward that specific melanosome impairments cause different forms of oculocutaneous albinism (OCA1-8). Further, we incorporate GPR143 that has been implicated in ocular albinism (OA1), characterised by a phenotype limited to the eye. Finally, we include the SLC38A8-associated disorder FHONDA that causes an even more restricted "albinism-related" ocular phenotype with foveal hypoplasia and chiasmal misrouting but without pigmentation defects. We propose the following retinal pigmentation pathway, with increasingly specific genetic and cellular defects causing an increasingly specific ocular phenotype: (HPS1-11/CHS: syndromic forms of albinism)-(OCA1-8: OCA)-(GPR143: OA1)-(SLC38A8: FHONDA). Beyond disease genes involvement, we also evaluate a range of (candidate) regulatory and signalling mechanisms affecting the activity of the pathway in retinal development, retinal pigmentation and albinism. We further suggest that the proposed pigmentation pathway is also involved in other retinal disorders, such as age-related macular degeneration. The hypotheses put forward in this report provide a framework for further systematic studies in albinism and melanin pigmentation disorders.


Introduction
Albinism is a clinically and genetically heterogeneous disorder related to melanin pigment. The term 'albinism' is derived from the Latin word albus, meaning 'white'. Ancient biblical texts mention Noah as the first person with albinism (Sorsby, 1958), and Pliny the Elder quotes from the Greek writer Isogonus of Nicaea: "a sort of people born in Albania, with eyes like owls, whereof the sight is fire-red; who from their childhood are grey-haired and can see better by night than by day" (Froggatt, 1960). The word "albino" itself is Portuguese and the first recorded use is by Balthazar Tellez in 1660 describing Africans with albinism that were seen along the African coast. A century later, the word "albino" came into more general use in the French, English and German languages. Nowadays, the term "people with albinism" rather than the potentially stigmatising term "albino" is commonly used. Albinism occurs throughout the animal kingdom (Bergsma and Brown, 1976).
The first scientific clinical description of albinism dates back to 1893, by G.M. Gould in the Journal of the American Medical Association (Caddy, 1894). Gould described photophobia, nystagmus, ametropia, iris transillumination, and amblyopia in human albinism. After the rediscovery of Mendel's laws in the early 1900s, Davenport (1908) and, subsequently, Waardenburg (Van Den Bosch and Waardenburg, 1956) recognized the hereditary nature of albinism. The disease was described by Garrod as an inherited metabolic disorder and ascribed the phenotype to a biochemical defect in melanin production (Garrod, 1908). Sheridan was the first to describe paucity of uncrossed optic nerve fibres in the Norwegian albino rat (Sheridan, 1965), confirmed by Creel using electrophysiological studies (Creel, 1971;Creel et al., 1970). In 1974, Creel, Witkop and King demonstrated a significant reduction of uncrossed nerve fibres at the optic chiasm in persons with albinism, the so-called chiasmal misrouting (Creel et al., 1974). Subsequently, Tomita et al. (1989) were the first to identify a mutation in the TYR gene causing human albinism. Bassi et al. (1995) identified the gene in which mutations cause ocular albinism 1 (OA1, also known as Nettleship-Falls OA): GPR143, with albinism features limited to the eye.
While albinism has fascinated scholars for centuries, only recently major advances have been made in the understanding of its molecular and cellular pathology. In this report, to minimize complexity, we focus on the melanin pigmentation pathway in the retinal pigment epithelium (RPE) in human albinism. Nonetheless, our discussion may also have implications for other cell types and disorders where melanosomal pigment is key.
The clinical diagnosis of albinism is based on examination of the eyes, skin, hair, and assessment of chiasmal misrouting. The diagnosis can be confirmed by pedigree analysis and DNA diagnostics of all known albinism disease genes. Of note, in approximately 25-30% of patients no disease-causing mutations have been identified (yet) (Kruijt et al., 2018;Lasseaux et al., 2018;Montoliu et al., 2014). Ocular examination includes best corrected VA measurement, slit lamp examination to assess iris translucency, and fundoscopy to assess pigmentation levels. Optical coherence tomography (OCT) scans are used to determine the degree of foveal hypoplasia, measured using the Leicester grading system (Thomas et al., 2011). OCT is also used to record the asymmetry of the ganglion cell layer thickness between the nasal and temporal areas of the retina, as found in patients with albinism (Brücher et al., 2019;Woertz et al., 2020a). Analysis of visual evoked potentials (VEPs) is used to assess optic nerve misrouting (Hoffmann et al., 2005a;Kruijt et al., 2019).
In oculocutaneous albinism (OCA), the pigmented skin, eyes, and hair are all affected to a variable degree (Fig. 1). In OA1, the hypopigmentation is mostly restricted to the eyes (iris and retina) (Schiaffino and Tacchetti, 2005). Nonetheless, some evidence suggests that OA1 may also be associated with mild hypopigmentation of the skin (Lewis, 1993a). In addition to OCA and OA1, Hermansky-Pudlak syndrome (HPS) and Chediak-Higashi syndrome (CHS) include albinism in their pathological spectrum (Bowman et al., 2019;Huizing et al., 2008). Albinism deafness syndrome(s), including the autosomal dominant Tietz syndrome, as well as Waardenburg syndrome type 2, caused by mutations in MITF, are clinically considered auditory-hypopigmentation syndromes (Chiang et al., 2009;Lakhdar et al., 2021;Smith et al., 2000;Vetrini et al., 2004), and are excluded from this report. MITF is only briefly mentioned in sections 8.1 and 9.2 as it is a transcription factor that affects expression of a number of melanosomal genes. Griscelli syndrome can also be considered a syndromic form of human albinism, but it has no consistent ocular phenotype (Mancini et al., 1998;Montoliu et al., 2014;Wasif et al., 2020), and is therefore excluded from this report. Finally, FHONDA (foveal hypoplasia, optic nerve decussation defects and anterior segment dysgenesis) is an autosomal recessive syndrome with remarkable similarities to albinism, including nystagmus, foveal hypoplasia and misrouting of the optic nerve fibres. However, an abnormal pigmentary phenotype is absent (Poulter et al., 2013).
Taken together, the phenotypic variability in human albinism is wide, both between and within genetic albinism subtypes. As illustrated in Fig. 1B, patients with OCA2 can show a relatively mild phenotype, whereas it is often more severe in other albinism subtypes. Thus, the severity of the disease differs from patient to patient. No single pathognomonic sign defining the diagnosis or prognosis currently exists (Kruijt et al., 2018).

A summary of different genetic forms of human albinism
While the prevalence of human albinism in Western societies ranges from 1:14,000 to 1:17,000, in African countries it lies between 1:1500 to 1:15,000 (Kruijt et al., 2018). Potentially, the difference in prevalence between western and African populations could partly be explained by mild pigmentation defects being obvious in dark-skinned individuals while not being recognized as albinism in lighter-skinned persons. There are multiple clinical forms of albinism with at least 22 different genetic disease genes involved (Table 1) . Currently, there are eight known autosomal recessive genetic types of OCA affecting hair, skin and eye pigment. Some of the implicated disease genes encode enzymes involved in the production of melanin, while others encode proteins regulating melanosome homeostasis. There are at least two autosomal recessive syndromes including oculocutaneous albinism: HPS, with 11 implicated genes, and CHS, caused by mutations in the LYST gene. Genetic defects in these syndromes involve the biogenesis of lysosome related organelles (LROs), including melanosomes (Bowman et al., 2019). There is one well-described X-linked genetic type of ocular albinism (OA1) presenting primarily in the eye. The gene implicated in OA1 is the G Protein-Coupled Receptor 143 (GPR143), which encodes a receptor most likely active on the apical membrane of the RPE and/or the melanosomal compartment (Lopez et al., 2008;Schiaffino et al., 1996). GPR143 has a central role in the molecular pathology of ocular albinism as described in the sections 2, 8, 9 and 10 below. Finally, the autosomal recessive FHONDA syndrome has a retinal phenotype remarkably similar to albinism, but without pigmentation abnormalities (Kruijt et al., 2022;Poulter et al., 2013;van Genderen et al., 2006). Because patients consistently exhibit a severe albinism-like retinal phenotype, we hypothesise that the implicated gene Solute Carrier Family 38 Member 8 (SLC38A8) functions downstream of GPR143. This is discussed in section 6.

A summary of the biochemistry of melanosomal pigmentation in the retina
Albinism is a disorder of melanin pigmentation. There are three major types of melanin pigment in the healthy human body: brown/ black eumelanin, red pheomelanin, and black neuromelanin. Eumelanin and pheomelanin are present in the eye and skin; neuromelanin can only be found in the brain. In this report we mainly focus on the eye and more specifically the RPE, as it relates to retinal development in albinism. The RPE primarily contains eumelanin (Weiter et al., 1986). Interestingly, the type, amount and localisation of melanin in the choroid correlates to the colour of the iris (Sturm, 2009;Wakamatsu et al., 2008). Eumelanin absorbs free radicals and photonic energy. It dissipates vast amounts of energy as heat, and releases excess toxic radicals. In contrast, pheomelanin absorbs less light of the ultraviolet and other wavelengths, but is less cytotoxic (Ortonne, 2002). For the RPE, a high ratio of eumelanin to pheomelanin seems most important for its photoprotective properties (Vincensi et al., 1998). The basic features of the biosynthetic pathways of eumelanin, pheomelanin and neuromelanin are presented in Fig. 2, and briefly described below. In reality, pigmentation is "not so black and white", since multiple genes, environmental factors and different (combinations of) biochemical intermediate levels or activity determine pigmentation and a myriad of associated colours in the different tissues .
In brief, eumelanin synthesis starts with the conversion of L-tyrosine into both L-DOPA and dopaquinone by the enzyme tyrosinase (TYR; OCA1) ( Fig. 2; left column). Subsequently, inside the melanosome, dopaquinone spontaneously transforms into (leuco-)dopachrome. Next, Fig. 2. Basic biochemical features of the biosynthesis of three types of melanin: eumelanin, pheomelanin and neuromelanin. Basic, because a large number of slightly different biochemical intermediates can produce a myriad of colour shades, based on the aforementioned main types. The production of black/brown eumelanin (left column; top to bottom) in particular is relevant for RPE pigmentation, and is produced as follows: the non-essential amino acid L-tyrosine is converted by tyrosinase (TYR, OCA1) to L-DOPA and dopaquinone. Subsequently, TYR and related enzymes (TYRP1; DCT) play a role in processing intermediates and eventually the production of the eumelanin pigment. Mutations in the aforementioned genes encoding these enzymes cause albinism (see also Table 1). Reddish pheomelanin production (centre column; top to bottom) also depends on L-DOPA production by TYR and the (subsequent) generation of dopaquinone. Dopaquinone reacts with cysteine and following some intermediate steps, is converted into the reddish pigment pheomelanin. Finally, neuromelanin (right column; top to bottom) is primarily found in the substantia nigra of the brain. It is also produced from L-tyrosine and L-DOPA, and specifically dopamine. Neuromelanin is produced in granules with a core chemically related to pheomelanin and a surface related to eumelanin (Ito, 2006). There is no evidence for the presence of neuromelanin in the retina or RPE, or that neuromelanin is affected by mutations causing albinism. A key difference is that, in the brain, tyrosine is primarily (but perhaps not exclusively) converted by tyrosine hydroxylase (TH), instead of by tyrosinase (TYR) in the RPE and skin. dopachrome is converted either into dihydroxyindole (DHI) followed by subsequent processing into indolequinone (IQ), again by TYR; or into dihydroxyindole carboxylic acid (DHICA) by dopachrome tautomerase (DCT; OCA8). Leucodopachrome is processed into indole-2-carboxylic acid quinone (ICAQ) by tyrosinase related protein 1 (TYRP1; OCA3). All three resulting compounds (IQ, DHICA and ICAQ) are finally polymerized into black/brown eumelanin (Hearing and Tsukamoto, 1991;Ito and Wakamatsu, 2008;Sugumaran, 2016).
Pheomelanin also relies on the TYR-dependent production of dopaquinone ( Fig. 2; middle column). Dopaquinone subsequently reacts with cysteine to produce different forms of cysteinyl-dopa. These cysteinyl-dopa variants are oxidised by TYR to produce cysteinyl-dopa quinones and, subsequently, benzothiazine intermediates. Oxidation of these intermediates results in the generation of pheomelanin polymers (Hearing and Tsukamoto, 1991). It is thought that pheomelanin can react spontaneously with sulfhydryl compounds resulting not in red, but yellowish pigment in the hair of some patients (Ray et al., 2007). In OCA2 and OCA4, the low pH in the melanosome is not optimal for TYR function: low levels of dopaquinone are produced that spontaneously result in pheomelanin production. Indeed, in OCA2 and OCA4, the reddish pheomelanin pigment is very obviously overrepresented, resulting in yellowish, sandy blond skin and hair.
The competitive metabolic activity down the eumelanin or pheomelanin pathway in melanocytes is at least determined by competition for the shared biochemical intermediate dopaquinone and the local activity of the Melanocortin 1 receptor (MC1R). The default synthesis (in the presence of MC1R antagonists) is that of pheomelanin. Independently from antagonists, this default synthesis happens also when MC1R is mutated, a common feature in red-haired people. MC1R can be activated by the α-Melanocyte-stimulating hormone that promotes the synthesis of black-brown eumelanin (Swope and Abdel-Malek, 2018). While MC1R is also expressed in primary RPE mouse cultures , and may protect mouse retinal cells against oxidative stress (Maisto et al., 2017;Rossi et al., 2016), its exact role in the RPE remains to be elucidated.
The third type of melanin, neuromelanin, is also produced from Ltyrosine ( Fig. 2; right column). However, tyrosine is not processed by TYR, but by tyrosine hydroxylase (TH) to L-DOPA. Subsequently, in neurons, L-DOPA is converted by the aromatic acid decarboxylase into the neurotransmitter dopamine, which is one of the biochemical intermediates toward the formation of neuromelanin (Rabey and Hefti, 1990). In the neural cell, a granule forms with a pheomelanin-related reactive core encapsulated by more inert eumelanin-like pigment (Ito, 2006;Zucca et al., 2014). Neuromelanin is found in the substantia nigra in the midbrain of primates (including humans) (Carballo-Carbajal et al., 2019). The substantia nigra degenerates in Parkinson's disease. The role of neuromelanin is thought to be similar to the eumelanin and pheomelanin pigments: scavenging potentially harmful oxidative stress-generated free radicals (LaVoie et al., 2005). Neuromelanin is considered to result from a protective process against excess cytosolic dopamine that can be cytotoxic (LaVoie et al., 2005). Interestingly, TYR (OCA1) is also expressed in the human substantia nigra (Miranda et al., 1984;Xu et al., 1997b), but its potential role in neuromelanin production is unclear. Recently, Carballo-Carbajal et al. (2019) constructed genetically modified mice in which Tyr was overexpressed in the substantia nigra, and found that neural cells started to produce neuromelanin that is normally not present in mice (Carballo-Carbajal et al., 2019). A potential relation between TYR and neuromelanin is also supported by a rare TYR mutation (p.V275F) implicated in albinism that has recently also been associated with a higher risk of developing Parkinson's disease (Lubbe et al., 2016). Taken together, these initial data suggest that TYR may also have a role in neuromelanin production or function. While eumelanin and pheomelanin are primarily produced early in life, neuromelanin takes decades to form. As such, it is not present in most laboratory animals, for example rodents (Carballo--Carbajal et al., 2019). Whether the albinism phenotype of commonly used laboratory animals could influence neuromelanin formation is unknown. Finally, it must be noted here that, although dopaminergic neurons, such as amacrine cells, are present in the neural retina (Pourcho, 1982), there is no evidence that these cells contain neuromelanin.

A summary of cellular pathology in human albinism
Pigment cells (melanocytes and RPE) are affected in albinism. Melanocytes originate from the neural crest and are particularly abundant in the eye (choroid, iris) and skin. The RPE is a neuro-epithelial cell layer of the retina lining the back of the eye. The choroid is a highly vascular tissue underneath the RPE and Bruch's membrane that supplies the outer retina with oxygen and nutrients, and removes metabolic waste (Brinks et al., 2021). Melanocytes are also found in other places in the body, such as the stria vascularis of the cochlea in the inner ear (Murillo-Cuesta et al., 2010) and in the heart valves (Mjaatvedt et al., 2005). In all these tissues, the melanosomal pigment is thought to reduce oxidative stress.
The intracellular site of pigmentation is the melanosome. This is an LRO in a melanocyte or RPE cell, in which melanin is produced and stored. Healthy melanosomes develop and mature through four defined subsequent stages (Schiaffino, 2010): Stage-I melanosomes are early endosomes. Stage-II melanosomes are elongated with longitudinal pre-melanosome protein (PMEL17) fibrils but without melanin. Stage-III have clear fibrils and pigmentation (Watt et al., 2009). In the fourth and final stage, the organelles are filled with melanin and their internal structure becomes optically obscured. The primary function of melanin lies in photo-protection of (adjacent) cells and tissues against light-induced oxidative damage (Boulton and Dayhaw-Barker, 2001). The most obvious sign to the outside world is that melanin colours cells. As described in the previous section, in RPE and skin there are two types of melanin present: eumelanin (dark brown-black) and pheomelanin (yellow-red) (Boulton and Dayhaw-Barker, 2001;Strauss, 2005). Besides melanin, other pigments are also present in the retina, for example: lipofuscin in the RPE; the pigments lutein and zeaxanthin next to macular photoreceptors; and melanopsin, uniquely found in retinal ganglion cells (RGCs). This report focuses on melanin pigment because of its key role in human albinism.

The proposed retinal pigmentation pathway in human albinism
In this report, we propose a retinal pigmentation pathway hypothesis. As introduced above, several clinical and genetic subtypes of albinism exist, which typically have nystagmus, foveal hypoplasia and misrouting of the optic tracts in common. The first subtypes are syndromic forms of albinism that affect general LRO genesis and function (including melanosomes) in multiple cell types and tissues including the eye, skin and hair (section 3). The second subtype, OCA, affects only eye, skin and hair pigmentation, because of either specific melanosome dysfunction, defective melanin production, or pigment cell differentiation (section 4). Next, we describe OA with near normal melanin production but affected melanosome formation, with typical ocular, but not skin features of albinism (section 5). And finally, FHONDA, which resembles albinism remarkably with nystagmus, foveal hypoplasia and misrouting, but without pigmentary abnormalities (section 6).
To date, genetic mutations in at least 22 disease genes have been implicated in all forms of human albinism (Table 1). Each of these disease-causing genes has a specific function, with the pathology ranging from general to specific: from general syndromic pathology, to more specific oculocutaneous and ocular albinism, to remarkably restricted in FHONDA. Guided by our own data, combined with those of the literature (Bassi et al., 1995;Fernández et al., 2021;Figueroa and McKay, 2019;McKay, 2019), we propose a new relationship between disease genes and function across various types of albinism, conceptualising one retinal pigmentation pathway. This may lead to a new functional genetic classification of human albinism, which may facilitate a better understanding of its aetiology (Fig. 3). All parts of the proposed pathway and its currently known regulatory aspects are discussed below (sections 3-11). After that, conceptual and potential therapeutic consequences for albinism and retinal pigmentation in general are reviewed and discussed (section 12).
3. Genetic defects affecting LROs, including the melanosome, lead to syndromic forms of albinism 3.1. Syndromic forms of human albinism: common features of the phenotype and pathology Two syndromic forms of human OCA exist: HPS and CHS. Typically, patients with these syndromes all have key, although variable, characteristics of OCA plus a number of additional clinical features. HPS and CHS patients typically present with bleeding problems, with the latter accompanied by a serious predisposition to infections and neurodegeneration. Both syndromes are characterised by a general cellular defect in the genesis or function of LROs (Olkkonen and Ikonen, 2006;Shiflett et al., 2002;Shotelersuk and Gahl, 1998). LROs are derived from the endosomal system and form specialised organelles in specific cell types, including melanosomes in pigment cells (Luzio et al., 2014;Marks et al., 2013). Consequently, in HPS and CHS, cells other than pigment cells are also affected. HPS and CHS and their wide range of underlying molecular and cellular mechanisms have recently been excellently reviewed (Bowman et al., 2019;Fernández et al., 2021). The relevant information in the context of albinism is summarised and updated below. For each of the syndromes, the specific phenotype, pathology and underlying molecular and cellular mechanisms in (retinal) pigment cells are consecutively discussed.
3.2. Syndromic forms of human albinism: disease genes, proteins and molecular mechanisms

Hermansky-Pudlak syndrome (HPS)
HPS is a heterogeneous autosomal recessive syndrome (Oh et al., 1998;Seward and Gahl, 2013). In the general population, the prevalence of HPS is approximately 1:1,000,000. However, in some more genetically isolated populations the prevalence is much higher; for example: up to 1:1800 in Puerto Rico (Seward and Gahl, 2013), likely due to founder effects. HPS consists of at least 11 genetic subtypes (HPS1-11; Table 1). Next, we describe the features that all HPS subtypes have in common, as well as the characteristics that are unique for one or more subtypes.
HPS patients have characteristic but variable features of OCA, such as mild to severe hypopigmentation of skin, hair and eye, and including a reduced VA. In addition, patients experience excessive bleeding and bruising due to the absence of platelet-dense granules (Seward and Gahl, 2013). Some subtypes also include immune deficiencies, lung defects and granulomatous colitis (Bowman et al., 2019). In HPS, the production of a number of organelles including melanosomes is affected in cells of different organs, such as the eye (RPE and choroid), and skin and hair melanocytes. The HPS proteins can be divided in four functional complexes, called the Biogenesis of LROs Complex (BLOC) − 1, − 2 and − 3, as well as the Adaptor Protein (AP) − 3 complex (Bowman et al., 2019). An overview of these complexes and their function is presented in Fig. 4. The role of (specific subunits of) BLOC and AP complexes may be cell-type specific, which could partially explain the variability of the HPS albinism syndrome features. Interestingly though, clinical features are remarkably similar for genetic subtypes associated with the same complex. Therefore, the BLOC and AP related pathology can be best understood by discussing the four different complexes (BLOC-1-3 and AP-3) in-depth, below.
The BLOC-1 protein complex consists of 8 subunits: BLOC1S1-8 (Table 1). Mutations in the BLOC-1 genes encoding DTNBP1 (BLOC1S8; HPS7), BLOC1S3 (HPS8), BLOC1S6 (HPS9), BLOC1S5 (HPS11) have been implicated in human HPS (Gwynn et al., 2004;Fig. 3. Proposed retinal pigmentation pathway in human albinism connecting genotype (top row), cellular defects (middle row) and phenotypes (bottom row) into a single framework. Disease genes of different types of albinism affect increasingly restricted cellular and molecular mechanisms and result correspondingly in increasingly specific phenotypes. Left column: the syndromic forms of albinism (HPS1-11, CHS) involve a generalised lysosome-related organelle (LRO) defect in different cell types, which also affects melanosomes in the RPE and melanocytes. Second column (OCA1-8 downward): the different types of OCA (OCA1-8) cause specific defects of the melanosome. Some of the implicated genes encode enzymes involved in melanogenesis or pigment cell differentiation, others appear to be involved in trafficking of melanosomal proteins or regulate melanosome physiology essential for melanin formation. According to the proposed retinal pigmentation pathway, HPS1-11, CHS and OCA1-8 defects all result in lower L-DOPA production and subsequent reduced activation of its receptor GPR143 (OA1). Third column (OA1 downward): OA1, with a phenotype restricted to the eye, and a feedback loop that results in macro-melanosomes formation, and melanin pigment abnormalities in the RPE. If GPR143 is dysfunctional, no downstream paracrine signalling to the neural retina occurs through this pathway, affecting retinal development, as illustrated by the OA1 pathology of the neural retina and optic nerve. Fourth column (SLC38A8 downward): mutations in SLC38A8, downstream in the proposed pathway have been implicated in the FHONDA phenotype, which has no pigmentation related defects at all and has a consistent albinism-like severe pathology of the neural retina (foveal hypoplasia) and optic nerve (misrouting). Huang et al., 1999;Li et al., 2003;Pennamen et al., 2020). Patients with BLOC-1 mutations have a relatively mild pathology without severe immune deficiency or lung defects (Cullinane et al., 2012;Li et al., 2003). BLOC-1 mediates budding and extension of membrane tubules from the endosome to LROs including the melanosome, enabling transport of essential cofactors and enzymes Dennis et al., 2016). These include cargos that have been implicated in albinism, such as TYR (OCA1), OCA2 and TYRP1 (OCA3) (Delevoye et al., 2009;Setty et al., 2007;Sitaram et al., 2012).
The BLOC-2 protein complex consists of three subunits: BLOC2S1-3. Mutations in these genes have been implicated in, respectively, HPS3, 5 and 6 Di Pietro et al., 2004;Gautam et al., 2004). Patients with HPS3, 5 and 6 present with a relatively mild pigmentation and bleeding phenotype (Huizing et al., , 2009Michaud et al., 2017). Neither granulomatous colitis nor lung disease are present. Mice with mutations in BLOC-2 subunits exhibit hypopigmentation in the RPE and melanosome clumping in choroidal melanocytes (Dennis et al., 2015). Relatively little is yet known about BLOC-2 function. Without BLOC-2, endosomal tubules make less stable connections with cellular vesicles, including the melanosome, and cargo can end up in the wrong intracellular destination. Recently, BLOC-2 was implicated in the formation of so-called Weibel-Palade bodies (Karampini and Voorberg, 2020;Sharda et al., 2020), an essential compartment for haemostatic and inflammatory depots in platelets and endothelial cells, possibly explaining the bleeding endo-phenotype. BLOC-1 and BLOC-2 can physically interact and possibly share some functionalities (Di Pietro et al., 2006). Indeed, in cultured BLOC-1 and BLOC-2 defective mouse or human melanocytes, trafficking of the same set of proteins (TYRP1, OCA2, TYR) is affected (Di Pietro et al., 2006) resulting in similar disease pathology.
The BLOC-3 protein complex has two subunits: BLOC3S1 and BLOC3S2 implicated in specific forms of HPS (respectively: HPS1 and HPS4) (Martina et al., 2003). HPS1 and 4 patients have nystagmus and hypopigmentation of the RPE and skin. In the choroid and hair follicles, enlarged melanosomes have been observed (Di Pietro and Dell'Angelica, 2005;Erickson, 1997). Whether the RPE has enlarged melanosomes, or not, is not clear. In addition to albinism features, HPS1 and HPS4 patients eventually develop lung fibrosis that is ultimately fatal without a lung transplant (Young et al., 2012). Around 20 to 30 percent of HPS1 and HPS4 patients additionally suffer from granulomatous colitis, also known as Crohn's disease (Seward and Gahl, 2013). The BLOC-3 complex is involved in intracellular cargo travelling to LROs, including the melanosome (Gerondopoulos et al., 2012;Wasmeier et al., 2006). In BLOC3S1 (Hps1) and BLOC3S2 (Hps4) deficient mutant mice, the RPE and skin melanocytes contain smaller and less pigmented melanosomes (Gardner et al., 1997;Nguyen and Wei, 2007;Suzuki et al., 2002). In contrast, the choroid and hair bulb melanocytes contain enlarged melanosomes with normal to high levels of pigment (Nguyen et al., 2002). Apparently different cell types are affected in different ways. In BLOC-3 deficient cultured mouse melanocytes, there is a decreased presence of tubules extending from the melanosome and reduced cargo trafficking . A lack of BLOC-3 activity may lead to reduced anterograde trafficking and recycling of specific compounds, resulting in small hypopigmented melanosomes. In addition, in vitro studies suggested that BLOC-3 deficiency can cause abnormal fusion of melanosomes with lysosomes or autophagosomes (Pols et al., 2013;Pryor et al., 2004). This could lead to melanosomal membrane and cargo accumulation, explaining over-sized melanosomes in some cell types.

Chediak-Higashi syndrome (CHS)
CHS is extremely rare, with fewer than 500 cases reported in the literature. CHS is an autosomal recessive syndrome, which involves OCA features: patients have hypopigmentation of skin, hair and eyes  as well as limited VA (Kritzler et al., 1964). Affected individuals have typical grey/silver hair, different compared to other types of albinism. Giant melanosomes are found in patient melanocytes and RPE (Boissy and Nordlund, 1997;Valenzuela and Morningstar, 1981). Remarkably, the CHS RPE is (almost) devoid of pigment, while the choroid seems relatively unaffected (BenEzra et al., 1980). Apart from OCA features, CHS patients suffer from bleeding and bruising problems and are prone to infections (Introne et al., 2017). Approximately 90% of patients with CHS suffer from severe immune system dysfunction, namely hemophagocytic lympho-histiocytosis (Toro et al., 1993), an overproduction of immune cells. CHS patients are frequently treated with bone marrow transplantation at an early age to overcome immune deficiencies. However, this treatment does not prevent CHS-associated neurodevelopmental or neurodegenerative problems (Tardieu et al., 2005).
CHS is caused by mutations in the lysosomal trafficking regulator (LYST) gene. LYST encodes a large protein of around 4000 amino acids of largely unknown function (Nagle et al., 1996;Ward et al., 2003). It

Fig. 4.
A basic overview of HPS trafficking complexes, BLOC-1-3 and AP-3. The BLOC and AP complexes function to aid the transport of membrane-bound cargo, to and from different LROs (including melanosomes). BLOC-1 is involved in the budding of membrane tubules, allowing them to extend from the endosome to the developing melanosomes. BLOC-2 supports stable connections between the extending membrane tubules and the extracellular vesicles. Without BLOC-2, cargo can arrive at the wrong destination, such as the Golgi or plasma membrane instead of the melanosome. The BLOC-3 complex controls antero-and retrograde trafficking between the melanosome and other cellular compartments. Finally, the AP-3 complex is involved in the budding of the endosomal vesicles used to transport cargo to other LROs. contains a domain which has been implicated in both fission or fusion of membranes (Cullinane et al., 2013) and sorting or trafficking of vesicles (Burgess et al., 2009). TYR and other LRO proteins are abnormally secreted (Zhao et al., 1994). Corroborating evidence was found in cell cultures and CHS skin histology: presence and secretion of relatively undifferentiated CHS LROs were observed. CHS melanocyte cultures showed formation of enlarged lysosomes as well as giant hypopigmented melanosomes. Apparently, skin CHS melanosomes were not able to transfer from melanocytes into keratinocytes (Boissy and Nordlund, 1997;Zelickson et al., 1967;Zhao et al., 1994). Cultured CHS patient B-cells suggested that LYST is involved in a microtubule-dependent sorting process of endosomal proteins to multivesicular endosomes (Faigle et al., 1998). Studies using cultured wild-type and defective LYST orthologue (beige) mouse macrophages and fibroblasts corroborate that LYST is involved in (abnormal) lysosomal fission (Durchfort et al., 2012). Recently, Lattao et al. (2021) found that the LYST Drosophila counterpart Mauve suppresses both endosomal vesicle fusion resulting in enlarged LROs, and CHS centrosome position defects.

Genetic melanosome defects: the phenotype and pathology
While syndromic forms of albinism affect the generation and function of LROs, including the melanosome in a wide range of cells, OCA involves more specific abnormalities affecting pigment cells only in the eye, skin, and hair. In OCA1, OCA3, and OCA8, synthesis of melanin is directly reduced as enzymes in the melanin biosynthetic pathway are mutated (Fig. 2). In OCA2 and OCA4, trafficking of specific melanosomal proteins is disrupted which leads to altered melanosome formation (Bin et al., 2015;Puri et al., 2000). This is accompanied (possibly also in OCA6) by a reduced intra-melanosomal pH, which affects TYR activity and leads to reduced melanin production (Bin et al., 2015;Brilliant, 2001;Vitavska and Wieczorek, 2013). The gene causing OCA7 (LRMDA) has been implicated in pigment cell-specific differentiation, but little is yet known about its function (Grønskov et al., 2013). Genetic linkage studies assigned the OCA5 locus to human chromosome 4, but no causative candidate disease gene has been identified yet (Kausar et al., 2013). In the next subsections we discuss each type of OCA, the associated genes and proteins, and pathogenic variants most likely affecting protein function.

OCA1: TYR, tyrosinase
OCA1 is an autosomal recessive disorder with a prevalence of around 1:36,000 in Western countries (Kruijt et al., 2018). The specific associated phenotype, which is highly variable, ranges from complete absence of pigmentation to (almost) normal pigmentation in hair, skin, and eyes. Ocular features include nystagmus, reduced VA, iris transillumination, foveal hypoplasia, and abnormal axonal crossing at the optic chiasm (King et al., 2003). TYR, located on human chromosome 11q14-q21 (Barton et al., 1988), contains five exons, resulting in a 2062 bp transcript (NM_000372.5). The TYR promoter region contains many binding sites for regulatory proteins, such as the melanocyte inducing transcription factor (MITF) that influences the expression of the gene and, partially, the activity of the encoded protein (Ray et al., 2007). Most of what we know about the key evolutionarily conserved regulatory regions in the human TYR gene originates from the numerous studies carried out with the Tyr − /− mouse (Seruggia et al., 2021).

Protein structure and function.
Human TYR (OCA1) encodes a protein of 529 amino acids, resulting in a ~80 kDa glycoprotein. It is a membrane-bound mono-phenol mono-oxygenase. The 2D structure of the human protein is presented in Fig. 5. The N-terminus of the protein consists of an 18 amino acid signalling domain targeting the protein to the endoplasmic reticulum (ER). Next to it, the protein contains two highly conserved cysteine clusters, the second of which is an epidermal growth factor (EGF)-like region, most likely involved in protein-protein interactions (Davis, 1990;García-Borrón and Solano, 2002). TYR has seven N-glycosylation sites and two central Cu 2+ binding sites (CuA and CuB) with a conserved cysteine cluster in between (García-Borrón and Solano, 2002). Towards the C-terminus, the CuB binding site is followed by a transmembrane domain and a dileucine and tyrosine-based motif essential for melanosomal targeting (Calvo et al., 1999;García-Borrón and Solano, 2002;Höning et al., 1996;Sandoval et al., 1994).
TYR expression is primarily seen in melanocytes and RPE cells. A complex process is needed to produce and deliver fully functional TYR to the melanosome. First, release from the ribosome into the ER lumen is accompanied by cleavage of an N-terminal signal sequence (Bouchard et al., 1989;Kwon et al., 1987;Yamamoto et al., 1989). Next, TYR moves from ER to the Golgi, where it reaches its active configuration by folding and glycosylation (Wang and Hebert, 2006). Mediated by ATP7A (MNK) (Petris et al., 2000;Setty et al., 2008), two copper ions are incorporated at the TYR CuA and CuB site (Fig. 5). Finally, TYR is transported from the trans-Golgi network to stage-II melanosomes (Costin et al., 2003). An essential factor for melanosomal localisation are the OCA2 and SLC45A2 (OCA4) proteins. In its absence, TYR is proteolyzed and secreted from melanocytes (Costin et al., 2003;Manga et al., 2001). Once TYR has fully matured it catalyses three reactions in the melanin pathway (Fig. 2). Of note, mutations in the aforementioned ATP7A gene itself cause Menkes disease, a rare X-linked disorder of copper metabolism. Menkes disease patients frequently present with, among others, blue irides and retinal hypopigmentation. Indeed, the mottled mouse model for Menkes disease is named after its blotchy pigmented fur, possibly due to X-inactivation. The potential relationship between Menkes disease and albinism warrants further investigation (Gasch et al., 2002;Mercer, 1998).

Genetic variation and mutations.
Mutations in TYR are the most frequently found cause of human albinism. Missense mutations in TYR are clustered to six domains of the encoded protein: A cluster affecting conserved cysteine residues at the N-terminal side, one around the EGFlike domain, two spanning, and one in between the two Cu 2+ -binding domains, and one located between the second Cu 2+ -binding and the transmembrane domain (Fig. 5). These clusters seem to be essential for protein function and likely affect proper folding of the protein, with mutations leading to retention and degradation in the ER Halaban and Moellmann, 1990;Toyofuku et al., 2002). One of the most common TYR mutations found in humans (p.T373K) alters a TYR glycosylation site, resulting in ER degradation Lasseaux et al., 2018), which is also found in the extreme dilution mottled mouse model (Lavado et al., 2005). Yet other (missense) mutations, outside the aforementioned domains, may result in reduced, but not absent, trafficking capability of the protein and lead usually to a milder phenotype . Generally, nonsense TYR mutations result in premature chain termination and loss of protein function (Ray et al., 2007). When no protein is delivered to the melanosome, a severe phenotype can be expected. Further mutation detection in TYR has been limited by the usual suspects, including clinical misdiagnosis, the presence of hypomorphic risk variants and mutation protocols only screening the coding regions. For example, the TYR risk variant p.R402Q, acts as a pathogenic variant only when combined with a fully penetrant pathogenic variant, together resulting in an albinism phenotype (Grønskov et al., 2019;Lasseaux et al., 2018;Monfermé et al., 2019). Also, recent studies in man and mice strongly suggest the presence of regulatory gene expression variants in the 5' non-coding region of TYR influencing pigmentation (Giménez et al., 2001;Michaud et al., 2021;Seruggia et al., 2021). To our knowledge, no apparent genotype-phenotype correlation has been observed.

Protein structure and function.
OCA2 encodes a melanosome membrane protein of 838 amino acids with a mass of 110 kDa (Rosemblat et al., 1994). The OCA2 protein has 12 transmembrane domains ( Fig. 5) (Rinchik et al., 1993) and has two dileucine motifs in its cytoplasmic tail. These dileucine motifs are essential for melanosomal localisation of the protein. Interestingly, the motifs interact with ubiquitous cytoplasmic adaptors AP-1 and AP-3 implicated in HPS (section 3.2.1) (Sitaram et al., 2009). Electrophysiological patch clamping of skin cells suggested that the OCA2 protein is a Cl − channel (Bellono et al., 2014). The channel is involved in regulation of the pH (Puri et al., 2000), which is essential for proper TYR function and melanin formation (Bin et al., 2015;Brilliant, 2001).

Genetic variation and mutations. Missense mutations in OCA2
are somewhat clustered in the central region of the second and fifth transmembrane domains, but are also present elsewhere ( Fig. 5) (Oetting and King, 1999). There are also a few loss of function nonsense mutations in different locations of the coding region (Oetting and King, 1999). An estimated 77% of OCA2 patients in sub-Saharan Africa carry a 2.7 kb deletion spanning exon 7 (Stevens et al., 1997). Interestingly, genetic defects, such as uniparental disomic inheritance, imprinting and (de novo) large deletions of the chromosome 15q11.2-q13 region, encompassing the OCA2 gene, cause Angelman syndrome or Prader-Willi syndrome (Sakazume et al., 2012;Spritz et al., 1997). A proportion of patients with the aforementioned syndromes have an albinism-like phenotype. Further sources of genetic variation influencing the OCA2 phenotype are copy number variations (Akahoshi et al., 2001) and variations outside the coding region (Donnelly et al., 2012;Han et al., 2008). A remarkable example is the single nucleotide polymorphism (SNP) (rs12913832) located 21.5 kb upstream of OCA2 in an intron of HERC2 (Eiberg et al., 2008). This SNP contributes to blue eye colour in the European population. However, HERC2 itself has no known role in pigmentation and the region around the SNP is hypothesised to be a distant regulatory region containing MITF motifs, which affects OCA2 expression (Eiberg et al., 2008;Levy et al., 2006;Sturm, 2009;Sturm et al., 2008;Visser et al., 2014).

OCA3: TYRP1, tyrosinase related protein 1
OCA3 has a prevalence of approximately 1:8500 individuals in Southern Africa, but in Western countries it is virtually absent (Rooryck et al., 2008). OCA3 was originally called 'rufous' or 'red' oculocutaneous albinism, since its phenotype is characterised by a copper-red skin and increased pheomelanin production . Ocular defects are, in general, not as severe as in OCA1 or OCA2 (Kromberg et al., 1990). The OCA3 gene, TYRP1, located on to the human chromosome 9p23 (Chintamaneni et al., 1991;Manga et al., 1997), consists of 8 exons, encoding 2896 bp of RNA (NM_000550.3). TYRP1 is the human homolog of the mouse Tyrp1 in the brown locus (Jackson, 1988).

Fig. 5.
Overview of missense mutations in disease genes for non-syndromic forms of albinism and FHONDA: The diagram shows the protein domains and distribution of missense mutations. We focussed on missense mutations, because of practical presentation and functional considerations. On the very top: for the TYR gene, we present a histogram showing the 166 pathogenic missense variants (so excluding non-pathogenic and nonsense variants) in the Leiden Open Variation Database (LOVD; http://lovd.nl). Clustering of missense mutations can be seen in TYR around the conserved cysteine residues before the EGF-like domain as well as in the EGF-like domain itself and around the metal-binding domains. Next (top to bottom): for each of the genes implicated in OCA as well as the OA1 (GPR143) and the FHONDA gene (SLC38A8) the most important protein domains are shown. Pathogenic variants derived from LOVD are shown for the OCA and OA1 genes. The (distribution of) mutations in genes indicate important domains for the function of the protein, which are discussed in the relevant paragraphs for each gene in the text. An overview of mutations in the FHONDA gene, SLC38A8, and related phenotype, have been recently published (Kruijt et al., 2021). 4.2.3.1. Protein structure and function. TYRP1 encodes a membranebound protein of 537 amino acids. The TYRP1 protein structure is presented in Fig. 5. The N-terminal end of TYRP1 consists of a highly conserved signalling domain. TYRP1, like TYR, contains two clusters of five cysteine residues, one of which, close to the C-terminus, is an EGFlike domain. TYRP1 also has a C-terminal end transmembrane domain and a tyrosine based dileucine motif for melanosomal targeting (García-Borrón and Solano, 2002;Vijayasaradhi et al., 1995). In contrast to the TYR Cu 2+ binding sites, TYRP1 has Zn 2+ binding sites that are presumed catalytically inactive in humans (Decker and Tuczek, 2017;Lai et al., 2017). TYRP1 residues are frequently glycosylated, but potential functional consequences are yet understudied (Xu et al., 1997a).
TYRP1 and TYR are evolutionarily conserved homologues which interact and stabilise each other at the cell membrane of the melanosome (Kobayashi and Hearing, 2007;Kobayashi et al., 1998;Orlow et al., 1994). In murine melanocytes, Tyrp1 functions as a DHICA oxidase during the biosynthetic conversion of tyrosine to eumelanin (Boissy et al., 1998;Sarangarajan and Boissy, 2001). Apparently, it has a different, currently unknown function in human melanocytes (Sarangarajan and Boissy, 2001). In addition, in the mouse, Tyrp1 modulates Tyr activity (Sarangarajan and Boissy, 2001).

Genetic variation and mutations.
There are only a few known missense mutations in TYRP1, which are present in the last two thirds of the coding region (Fig. 5). Most importantly, three missense mutations were identified in the first and second metal binding domains. These mutations possibly influence the ability of the protein to be properly loaded with zinc ions, rendering it inactive. Nonsense mutations in TYRP1 leading to premature chain termination and loss of protein are relatively common. There are three hotspots of these in TYRP1. The first is around the first cysteine cluster, thought to be involved in proteinprotein interaction. The second one is just after the first metal binding domain, and the third hotspot is located the second metal binding domain. These hotspots possibly locate to regions of the gene that are sensitive to replication errors.

OCA4: SLC45A2, solute carrier family 45 member 2
OCA4 has a prevalence of around 1:100,000 in Caucasian populations. However, OCA4 is very common in East Asia and Japan, where it accounts for up to 24% of all albinism cases (Inagaki et al., 2004;Suzuki and Tomita, 2008). Patients with OCA4 can be as severely hypopigmented as OCA1 patients, but in some patients pigmentation levels increase in the first decade (Inagaki et al., 2004;Rundshagen et al., 2004). Most patients present with foveal hypoplasia and chiasmal misrouting as well as nystagmus and poor VA, but some have only mild ocular features (Kruijt et al., 2021). OCA4 is caused by mutations in the human solute carrier family 45 member 2 (SLC45A2) gene . SLC45A2 is located at Giemsa band 5p13.3. The gene has seven exons, which produces three alternatively spliced transcripts: a long 2775 bp transcript (NM_001012509.4), and two shorter ones (NM_016180.5, 1728 bp transcript; NM_001297417.4, 1122 bp transcript). SLC45A2 encodes an integral melanosomal membrane protein (Bin et al., 2015). The human gene is the orthologue of the mouse underwhite (uw) gene .
Indeed, melanocytes of the mutant mouse orthologue for SLC45A2, uw, show defective TYR trafficking and a lower melanosomal pH (Bin et al., 2015;Le et al., 2020;Puri et al., 2000). A low melanosomal pH is detrimental for TYR activity.

Genetic variation and mutations.
Missense and nonsense mutations in SLC45A2 have been identified but do not appear to be clustered to particular domains of the protein (Fig. 5). Mutations affect protein stability and transport activity that result in complete loss of melanin production in vitro (Cook et al., 2009). Cultured human melanocytes with common African or Caucasian SLC45A2 SNPs express different levels of SLC45A2 transcript and show differential TYR activity (Cook et al., 2009). A number of OCA4 mutations were reproduced in the evolutionarily conserved homologous regions of a rice sucrose transporter, and confirmed a corresponding severe decrease in transport activity (Reinders and Ward, 2015). Finally, the SLC45A2 gene and its promoter region have also been implicated in natural skin and eye pigmentation variation (Graf et al., 2007;Mengel-From et al., 2010;Sturm and Larsson, 2009).

OCA5: chromosomal locus 4q24
The OCA5 locus (4q24) was found in a family presenting white skin, golden yellow hair, low VA, nystagmus and foveal hypoplasia (Kausar et al., 2013). So far, the causative gene has not been identified. The relevant genomic bin includes at least 14 candidate OCA5 genes. However, none of these genes has, apparently, a direct functional relationship with the melanogenesis pathway. One candidate gene, MANBA, has been implicated in lysosomal function, while other genes, SLC9B1, SLC9B2, and SLC39A8 are of the solute carrier family. Screening of candidate genes (SLC9B1 and SLC9B2) did not yield any pathogenic variants (Kausar et al., 2013), so further research is warranted to identify the OCA5 disease gene. 4.2.6.1. Protein structure and function. SLC24A5 encodes a 500 amino acid solute carrier protein that functions as a K + dependent ion exchange transporter for Na + /Ca 2+ (NCKX5). SLC24A5 contains two clusters of five transmembrane domains and a large cytosolic loop (Fig. 5). Its proper localisation is dependent on the cytosolic loop (Schnetkamp, 2004(Schnetkamp, , 2013. The SLC24A5 protein is located in the trans-Golgi network in cultured melanocytes, HEK293 cells, and the melanoma cell line MNT1, but its location in the RPE has not been studied (Ginger et al., 2008;Lamason et al., 2005;Rogasevskaia et al., 2019).
While there is a pigmentation phenotype in OCA6 patients, it is remarkable that there is no clear evidence for presence of SLC24A5 in wild-type melanosomes. Possibly it has an early vesicle function in the development of melanosomes. Alternatively, SLC24A5 may regulate transcription of MC1R or affect cholesterol homeostasis, both affecting melanogenesis with an apparently severe effect on melanin (Wilson et al., 2013). Epidermal melanocytes from OCA6 patients have a relatively high number of immature melanosomes (Wei et al., 2013), suggesting melanosomal maturation is disturbed. Of note, zebrafish OCA6 models with mutations in the slc24a5 orthologue are called the golden mutants, exhibiting fewer and smaller melanosomes in the eye and skin (Lamason et al., 2005). A similar phenotype was observed in Slc24a5 − /− mice (Vogel et al., 2008). Thus, SLC24A5 may have an opposite regulatory role compared to GPR143, which is involved in macro-melanosome formation. 4.2.6.2. Genetic variation and mutations. Only three SLC24A5 missense mutations have been identified, which are present in the second and fourth transmembrane domain: p.A115E (Morice-Picard et al., 2014), p. R174K and p.S182R (Bertolotti et al., 2016) (Fig. 5). The potential functional consequences of these mutations were studied in the closely related protein family member NCKX4, using a Ca 2+ transport assay (Jalloul et al., 2016). While protein expression levels were not affected, the p.R174K mutation resulted in a more than 70% decrease in Ca 2+ transport, while the other two variants showed no exchanger activity at all. Defective SLC24A5 leads to reduced pigment biosynthesis (Lamason et al., 2005). Ginger et al. (2008) overexpressed SLC24A5 carrying a p. A111T nsSNP in "High Five" insect cells, resulting in a reduced Ca 2+ exchange over the plasma membrane. SLC24A5 maintains the proper melanosomal ion concentrations which possibly, like OCA2 and SLC45A2, contributes to melanosomal pH. Chain-terminating nonsense mutations, resulting in loss of protein function, have been identified across the entire SLC24A5 coding region. Like other albinism genes, genetic variation in SLC24A5 contributes also to variation in natural human pigmentation. For example, the SLC24A5 nsSNP, p.A111T, in Caucasians has been implicated in light skin (Ginger et al., 2008;Stokowski et al., 2007).

OCA7: LRMDA, leucine-rich melanocyte differentiation related
Mutations in the OCA7 gene (LRMDA) were found in a Faroese albinism family and confirmed in a Lithuanian patient (Grønskov et al., 2013). OCA7 patients are not completely devoid of pigmentation but usually have lighter skin than healthy relatives. In addition, typical ocular albinism features have been reported, including nystagmus, reduced VA, and misrouting of the optic nerves (Grønskov et al., 2013;Kessel et al., 2021;Kruijt et al., 2018). OCA7 is associated with mutations in the leucine-rich melanocyte differentiation related (LRMDA; also known as C10orf11) gene located to human chromosomal region 10q22.2-q22.3. The gene has seven exons coding for two transcripts, one of 3662 bp (NM_001305581.2) and one of 3692 bp (NM_032024.5).
4.2.7.1. Protein structure and function. The LRMDA gene encodes a protein of 198 amino acids. It is a member of the leucine-rich repeat (LRR)-containing protein family (CT)-LRR (Fig. 5) (Bella et al., 2008;Grønskov et al., 2013). LRMDA localises to melanoblasts and melanocytes of foetal skin epidermis (Grønskov et al., 2013). CT-LRR proteins have multiple functions, including RNA processing, extracellular matrix assembly, cell adhesion or signalling and neuronal development (Bella et al., 2008). Knockdown of the LRMDA zebrafish homolog c10orf11 (lrmda) resulted in reduced pigmentation and fewer pigmented melanocytes (Grønskov et al., 2013). The phenotype in the fish was rescued by wild-type, but not mutant c10orf11 mRNA. Intriguingly, LRMDA expression was neither found in embryonic RPE of seven weeks, nor in our own retinal organoid dataset (section 7.2; Supplementary Fig. 2) (Wagstaff et al., 2021b). This is remarkable because a neural retinal albinism phenotype is present in these patients, and warrants further investigation.

Genetic variation and mutations.
Only a few mutations in LRMDA causing albinism have been found. A study in 23 unrelated Iranian albinism patients reported a p.N89K missense mutation (Fig. 5) (Khordadpoor-Deilamani et al., 2016). The functional consequence of these mutations remains to be elucidated.
Next to the aforementioned missense mutation, both LRMDA stop and frameshift mutations have also been implicated in OCA7. The stop mutation is located very close to the C-terminus at residue 194 of 198 (p. R194*) (Grønskov et al., 2013). The frameshift mutation is located at residue 23 of the protein (p.A23Rfs*39) likely resulting in loss of functional protein (Grønskov et al., 2013). 4.2.8.1. Protein structure and function. DCT (also known as tyrosinase related protein 2, TYRP2) encodes four protein isoforms of 552, 519, 456 and 208 amino acids. In Fig. 5, the DCT structure is given for isoform 1 (519 aa). DCT is highly similar to TYR and TYRP1. The three proteins contain an evolutionarily conserved EGF-like domain (Fig. 5). DCT, like TYRP1, also has two Zn 2+ -binding sites essential for catalytic activity (Olivares and Solano, 2009). A transmembrane domain is located at the C-terminal end of the protein followed by two tyrosine-based motifs for melanosomal targeting (García-Borrón and Solano, 2002). Finally, there are several sites for possible glycosylation. Knockout of Dct in mice results in altered pigmentation, which is called the slaty phenotype (Budd and Jackson, 1995;Jackson et al., 1992). Just as in OCA8 patients, pigmentation of the RPE is decreased in Dct − /− mice (Fig. 2). DCT functions as a biochemical intermediate in the eumelanin biosynthetic pathway   (Fig. 2; section 1.3).

Genetic variation and mutations.
So far, five mutations in DCT, associated with albinism, have been published. Pennamen et al. (2021) described two missense mutations affecting conserved cysteine residues (p.C40S and p.C61W) in the first EGF-like domain. One additional missense mutation was recently identified: p.G59V, that most likely affects TYRP2 glycosylation and results in ER retention of the protein.

OA1: the specific phenotype
OA1 is the only proven form of albinism inherited in an X-linked recessive fashion. This disease subtype has a prevalence of at least 1:60,000 (Rosenberg and Schwartz, 1998). The phenotype of OA1 is mainly restricted to the eye, while pigmentation of skin and hair are normal. Nonetheless, anecdotal evidence suggests that male OA1 patients might have somewhat pale skin as well (Lewis, 1993a). Typically, ocular pathology in affected males is on average more severe than in OCA1 or OCA2, and characterised by iris translucency, foveal hypoplasia and chiasmal misrouting (Schiaffino, 2010). On a cellular level, fewer and abnormally enlarged melanosomes are present in the RPE as well as in skin, iris and choroid melanocytes (Garner and Jay, 1980;O'Donnell et al., 1976;Wong et al., 1983). Intriguingly, while pathology is limited to the eye, macro-melanosomes are present in both eye and skin (Schiaffino, 2010;Schiaffino and Tacchetti, 2005). In female carriers, punctate iris translucency and a typical pattern of peripheral mottled retinal pigment changes probably due to random X chromosome inactivation can be seen in the fundus (Lang et al., 1990;Oetting, 2002).
In Gpr143 − /− mice, the hair colour pigmentation is indistinguishable from that of wild-type pigmented mice. However, in mice, coat colour is governed by hair pigmentation, unlike humans, where the skin pigmentation is primarily determined by melanocyte-keratinocyte interactions outside the hair bulb. This suggests that also hair bulb melanocytes are not affected in OA1 subjects . Below we discuss the OA1 gene, its protein function, and the relevant effects of mutations in this gene.

Protein structure and function
GPR143 consists of 404 amino acids and is a G protein-coupled receptor (GPCR)  with seven membrane-spanning domains (Sone and Orlow, 2007). Similar to other GPCR family members, GPR143 has two extracellular loops, two evolutionarily conserved cysteine residues (Schiaffino et al., 1999), and a DRY-like motif in transmembrane domain three (Sone and Orlow, 2007). In addition, GPR143 contains endosomal/lysosomal targeting motifs: a dileucine motif in both the third cytosolic loop and sixth luminal/extracellular loop (Sone and Orlow, 2007), and a tryptophan-glutamic acid doublet in the C-terminal tail . Relevant for downstream GPCR signalling, both the second and third cytosolic loops as well as the C-terminal tail contain G protein-coupling domains Schiaffino et al., 1999).
Localisation of GPR143 was studied in both pigmented and nonpigmented cells. In general, GPCRs are localised to the plasma membrane (classical signalling), or the endosomal compartment (endosomal signalling) (Liccardo et al., 2022;Pavlos and Friedman, 2017). The localisation(s) of GPR143 in the human RPE cell in vivo remains to be elucidated. In non-pigmented cultured cells (COS-7; HeLa; COS-1; CHO) transfection of recombinant GPR143 showed various possible subcellular localisations of the gene product, probably depending on culture conditions and difference in study design. GPR143 was not only localised to intracellular vesicles such as late endosomes/lysosomes (Burgoyne et al., 2013;Piccirillo et al., 2006;Schiaffino et al., 1999;Sone and Orlow, 2007), but also to the plasma membrane (De Filippo et al., 2017a, 2017bInnamorati et al., 2006;Khristov et al., 2018;Lopez et al., 2008). Supporting these data, recombinant GPR143 localisation in yeast Saccharomyces cerevisiae was found in both late endosomes and the plasma membrane (Staleva and Orlow, 2006). Others however showed only intracellular localisation of GPR143, regardless of expression level and culturing in the absence or presence of tyrosine (De Filippo et al., 2017a).
In human RPE, in situ labelling of GPR143 showed positive signals both intracellularly and on the plasma membrane, a typical result for GPCR localisation studies to date (Lopez et al., 2008;McKay, 2019;Pavlos and Friedman, 2017). In human donor RPE, 3.5 ± 0.7% of the total GPR143 was present on the apical plasma membrane surface of RPE (Lopez et al., 2008). In cell cultures, intracellular localisation apparently depends on the level of tyrosine in the medium (Lopez et al., 2008;Schiaffino et al., 1996). GPR143 localisation studies in GPR143 KO mice were not conclusive (Bassi et al., 1996;Cortese et al., 2005;Incerti et al., 2000).
Recently, L-DOPA was identified as one of the ligands and dopamine as a competitive inhibitor of GPR143 (Goshima et al., 2014;Lopez et al., 2008;McKay, 2019). The N-terminal ligand binding part of GPR143 extends into the extracellular space or melanosomal lumen, while its C-terminal is in the cytoplasm. (Extracellular) activation of the GPR143 RPE receptor results in downstream intracellular GPR143 signalling. The central role of L-DOPA and GPR143 signalling in albinism pathology and regulation of the retinal pigmentation pathway is further explored in sections 5.3 and 8-11 below. Possibly, the G protein subunit Gαi3 (GNAI3) may be essential for GPR143 downstream signalling through a second messenger cascade. Indeed, in mice, Gαi3 interacts with Gpr143, and the independent knockout of Gαi3 and Gpr143 results in a similar phenotype (Young et al., 2008(Young et al., , 2011. Constitutively active Gαi3 can rescue the Gpr143 knockout phenotype. While mouse Gpr143 interacts with Gαi3 (Young et al., 2008(Young et al., , 2011(Young et al., , 2013, the G protein with which human GPR143 interacts remains unclear Schiaffino et al., 1999). A potential direct physical interaction between GPR143 and TYR has also been proposed, since these proteins co-immuno-precipitate in a biochemical assay (Cortese et al., 2005). Whether this is the case in physiological conditions remains to be elucidated. In any case, GPR143 probably does not regulate TYR activity directly as in Gpr143 mutant mice Tyr activity is not impaired (Cortese et al., 2005).

Mutations
A variety of mutation types has been identified along the whole sequence of GPR143. GPR143 missense mutations can functionally be divided in two types: the first alters cellular distribution, the second type affects receptor signalling. Examples of the first type are: p.Q124R, p. A138V, p.S152N, p.G229V, p.T232K, p.E235K, p.I244V and p.E271G Addio et al., 2000). The second type of mutations (p.R5C, p. G35D, p.D78N, p.G84D, p.C116S, p.G118E, p.W133R, p.A173D, p. I261N, p.W292G and p.T290del) are mainly found in the second and third cytosolic loop (Fig. 5) (d' Addio et al., 2000;Innamorati et al., 2006;Schiaffino et al., 1999). Functional in vitro studies using COS-7 cells showed that GPR143 ligand subunits Gβ and Gαi did not bind at all (p.C116S), or hardly (p.D78N) to mutant proteins (Schiaffino et al., 1999). 23 nonsense and frameshift mutations are archived in the Leiden Open Variome Database (LOVD). They are found across the gene and most likely result in premature chain termination and loss of protein function.

GPR143: the molecular mechanism of ocular albinism
As described above, the OA1 phenotype, caused by mutations in GPR143, is limited to the eye. Both OA1 RPE and skin have relatively few enlarged macro-melanosomes (Lewis, 1993a;Schiaffino and Tacchetti, 2005), but only the RPE has decreased melanin content (Cortese et al., 2005;Schiaffino, 2010). The (defective) GPR143 receptor plays a key role in ocular albinism (McKay, 2019). In this section, we first discuss the ligands activating or inhibiting this receptor and next potential downstream events, such as secretion of growth factors and exosomes as well as the formation of macro-melanosomes.
One important ligand of GPR143 is L-DOPA (Lopez et al., 2008;McKay, 2019) (section 1.4, Fig. 2). We are aware of the multiple potential mechanistic actions of L-DOPA, which are reviewed elsewhere (De Deurwaerdère et al., 2017). However, In the context of this report, we only consider the role of L-DOPA in the proposed retinal pigmentation pathway. L-DOPA is produced when tyrosine is converted to melanin in the RPE melanosome (Hearing and Tsukamoto, 1991). L-DOPA interacts with GPR143, most likely at the apical surface of the RPE or at the lumenal side of the melanosome membrane (Basrur et al., 2003;Lopez et al., 2008;Piccirillo et al., 2006). The fundamental role of L-DOPA in albinism had been anticipated by several studies from Glen Jeffery's laboratory since 1998, stating its deficit in the eyes of persons with albinism and its role as a cell-cycle regulator (Jeffery, 1998). Thereafter, using rats as experimental model, they proposed that L-DOPA might regulate the time at which retinal cells exit mitosis (Ilia and Jeffery, 1999), rod photoreceptor production (Ilia and Jeffery, 2000), and cell cleavage orientation in the developing retina . They also confirmed similar properties associated with dopamine (Kralj-Hans et al., 2006). Experiments by  elegantly illustrated the role of L-DOPA in albinism associated retinal pathology. They constructed transgenic mice overexpressing TH under control of the Tyr promoter, in a spontaneously mutated Tyr defective strain (NMRI) . These mice are not able to convert tyrosine into melanin but are able to produce L-DOPA in the RPE (Fig. 2). Indeed, introduction of TH in the aforementioned mice rescued the neural retinal albinism pathology . Thus, the lack of L-DOPA production in the RPE appears to be crucial in the albinism retinal phenotype, while the lack of melanin is not. This hypothesis was corroborated by Murillo-Cuesta et al. (2010), who found, in the same albinism mouse strain that L-DOPA, produced through TH in the cochlea, rescued age-related and noise-induced hearing loss associated with albinism (not systematically described in albinism patients). L-DOPA is a ligand of GPR143 (Lopez et al., 2008). Genetic defects in GPR143 cause OA1, while retinal melanin is present (Cortese et al., 2005;O'Donnell et al., 1976;Wong et al., 1983). Given the aforementioned phenomena, we hypothesise that defects in GPR143 or its activation by L-DOPA is central in the retinal pathology of albinism.
The regulation of GPR143 activity is complex: L-DOPA (from RPE and neural retina) and dopamine (from neural retina) can modulate GPR143 activity in an excitatory or inhibitory way, respectively (Lopez et al., 2008). L-DOPA is present in embryonic stages of healthy retinal development in WT mice but not in albino animals (Roffler-Tarlov et al., 2013). Dopamine production starts postnatally in the neural retina (Roffler-Tarlov et al., 2013), when, at the same time, retinal development and maturation still continues (Boothe et al., 1985;Hendrickson et al., 2012;Tian, 2004). In addition, L-DOPA and dopamine production in the retina appears to be under circadian control (Baba et al., 2017). These regulatory aspects of the retinal pigmentation pathway are discussed in detail in section 9. We reviewed a number of clinical trials testing L-DOPA treatments in section 12.

Phenotypic characteristics of FHONDA
The FHONDA albinism-related ocular phenotype is generally more severe and more consistent between patients than any of the albinism subtypes. VA is worse than in OCA: median VA is 0.7 logMAR (IQR 0.6-0.8) for FHONDA while median VA for OCA1 is 0.5 logMAR (IQR 0.3-0.7) and for OCA2 0.5 logMAR (IQR 0.2-0.7). While foveal hypoplasia grades in OCA range from 0 to 4, foveal hypoplasia in FHONDA is limited to grade 3 and 4 (Kruijt et al., 2022;Thomas et al., 2011). Nystagmus and chiasmal misrouting have been uniformly present in reported FHONDA cases, compared to 93% and 84% detected in albinism respectively (Forsius et al., 1964). Anterior segment dysgenesis is only present in 19% of FHONDA patients, while general prevalence in ophthalmology clinic patients is between 7 and 32% (Forsius et al., 1964;Ozeki et al., 1997;Rennie et al., 2005). Thus, the association of this sub-phenotype is most likely due to chance. No pigment abnormalities are present in FHONDA patients. Our hypothesis is that FHONDA is at the end the proposed retinal pigmentation pathway (Fig. 3), suggesting that retinal development may be more severe and consistently affected, lacking potential upstream compensatory mechanisms present in (other types of) albinism. This is also in line with the clinical observations in FHONDA. Below we discuss what is known about the gene implicated in FHONDA, the protein and known pathogenic variants.

. Protein structure and function
Autosomal recessive FHONDA is caused by mutations in SLC38A8 (Poulter et al., 2013). The gene has nine exons resulting in a 1672 bp transcript (NM_001080442.3). It encodes a 435 aa protein with 11 transmembrane domains (Fig. 5). SLC38A8 belongs to the SLC38 family of sodium-coupled neutral amino acid transporters (Mackenzie and Erickson, 2004). Indeed, SLC38A8 has a preference for sodium dependent transport of glutamine, alanine, arginine and histidine in the central nervous system (CNS) (Hägglund et al., 2015). In mice, RNA expression of SLC38A8 was found in the eye but not in the skin (in contrast to many albinism disease gene products) (Perez et al., 2014). In human donor eyes, immunohistochemical staining showed protein expression of SLC38A8 in the neural retina and possibly the RPE. In addition, protein expression was found in human foetal and adult brains, in neuronal cell bodies and axons as well as in the spinal cord (Poulter et al., 2013). Expression in the adult retina, and other parts of the CNS, suggests that SLC38A8 function is not limited to neural development alone. In genetically modified mice, Slc38a8 mutations recapitulate the phenotype observed in humans, with distorted optical tract but without pigmentary alterations (Lluis Montoliu, unpublished data).

Genetic variation, mutations and mechanism
So far, 37 mutations in SLC38A8 have been implicated in FHONDA (Fig. 5). 21 of them are missense mutations. 16 out of 21 missense mutations possibly affect transport function, as they are found in or close to transmembrane domains. Three missense mutations, p.T87I, p.E233K and p.D283A, affect hydrophobicity near the channel pore, potentially disrupting sodium dependent transport. Yet another mutation, p.E233K, affects hydrogen bonding capabilities and possibly protein folding (Poulter et al., 2013;Toral et al., 2017). Two mutations, p.D283A and p. A282del, are located in the fourth extracellular loop and may affect subcellular localisation of the protein (Poulter et al., 2013;Toral et al., 2017). In total, 14 loss of function mutations have been identified, leading to premature chain termination (Kruijt et al., 2022).

Expression of genes implicated in albinism in human RPE, neural retina and retinal organoids
In the previous sections we discussed all currently known disease genes implicated in (ocular) albinism and FHONDA, and their respective phenotypes. In order to have an impression when function or molecular pathology could set in, we summarised and compared five RNA expression datasets of (disease) genes involved in the different types of albinism. Three out of five sets contain data from all major human retinal cell types together, including RPE, during development ( Fig. 6  and Fig. 7). Two out of these three are derived from wild-type human retinal organoid in vitro models (GSE119274) (Kim et al., 2019;Wagstaff et al., 2021b), and one out of three is a healthy human foetal retina developmental control data set (GSE104827) (Hoshino et al., 2017). Note that in these samples the relative size of the whole retina increases over time compared to the RPE and thus in these datasets the relative expression of RPE genes will also generally decrease. It is difficult to directly compare the expression data sets, since the two human retinal organoids sets were generated with different protocols and the human foetal expression data suffers from limitations intrinsic to using human donor material. Also, in all three datasets, not all relevant genes were present, expression was calculated in different ways, and time points studied varied. Next to the aforementioned organoid and foetal datasets, we also include embryonic stem cell (ESC) derived RPE-only expression data sets to which the same study limitations apply (Supplementary Fig. 1 and 2) (Bennis et al., 2017;Smith et al., 2019). Obviously, the expression data presented are important for comparison of human organoid models with the physiological situation, and are relevant for the construction of new models. Adding more -omics datasets like this in the future will give a more complete picture of the aetiology of human albinism and related disorders.

RNA expression of genes implicated in syndromic albinism
As proposed in section 3, syndromic forms of albinism involve a disruption of general cellular processes related to LROs generation or function. LROs are present in many different cell types and tissues resulting in a wide range of symptoms affecting patients. From the earliest stages of (retinal) development onward, all disease genes implicated in syndromic forms of albinism seem to be consistently expressed, albeit at different levels, in all datasets ( Fig. 6; Supplementary Fig. 1). Mostly, this expression does not change dramatically over time. In the full retina entities, relatively high expression is found for AP3D1 and increasingly for LYST, and perhaps BLOC1S6. Tentative lower expression is found for HPS1, HPS6, DTNBP1 and BLOC1S3 (Fig. 6). Data from the RPE-only expression datasets suggest that AP3D1 and BLOC1S6 are relatively high expressed in RPE ( Supplementary  Fig. 1). These data confirm the very early onset of expression of disease genes involved in syndromic albinism in human (retinal) development. Indeed, as these genes have a role in multiple LROs in multiple cell types, it is expected that they are ubiquitously present.

RNA expression of genes implicated in oculocutaneous albinism
The genes implicated in OCA, described in section 4, are most likely involved in the function and pathology of pigment cells. Using the same aforementioned gene expression datasets, we analysed the timing and level of OCA disease gene expression in the developing retina (Fig. 7) and RPE (Supplementary Fig. 2). In contrast with the disease genes involved in syndromic forms of albinism, expression of OCA genes appears to be relatively low and remains so during early stages of retinal  (Wagstaff et al., 2021b) shows relatively high expression of all genes implicated in HPS and CHS at all developmental time points. (B) Expression of these genes is more variable in the other retinal organoid development dataset from Kim et al. (2019) using another differentiation protocol. However, data confirm that expression of all genes is present at all time points. (Note that no expression data was available for BLOC1S5 and BLOC1S6 in this dataset) (C) In foetal retina expression is also present during all developmental stages (Hoshino et al., 2017). The early onset and high RNA expression of the genes implicated in syndromic forms of albinism confirm their essential role from a very early developmental stage onward. Indeed these genes appear to be already involved in LRO development in largely undifferentiated embryonic stem cells (start of retinal organoid differentiation) and possibly multiple cell types present (in the retina), including the RPE with its melanosomes. Of note, the relative cellular contribution of the RPE compared to the expanding neural retina becomes less during development of the whole organoid samples and foetal retina. This may affect the relative gene expression measured in the aforementioned tissues, especially for cell-type specific genes. This potential effect is not obvious in the gene expression patterns observed here. development (Fig. 7). For example, the expression of TYR, SLC45A2 and SLC24A5 in all datasets is somewhat variable but generally low during all developmental stages studied. This might confirm that, in OCA, the pathology of pigment cells uniquely is implicated, which develop at later stages. Apparent exceptions are: DCT, which is relatively highly expressed at all time points in all data sets; and OCA2, which is relatively highly expressed in two out of three datasets. The high expression of DCT in comparison to, for example, TYR that functions in the same melanin production pathway, might reflect the hypothesis that DCT is involved in additional functionalities. These possibly include the clearing of toxic quinone metabolites in the retina (Michard et al., 2008), or neural progenitor proliferation (Jiao et al., 2006). Interestingly, although OCA2 levels are different in the models, the OCA2 expression pattern seems to rise, peak, fall, and then perhaps stabilise during development. It is tempting to speculate that this is correlated with the transient role OCA2 plays in melanosome development (Bellono et al., 2014). While most OCA genes are expressed in the RPE-only data sets (Supplementary Fig. 2) and, most recently, in preliminary expression data provided by George et al. (2022) and Bakker et al. (data not shown) in iPSC derived RPE cultures, no further firm conclusions or hypotheses can be drawn on the basis of currently available RNA data.

RNA expression of GPR143
The GPR143 protein is localised to the RPE apical membrane or to the RPE melanosomal compartment, or both (for a full description see section 5.2). GPR143 controls melanosome formation in pigment cells and is implicated in the OA1 phenotype that is largely restricted to the eye. As we discussed elsewhere in this report (sections 2, 5.3, 8-10) the protein plays a central role in the retinal pigmentation pathway.
We observed a temporal decrease of GPR143 gene expression in all aforementioned datasets during full retinal development (Fig. 8). However, in the RPE-only datasets, GPR143 expression seems to be stable during retinal development. Possibly these expression patterns  (Hoshino et al., 2017) (C). In general, expression of OCA disease genes is relatively low. TYR gene expression appears to be lower than TYRP1 expression, at least for a number of time-points. DCT expression is high in all datasets but more variable over time in the dataset by Kim et al. (2019) and foetal retina. SLC45A2 and SLC24A5 appear to have a very short window where expression is present. OCA2 expression appears to have a transient peak and then stabilises in all datasets. Note that no data was available for LRMDA expression and expression is virtually absent in our organoid expression data set, while expression of this gene was not reported in the organoid dataset of Kim et al. (2019) and the foetal retina dataset (Hoshino et al., 2017). Of note, the relative cellular contribution of the RPE compared to the expanding neural retina becomes less during development of the whole organoid samples and foetal retina. This may affect the relative gene expression measured in the aforementioned tissues, especially for cell-type specific genes.
reflect the expanding cellular contribution of the neural retina (where GPR143 is not present) compared to the RPE monolayer (and thus GPR143 gene expression) to the developing retina. Obviously, further studies are needed to shed light on this issue.

RNA expression of SLC38A8
SLC38A8 localisation and expression has been found in the neural retina (Poulter et al., 2013). When mutated in FHONDA patients, it causes a severe albinism-like phenotype restricted to the neural retina. Pigmentation is not affected in FHONDA. Therefore, we hypothesised it . Note that SLC38A8 was not analysed in the data sets of Bennis and Wagstaff. GPR143 expression in the retinal organoid (A1, A2) and human foetal retina (A3) seems to follow an overall pattern of high expression of GPR143 in early development and lower (or undetectable) in later stages of retinal development. This expression pattern may possibly be explained by the fact that GPR143 has been localised to the RPE, and relative cellular contribution of the RPE compared to the expanding neural retina becomes less during development. Indeed, the Smith data (A4) suggest comparable expression in iPSC-derived RPE and foetal retina (albeit at one time point) and the Bennis expression data (A5) suggest comparable GPR143 expression in stem cell derived early and more mature RPE stages. SLC38A8 expression (B1) is present in the (cone-enriched) organoid differentiation expression data by Kim et al. (2019) but it seems to be variable. (B2) In human foetal retina SLC38A8 expression is relatively low and also variable (Hoshino et al., 2017). Interestingly, the Smith SLC38A8 expression data (B3) suggests that SLC38A8 transcription takes place in the RPE, while the SLC38A8 protein was previously localised to the neural retina (Poulter et al., 2013). Further data is needed to discern whether SLC38A8 is expressed in multiple retinal cell types, or whether its RNA or protein is transported from RPE to neural retina.
to function downstream from GPR143 in the retinal pigmentation pathway (section 2). Again, we reviewed and compared the temporal retinal RNA-seq expression data of SLC38A8 during development in two organoid and one foetal datasets (Fig. 8) (Hoshino et al., 2017;Kim et al., 2019;Wagstaff et al., 2021b). Remarkably, no SLC38A8 expression was detected in our own retinal organoid differentiation data (Wagstaff et al., 2021b) while it is temporally variable in the second organoid expression data set (Fig. 8B1) (GSE119274) (Kim et al., 2019), as well as in foetal retina (Fig. 8B2) (GSE104827) (Hoshino et al., 2017). Given the potential role of SLC38A8 cone specification in patients , it is of interest to speculate that the expression of SLC38A8 is a bit elevated in the cone-rich organoid data (Kim et al., 2019), compared to the Wagstaff retinal organoid data, where it is virtually absent (not shown); an issue that needs further investigation. Perhaps even more surprisingly, in the RPE-only dataset of Smith et al. (2019) RNA expression of SLC38A8 was detected. This raises the intriguing possibility that SLC38A8 may be transcribed in the RPE, and that transcripts or protein is transferred to the neural retina. Alternatively, SLC38A8 may be transcribed in more than one retinal cell type. Future detailed studies need to be performed to shed light on this issue. Of note, SLC38A8 expression is below detection level at some time points in foetal retina, so it may be that expression in the other datasets is temporarily present between the actual time-points measured.

GPR143-activated paracrine RPE growth factor signalling
In the retina, expression of GPR143 is limited to the RPE (Bharti et al., 2006;Surace et al., 2000). However, foveal hypoplasia and chiasmal misrouting are deficiencies of the neural retina. In this section, we describe the potential (indirect) regulatory role of GPR143 in paracrine signalling from the RPE to the neural retina. Paracrine signalling in the proposed retinal pigmentation pathway may consist of PEDF, VEGF secretion and exosome release from the RPE, which are discussed one by one below.
In the healthy RPE, both GPR143 and PEDF RNA expression are present during retinal development in retinal organoids (Wagstaff and Bergen, unpublished). In the healthy (foetal) situation, PEDF is secreted mainly from the apical RPE toward the neural retina ( Fig. 9) (Fields et al., 2017;Maminishkis et al., 2006). We hypothesise that this is (in part) stimulated by GPR143 activation (Falk et al., 2012;Lopez et al., 2008). In syndromic or oculocutaneous albinism, the conversion of tyrosine to L-DOPA (Fig. 2) is decreased and local L-DOPA levels are depleted. We hypothesise that this leads to reduced activation of the GPR143 receptor and less apical PEDF secretion from the RPE. Indeed, in human primary RPE cultures, reduction in PEDF secretion caused by pharmacological blocking of TYR can be rescued by L-DOPA supplementation (Falk et al., 2012;Lopez et al., 2008).
PEDF is a neurotrophic and antiangiogenic factor primarily secreted apically by foetal RPE (Fields et al., 2017;Maminishkis et al., 2006). Its antiangiogenic role may be important for the development of the foveal avascular zone and the subsequent formation of the fovea (Provis et al., 2013). Consequently, the potential lack of PEDF secretion in albinism may contribute to an underdeveloped macular area and foveal hypoplasia. Furthermore, PEDF has also been implicated in long range axon formation (Barnstable and Tombran-Tink, 2004). In addition, PEDF promotes cell survival in immature cerebellar granule cell cultures from chick spinal cord but not in mature granule cells (Araki et al., 1998). This suggests a potential neurotrophic role for PEDF in neuronal and possibly retinal development.
In contrast to increased apical PEDF secretion, activation of GPR143 may lead to decreased VEGF secretion from human primary RPE cultures (Blaauwgeers et al., 1999;Falk et al., 2012). VEGF is secreted mainly from the basolateral side of the foetal RPE (Fields et al., 2017;Maminishkis et al., 2006). In the healthy developmental situation, VEGF secretion has been implicated in the development of the choroid, and, in low levels, in development and maintenance of the foveal avascular zone (Provis, 2001). In the adult situation, basolaterally secreted VEGF acts on the choroid and maintains retinal differentiation. In the elderly, basolateral RPE secreted VEGF plays a key role in choroidal neovascularization (in age-related macular degeneration; AMD) (Adamis et al., 1993;Blaauwgeers et al., 1999;Witmer et al., 2003). Whether VEGF is also secreted in significant levels from the apical RPE, and what the implications would be for neural retinal development and disease, is not clear. In albinism, we hypothesise that low L-DOPA levels could lead to less activation of the GPR143 receptor, and therefore locally increased VEGF secretion. In albinism, excess VEGF could possibly promote vascularisation, impede formation of the avascular zone, and contribute to hypoplasia of the fovea. Of note, GPR143 signalling also regulates MITF expression in the RPE. MITF is a transcription factor not only controlling melanogenesis but also PEDF expression (Zhu et al., 2011). Thus, the potentially increased release of PEDF seen after GPR143 activation is possibly also modulated by MITF.

Apical RPE exosome release and GPM6A
Following the retinal pigment pathway hypothesis, GPR143 activation possibly leads to a decrease of secreted exosomes from primary RPE cultures (Locke et al., 2014;Lopez et al., 2008). Exosomes develop through fusion of intracellular multivesicular endosomes with the plasma membrane that result in release of vesicles from the cell (Lo Cicero et al., 2015;Raposo and Stoorvogel, 2013). RPE exosomes carry a diverse range of biomolecules (lipids, proteins, RNAs) that are key to In summary, GPR143 localises to the RPE plasma membrane, the melanosome and endosome (for detailed explanation, see section 5.2). For the sake of mechanistic clarity, we show here only the apical GPR143 activation. GPR143 activation leads to increased apical secretion of PEDF and a decreased secretion of VEGF (primarily basolateral) by the RPE. PEDF is involved in inhibiting vascular growth, VEGF has been implicated in neo-vascularisation. Exosomes are secreted from the RPE in a polarised fashion from the RPE. Of note, apical RPE exosomes contain different sets of biomolecules than basal exosomes. Here, only the albinism-relevant apical exosome release is shown. Experimental data suggests apical exosome release decreases upon GPR143 stimulation. Altered release of PEDF and exosomes (and possibly also VEGF) toward the neural retina could contribute to the neural retinal pathology of albinism. intercellular signalling and processing of cellular waste (Colombo et al., 2014;van der Pol et al., 2012). In the healthy condition, the diameter of exosomes ranges between 30 and 100 nm (Lo Cicero et al., 2015).
RPE exosomes are released both apically, toward the neural retina, and basolaterally toward the Bruch's membrane and choroid (Klingeborn et al., 2018). The release of exosomes from (adult primary porcine) RPE is polarised as many proteins found in apical exosomes are different from basolateral ones (Klingeborn et al., 2017). Whether exosomes are released at similar rates and with similar contents during development remains to be elucidated. For our present analyses, we used the best data sets available on RPE exosome contents. Possibly, apical RPE exosome release and signalling are involved in proper retinal development and function and they are, at least in part, likely to be relevant to the ocular albinism phenotype. Given the abnormal development of the neural retina in albinism, we focus here on the apical RPE exosome release towards the neural retina.
Interestingly, RPE cells, like other cell types, increase release of exosomes upon stress (Atienzar-Aroca et al., 2016;King et al., 2012). The RPE in albinism may be in a continuously stressed state due to its inability to form pigment that otherwise photo-protects the tissue. In donor eyes with vitreous and neural retina removed, exosome release is halted by supplementation of excess L-DOPA possibly through GPR143 activation (Locke et al., 2014). Note that this study is performed in adult donor eyes and the developmental situation may differ. Conversely, GPR143 may be inactivated by genetic modification or dopamine supplementation, which may ultimately lead to increased exosome release. What role these exosomes may play in healthy retinal development and in albinism depends on their content. Below we use a bioinformatics approach to identify RPE exosome constituents potentially relevant for the ocular pathology of albinism.
Limited by the data yet available, we aim here to identify potential relevant exosome bio-molecules in the context of albinism. Klingeborn et al. (2017) previously identified 55 unique proteins in the apical RPE exosomes of first passage primary porcine RPE cultures (obtained from adult eyes) (Supplementary Table 1). In order to identify albinism relevant entries, we compared this apical exosome protein set (Supplementary Table 1) with a list of gene ontology entries related to both neural development and potentially involved in albinism pathology (Supplementary Tables 2A and 2B). The overlap between these data sets (Supplementary Tables 1 and 2B) yielded a single entry: Glycoprotein M6A (GPM6A) (Fig. 10). Interestingly, substantial in vitro and in vivo experimental evidence support the notion that GPM6A may be highly relevant for retinal and CNS development and maturation. Additionally, taken the bioinformatic and experimental data together, GPM6A may also potentially be relevant for neural albinism pathology. This evidence consists of two relevant groups: (1) studies directly involving the developing (neural) retina or optic nerve and (2) neuronal studies involving that other part of the CNS: the brain.
Overexpression of Gpm6a in mouse retinal progenitor cells resulted in enhanced neurite outgrowth . Indeed, in mice, mutations in Gpm6a resulted in hypoplasia of axon tracts in the brain (Mita et al., 2015). In the neonatal mouse, Gpm6a protein is present in RGC axons . In adult mouse retina, Gpm6a has been localised to the inner retinal layers . In a study involving of the Drosophila GPM6A orthologue M6, mutants showed a defective visual system development: reduced M6 levels triggered mild eye defects, including a disorganised ommatidium array, defects in ommatidium shape and/or ommatidium fusion, defective or missing bristles and a reduced phototactic response to light (Zappia et al., 2012).
In culture, GPM6A has been implicated in filopodia formation in neuronal cells (Honda et al., 2017). Filopodia are cellular projections that have been linked to neural differentiation and projection as well as dendrite and synapse formation (Brocco et al., 2010). Genetically modified cultured mouse stem cells showed that Gpm6a is involved in the differentiation of neurons (Michibata et al., 2008(Michibata et al., , 2009. Also, Gpm6a dynamically supports dendritic spine and synapse formation in mouse primary hippocampal neuronal cultures (Brocco et al., 2010;Formoso et al., 2016).
Taken together, experimental in vitro and in vivo as well as in silico data support our hypothesis that GPM6A is involved in retinal or optic tract (projection) development and can thus be relevant for the development of the neural pathology in albinism, downstream in the proposed retinal pigmentation pathway. Obviously, this hypothesis needs further substantiation in future studies.

GPR143-activated autocrine RPE growth factor signalling
According to our proposed retinal pigmentation pathway, downstream GPR143 signalling may result in PEDF and VEGF secretion from the RPE. This has not only an effect on the adjacent tissues, but also an autocrine effect on the RPE itself. These effects have both been studied in in vitro and in vivo models. Obviously, in vitro data should be considered with caution, and results obtained depend on research questions posed, study design and biomarkers used. For example, cultured ARPE-19 cells may not respond to a VEGF stimulus in the same manner as primary foetal human RPE cultures do. Also, ARPE-19 cells possibly behave more like aged RPE, while foetal (human) RPE appears to be more representative for healthy young human RPE (Ablonczy et al., 2011). Nonetheless, both are relevant to study. The autocrine effects of PEDF and VEGF on the RPE are summarised below (Supplementary Fig. 3).
Autocrine effects of RPE PEDF secretion may be beneficial in development and homeostasis of the retina. We will discuss first the role of PEDF in development and next its role in homeostasis. Beneficial developmental effects of PEDF supplementation to primary rat RPE cultures include larger RPE cell bodies, enhanced development of tight junctions to neighbouring cells, increased number of phagocytic vesicles and more melanosomes with increased maturity, and PEDF expression was correlated with human embryonic stem cell-derived RPE maturity (Al-Ani et al., 2020;Malchiodi-Albedi et al., 1998). Experimental evidence also suggests that PEDF positively affects mature RPE homeostasis: first, PEDF protects cells including rescued barrier function and mitochondrial function from sodium-iodate or H 2 O 2 induced oxidative stress Ho et al., 2006;Nadal-Nicolas and Becerra, 2018;Subramanian et al., 2016;Wang et al., 2019). Second, PEDF allows the RPE to form a stable monolayer by inhibiting proliferation and migration in vitro (Farnoodian et al., 2015;Ma et al., 2012b). Finally, in PEDF receptor (PNPLA2) knockout models, including mice and choroidal or RPE explants, photoreceptor outer segment phagocytosis was low Fig. 10. Venn diagram of the comparison between the apical RPE exosome secreted proteome dataset (Klingeborn et al., 2018); yellow circle on the left) and genes related to neural retina development (annotation see Supplementary  Table 1; green circle on the right). Both phenomena may be, independently, implicated in human albinism. We found, data-driven, that one single entry was shared among the two datasets: Glycoprotein M6A (GPM6A). GPM6A is a CNS specific trans-membrane protein, a member of the PLP/DM20 proteolipid family. In the developing healthy mouse retina, Gpm6a is expressed by RGCs, and the encoded protein product was localised to RGC axons (366). The protein acts as a nerve growth factor (NGF) dependent Ca-channel (367). GPM6A has functionally been implicated in neuronal migration, differentiation and projection in the CNS. Indeed, GPM6A is a good candidate signalling molecule involved in retinal development and albinism (see text). compared to controls (Bullock et al., 2021). Taken together, the autocrine effect of PEDF signalling on the RPE itself appears essential for proper function and homeostasis of the RPE.
Not only PEDF, but also mechanistically coupled VEGF secretion from the RPE plays an important role in both the healthy and diseased RPE state. Indeed, some VEGF autocrine stimulation of RPE seems essential and has a (positive) effect on RPE homeostasis, as illustrated by the following examples: First, VEGF plays a critical role in survival and maintenance of RPE integrity shown in murine retina eye developmental studies (Ford et al., 2011). Anti-VEGF treatment of ARPE-19 and long-term anti-VEGF treatment of primary porcine RPE in vitro cultures cause a decrease in cell viability (but also reduces RPE65 expression) (Brinkmann et al., 2022;Tsujinaka et al., 2015). Last, VEGF may enhance the intrinsic protective capability of the RPE to deal with oxidative stress. On one hand exposure to cigarette smoke, advanced glycation end products and lipid peroxidation products induced oxidative stress and also induced expression of VEGF (Bergmann et al., 2011;Bertram et al., 2009;McFarlane et al., 2005;Tsujinaka et al., 2015). Additionally, the oxidative stress stimulated VEGF induced its own expression in an autocrine manner (Rossino et al., 2020). On the other hand, in oxidatively stressed ARPE-19 cell cultures (by cigarette smoke), VEGF treatment increased phagocytosis and rescued cell death (Chu et al., 2013). Indeed autocrine VEGF signalling enhanced ARPE-19 cell survival under oxidative stress (by H 2 O 2 ) (Byeon et al., 2010).
In contrast to the aforementioned positive effects, increased VEGF expression may also have detrimental effects on barrier function and maintenance of cellular RPE identity: Indeed, VEGF reduces RPE barrier function via VEGF receptor 2 (VEGFR2) activation in primary porcine RPE and ARPE-19 cells, as measured using trans-epithelial resistance (Ablonczy and Crosson, 2007). There is a molecular cross-talk in the RPE between PEDF and VEGF inside the RPE, with PEDF inhibiting VEGFR2 signalling, securing its barrier function (Ablonczy et al., 2009(Ablonczy et al., , 2011 and also VEGF induces PEDF expression via VEGFR1 (Ohno-Matsui et al., 2001) suggesting a regulatory interaction between the two systems. This counterbalance was shown in RPE from smoker patients, where VEGF-to-PEDF ratio was increased in human RPE from smoker patients with AMD (Pons and Marin-Castaño, 2011). Further, dysregulation of RPE barrier function by VEGF can result in retinal oedema in rabbits (Dahrouj et al., 2014). Finally, in proliferative vitreoretinopathy, the RPE contracts, proliferates and migrates. In adult human primary RPE cultures, VEGF induces contraction, proliferation and migration of RPE cells and the RPE acquires a more mesenchymal phenotype (Hoffmann et al., 2005b;Kehler et al., 2012;Ma et al., 2012a).

Regulatory aspects of the retinal pigmentation pathway
Above, we proposed a new functional genetic classification of albinism, and formulated a retinal pigmentation pathway (section 2, Fig. 3). We also described details of this pathway one by one. It is now of interest to elaborate further on the regulatory aspects that, in part, determine the activity of the proposed pathway (Fig. 11) and may contribute to the variability seen in the albinism phenotype. The activity of regulatory hubs in the pathway may be driven by variation in (a combination of) genetic factors, as well as biochemical, metabolic and environmental cues. Main regulatory entities include competitive activation and inhibition of GPR143 by L-DOPA/dopamine, as well as the regulation of (macro)melanosome formation and oxidative stress. Below, we will discuss the former two entities, while oxidative stress is discussed in the context of AMD and retinopathy of prematurity (ROP) in section 12.4.

Competitive inhibition and activation of the GPR143 receptor
As described briefly above (section 5.3), the activation of the GPR143 receptor by its ligand L-DOPA, and competitive inhibition by dopamine could play an important role in the proposed retinal pigmentation pathway (Fig. 11) (Falk et al., 2012;Locke et al., 2014). Please note that dopamine is not present during embryonic stages of retinal development (Roffler-Tarlov et al., 2013). In addition, we assume most activity of GPR143 takes place at the plasma membrane as that is typical for GPCRs (Pavlos and Friedman, 2017). However, if the main site of activation of GPR143 in vivo is in the melanosome competitive inhibition by dopamine may be less or not present. Previous studies in Fig. 11. Four (potential) regulatory aspects of the retinal pigmentation pathway (itself depicted in Fig. 2): (half shaded boxes) circadian control of the activity of the pathway; (1) variation in L-DOPA and dopamine release, and their effect on GPR143 signalling; (2) regulation of melanosome formation through GPR143 signalling and circadian rhythm; (3) reduced capacity of the RPE to handle and process oxidative stress (over time). Obviously, the regulation of the activity of the proposed pathway involves many other factors, including intrinsic (age-related factors, pleiotropic effects and additional feedback loops), genetic (MITF activity, relevant genetic variants in disease genes or regulatory sequences), and environmental or metabolic factors (diet-driven), which are areas of future research. Here, we focus on the aforementioned aspects. Half shaded boxes: L-DOPA and dopamine are, respectively, receptor activating and inhibiting ligands of GPR143. L-DOPA is produced by the RPE melanosome as tyrosine is converted into melanin. L-DOPA and/or dopamine is also secreted from the (adult) neural retina to the RPE during the day. The aforementioned processes are under circadian control (Hardman et al., 2015) (shaded boxes) and therefore may regulate the activity of the retinal pigmentation pathway. (ad 1) L-DOPA and dopamine competitively modulates GPR143 signalling. (ad 2) GPR143 intra-cellular feedback signalling in RPE cells result in the production of macro-melanosomes and reduced melanin production, which also affects the local L-DOPA production. Thus, the effectiveness of this feedback loop (different in retina and skin) may constitute a regulatory aspect of the activity of the pathway. (ad 3) Finally, oxidative stress, partly due to reduced pigmentation, affects general health of the RPE by oxidising a large range of local biomolecules (discussed in section 12.4). However, oxidative stress may also harm the specific structure or function of the melanosome, and thus the specific (protein) activity of the proposed retinal pigmentation pathway.
foetal retinal tissue and animal models showed that L-DOPA has an essential regulatory role regarding cell fate and differentiation in retinal development (Kralj-Hans et al., 2006;Reis et al., 2007). Excess L-DOPA produced in the initial step of melanogenesis is released from the RPE to the subretinal space. To complicate matters, in the adult situation, L-DOPA and dopamine can also be released from the neural retina to the subretinal space (Reis et al., 2007). In the neural retina, L-DOPA is a precursor for the neurotransmitter dopamine. Indeed, in contrast to the neuro-epithelial RPE, dopaminergic neurons in the retina express the enzyme TH. This enzyme is essential for the direct conversion of L-DOPA into dopamine (Fig. 2) (Reis et al., 2007). If dopamine is released from the neural retina and reaches the RPE (Rajan et al., 2000), it could compete with and block binding of L-DOPA to GPR143 and inhibit downstream signalling (Lopez et al., 2008). At the same time, L-DOPA could be released from the neural retina and stimulate GPR143. Unlike L-DOPA, dopamine does not activate a downstream G protein-coupled receptor response along with characteristic Ca 2+ influx into the cell (Lopez et al., 2008). Taken together, we hypothesise that L-DOPA is released into the subretinal space by both the RPE and neural retina, activating GPR143, while dopamine is released from the neural retina, competitively inhibiting GPR143. L-DOPA and dopamine, exerting opposite effects, are therefore locally competing for the GPR143 binding site. Uptake of dopamine by the RPE would remove it from the subretinal space where the ligand-binding domain of GPR143 is located and decrease its inhibitory effect. A shift in L-DOPA to dopamine ratio in the subretinal space could possibly result in a shift in activation/inhibition of GPR143.
Dopamine release from retinal neurons follows a circadian rhythm (Jackson et al., 2011;Jin et al., 2015). During the day, active neurons secrete dopamine that builds up in the retina and is (partially) released toward the RPE (Rajan et al., 2000). On the surface of the RPE, dopamine can inhibit GPR143 activity. During the night, neurons are less active and produce less dopamine. Possibly, this mechanism controls the activity of the pigmentation pathway to uniquely accommodate for light induced stresses of the retina during the day. After all, during the night the retina does not have to process light-induced stressors. In addition, both L-DOPA and melanin production by the RPE follow a circadian rhythm. Expression of melanogenic enzymes (TYR, TYRP1, DCT) and production of melanin in hair follicle melanocytes are under circadian control, regulated by the light dependent expression of the clock genes BMAL1 and PER1 (Hardman et al., 2015). These circadian clock genes are active during the day, and melanogenesis is suppressed during this period. The circadian clock appears to be active in the RPE as well and could therefore possibly regulate L-DOPA and melanin production (Bagchi et al., 2020;Milićević et al., 2021). Whether the aforementioned clock regulated processes act in sync, is currently not clear.
Nonetheless, this circadian control should be taken into account when designing experiments studying the regulation and activity of the proposed retinal pigmentation pathway or experimental therapies. Of note: since albinism pathology starts early in retinal development in the womb, and relevant foveal maturation occurs after birth, the circadian regulation of the retinal pathway may be different in these two phases of life. Clearly, this issue also warrants further investigations.

GPR143 and its role in formation of macro-melanosomes
GPR143 activation and downstream signalling most likely controls intracellular formation of vesicles, as highlighted by at least two pathological OA1 features. These are the formation of fewer and abnormally enlarged melanosomes (Cortese et al., 2005;Schiaffino and Tacchetti, 2005) and, possibly, the affected RPE exosome vesicle release (section 8.3) (Locke et al., 2014). The macro-melanosome pathology appears to be more severe in RPE and choroid than in hair or skin (O'Donnell et al., 1976). In Gpr143 − /− mutant mice, fewer early melanosomes are produced which subsequently show overgrowth as they mature (Cortese et al., 2005). As previously published, GPR143 may control early melanosome formation in two ways: first, in early stages of healthy melanosome development, GPR143 signalling prevents lysosomal fusion of late endosomes or melanosome precursors (Burgoyne et al., 2013), so that multiple distinct melanosomes can form. Consequently, GPR143 inactivation (in OA1) would lead to fewer enlarged melanosomes. Second, GPR143 signalling may also regulate the expression of MITF, a master regulator of the RPE and melanocyte development (Levy et al., 2006). MITF (co-)controls the expression of genes encoding melanosomal proteins, such as TYR and the structural fibril protein PMEL. In theory, lower levels of GPR143 signalling and a subsequent decrease of MITF activation can lead to lower amounts of PMEL, which allows fewer melanosomes to be formed (Burgoyne et al., 2015;Levy et al., 2006;Ma et al., 2019). Falletta et al. (2014) proposed that fewer early-stage melanosomes, as controlled by GPR143 signalling, lead to an increased ratio of melanosomal proteins to melanosome number. This would lead to proteins accumulating in fewer melanosomes with more pigment production and increased growth compared to normal sized melanosomes (Cortese et al., 2005;Falletta et al., 2014).

Do mechanisms underlying foveal hypoplasia and optic tract misrouting overlap?
Obviously, foveal hypoplasia and misrouting are two distinct entities with many differences. These include clinical, anatomical, cellular, molecular and functional phenomena. Molecular differences relate to molecular and developmental cues affecting the development of the retina and optic tracts (Fig. 12). Functional differences can be defined as developmental timing and projection of the optic tracts. The subject of foveal formation (Provis et al., 2013) and optic nerve (mis-)routing (Hoffmann et al., 2005a) has been reviewed elsewhere in depth.
Considering the clinical and pathological overlap between the various forms of albinism in the context of our single proposed pathway, the question arises whether foveal hypoplasia and optic nerve misrouting share one or more common molecular or cellular mechanisms. Indeed, these two entities are consistent defects observed among all genetic albinism subtypes, suggesting that these are intrinsically linked in albinism disease pathology and part of a final common pathway. FHONDA, with its most restricted but severe albinism-like phenotype and with its corresponding SLC38A8 disease gene expression in the neural retina and not in the RPE (Poulter et al., 2013), suggests a (partly) shared mechanism for foveal hypoplasia and chiasmal misrouting originating in the neural retina.
A vast amount of data on this subject comes from previous animal studies. It should be noted here that in a number of animal models of albinism investigated, the ocular phenotype is slightly different from the one observed in humans. For example, in rodents, the fovea does not exist. However, there is an increase of photoreceptors in the central retina, which can be quantified and used to compare differences between albino and wild-type murine retinas (Jeffery et al., 1997). In rabbits, the fovea is represented by a visual streak, whose presence is significantly reduced in albino animals (Jeffery et al., 1997). Various degrees of optic tract misrouting exist in different animal models which are usually investigated by anterograde labelling methods (Jeffery et al., 1994) or optokinetic tests . In the context of this article, we only summarise the most relevant aspects of the development of the optic nerve (section 10.1), misrouting of the optic tracts (section 10.2), development of the fovea (section 10.3), and its hypoplasia in albinism (section 10.4) to address the main specific question here: In the context of the proposed retinal pathway, do a number of mechanisms underlying foveal hypoplasia and optic tract misrouting overlap?

A summary of molecular/cellular mechanisms underlying development of the optic nerve
The retina transmits visual information to the brain via the RGCs.
RGCs are situated in the innermost layer of the retina, and receive their input from the photoreceptors via the bipolar and amacrine cells. They possess long axons encapsulated by glia to form the optic nerve. The optic nerves extend to the optic chiasm where they partially decussate. Most of the axons of the optic nerve terminate in the lateral geniculate nucleus. There, information is relayed to the visual cortex, while other axons terminate in the pretectal area. In humans, RGCs are one of the earliest-born cell types of the future neural retina, developing between foetal week five and foetal week 18 (Pacal and Bremner, 2014). RGC genesis follows a wave of Sonic hedgehog (Shh) signalling (Wagstaff et al., 2021a) developing first in the central retina before spreading to the periphery (Erskine and Herrera, 2014;McCabe et al., 1999). Once born, RGC axons need to be guided out of the retina into the optic nerve. This is achieved using a spectrum of intra-and extracellular signalling molecules alongside physical cues that attract, repel or guide RGC axons (Deiner et al., 1997;Erskine and Herrera, 2014). Examples of these signalling molecules are the aforementioned Shh; the family of Slit proteins, Netrin-1, Sema5A (Fig. 12) (Deiner et al., 1997;Erskine and Herrera, 2014) and possibly, as described in section 8.2 of this report: GPM6A.

Fig. 12.
Overview of axon guidance throughout the retina and optic tract. As ganglion cells are born, guidance cues are present already in the retina to direct outgrowing axons through the proper path. First, the Slit2 protein is expressed just below the ganglion cell layer, which restricts axons from growing down into the other layers of the retina. More repellent cues are expressed toward the retinal periphery, while attractant cues are expressed towards the optic disc. At the optic disk, Netrin-1 is highly expressed, channelling the axons out through the back of the retina and down the optic tract. Slit2 is also found on either side of the optic chiasm, possibly ensuring that no crossing occurs other than at the chiasmal midline. Once axons reach the optic chiasm, they are divided into the contralateral and ipsilateral by further guidance cues. Contralateral ganglion cells (shown in red) respond to attractants such as VEGF, Sema6D, Plexin-A1 and NrCAM. Ipsilateral ganglion cells (shown in blue) are repelled at the midline due to the interaction of Ephrin-B2 and its ligand, found only in ipsilateral ganglion cells. The ipsilateral axons instead travel straight down towards the lateral geniculate nucleus.
Once RGCs arrive at the optic chiasm, axons usually cross at the midline and grow towards the contralateral side of the brain towards their target. However, in species with binocular vision, including humans, a subset of axons do not cross the chiasmal midline and instead grow towards the ipsilateral side of the brain. The nasal side of the retina gives rise to contralateral RGCs, while the temporal retina yields ipsilateral RGCs. Although not unquestioned (Larsson, 2011), the amount of crossing appears to reflect the importance of binocular vision: in most fish and birds, where binocular vision is often not present, all of the RGC projections cross at the optic chiasm and the excellent binocular vision of birds of prey is achieved by converging retinal information in the brain by way of the thalamofugal pathway (Güntürkün and Hahmann, 1999). However, recent data does suggest that in some species of fish, ipsilateral projections do exist that do not correlate with the traditional positioning of the eyes in binocular vision (Vigouroux et al., 2021). In mice, only around 3-5% of RGC axons do not cross the midline, whereas in humans it rises to roughly 50% (Erskine and Herrera, 2014;Kupfer et al., 1967). EPHA6 is highly expressed in early developmental stages of the temporal retina, and may give rise to an excess number of ipsilateral RGCs (Lambot et al., 2005;Provis et al., 2013).
Several previously identified guidance molecules are expressed in the optic chiasm to attract or repel RGC growth cones throughout the midline. These are discussed one by one below for the contralateral and ipsilateral RGCs. Contralateral RGCs respond to attracting molecules, such as VEGF and Semaphorin 6D (Sema6D), which are both expressed at the chiasmal midline (Erskine et al., 2011;Kuwajima et al., 2012). VEGF acts through the receptor neuropilin1 (Nrp1) (Erskine et al., 2017). This receptor is expressed by contralateral RGCs only, and is crucial for midpoint crossing: Nrp1 − /− mice present with smaller proportion of contralateral RGCs (Erskine et al., 2011(Erskine et al., , 2017. Sema6D specifically attracts contralateral axons, via interaction with Plexin-A1 and NrCAM (Kuwajima et al., 2012). Indeed, a loss of SEMA6D results in a decreased number of contralaterally projecting RGC axons (Kuwajima et al., 2012).
Ipsilateral RGCs respond to repellent cues, such as resulting from the interaction between tyrosine kinase receptor EPHB1 and its ligand, Ephrin-B2. EPHB1 is uniquely expressed in ipsilateral RGC axons, and once bound to its ligand, creates a repellent response that results in ipsilateral axons not crossing and growing away from the chiasmal midline (Petros et al., 2010). The expression of EPHB1 is controlled by ZIC2, a known marker and transcription factor in ipsilateral RGCs (García-Frigola et al., 2008). Taken together, RGC axons appear to cross the chiasm depending on their retinal area of origin, which receptors they express and the resulting attractant or repellent processes at different stages. Once the RGCs exit the optic chiasm, they continue down the optic tract towards their specific targets in the brain. Although guidance molecules have been implicated in this process, brain targets of the optic tracts fall outside the scope of this review.

A summary of misrouting in albinism
Numerous albinism studies have been carried out in mice, for obvious reasons, but it is also clear that chiasm formation is fundamentally different in mice and men (Neveu and Jeffery, 2007). Consequently, translation of animal model findings to the human situation remains a challenge.
In albinism, the crossing of RGC axons at the optic chiasm is severely altered (Ather et al., 2019;Puzniak et al., 2019). The optic chiasm in albino animal models usually exhibits an increased number of contralateral RGCs crossing the chiasmal midline at the expense of ipsilaterally projecting RGCs. Previous studies demonstrated that the albinism-related defects in axon guidance are primarily due to the location and/or properties of retinal cells, not to midline cues: retinal explants from pigmented healthy mice grew normally in the presence of either pigmented or albino midline cells. In contrast, retinal explants from albino mice grew differently compared to pigmented controls (Marcus et al., 1996). This was corroborated by the observation that albino mouse models have a decreased number of RGCs expressing EPHB1, an ipsilateral-specific receptor responsible for repelling axons in the chiasm (Rebsam et al., 2012). However, the effects observed in albino models are not identical to EPHB1 null models (Rebsam et al., 2009). Thus, this observation suggests that a lack of EPHB1 expression is not the sole reason for a reduction in ipsilateral projecting RGCs. Interestingly, Iwai-Takekoshi et al. (2018) found that activation of Wnt signalling reduces ipsilateral RGCs in pigmented wild-type retina. Thus, evidence exists that the origin or location of retinal cells, rather than the guidance cues, determines the (non) crossing fate of RGC axons at the optic chiasm. During normal retinal development, ipsilateral RGCs develop earlier and during a shorter time interval compared to contralateral ganglion cells (Dräger, 1985).
In albino mice retinas, there is a temporal shift in neurogenesis resulting in uneven development of retinal cell types (ganglion cells and cone photoreceptors) compared to wild-type (Dräger, 1985). Indeed, this initially altered molecular specification might correlate with a decrease of ipsilateral axon specific transcription factor ZIC2 positive RGCs in albinism (Herrera et al., 2003). In addition, the aforementioned temporal shift in retinal development could also explain a decrease in photoreceptors in the albino retina, as the timing determines ultimate retinal cell fate. A delayed peak in albino outer nuclear layer cell proliferation compared to pigmented retinal cells may account for altered cone numbers (Ilia and Jeffery, 2000). This animal model based delay-in-neurogenesis-hypothesis is further supported by the fact that the first RGCs that reach the optic chiasm early during development are more likely to stay ipsilateral (Baker and Reese, 1993). Consequently, an albinism-related temporal shift in retinal development would result in RGCs arriving later at the optic chiasm, meaning fewer would stay ipsilateral. Whether this is due to a different fate determined at birth of the RGC, processes at the chiasm, or both, is presently unclear.

A summary of molecular/cellular mechanisms underlying development of the fovea
The fovea is the cone-rich area of the retina, located in the centre of the macula, developing from foetal week 12 in humans. The highest densities of cone photoreceptors and RGCs are found in the fovea. Every foveal cone connects to roughly 1-3 ganglion cells, whereas in the peripheral retina the ratio increases to 4-6 cones per ganglion cell (Ahmad et al., 2003;Sjöstrand et al., 1999). Together, this enables highly detailed central vision (Curcio and Allen, 1990;Wässle et al., 1989). The fovea can be separated into three different regions. The fovea centralis, also known as the foveal pit, is the centre of the fovea where the density of cone photoreceptors is highest. The fovea centralis is also the location of the rod-free zone and the foveal avascular zone, an area including and surrounding the foveal pit that is devoid of blood vessels and is critical for the formation of the foveal pit. The foveal avascular zone is defined before the foveal depression is formed, and indeed, previous research suggested that a fovea develops only in combination with an avascular zone (Provis et al., 2013). Surrounding the foveal pit is the parafovea, which extends roughly 1-3 mm from the central fovea and is home to the foveal walls. The walls are the thickest part of the retina, as it includes all of the cells that have been displaced from the foveal pit (Bringmann et al., 2018). The fovea is surrounded by the perifovea, located roughly 3-5 mm from the centre (Turan-Vural et al., 2014).
The fovea develops initially in utero by displacing inner retinal layers to form the foveal pit around foetal week 25 (Hendrickson et al., 2012). Foveal maturation extends long after birth and continues until 13 years of age. At that time, cone numbers have increased in the area and become tightly packed in a hexagonal configuration . Indeed, foveal cone density of young adult retinas can be 20-30 times higher compared to foetal cone density (Yuodelis and Hendrickson, 1986). The complete development and maturation of the fovea is poorly understood, but it is thought that the formation of the foveal pit and the central packing of cone cells are independent of each other.

A summary of foveal hypoplasia in albinism
In albinism, developmental retinal abnormalities occur, including foveal hypoplasia (lack of foveal pit). This can be clearly seen using OCT (section 1.1). Possible explanations for hypoplasia as a result of the dysfunctional activity of the central pigmentation pathway, in particular the activity of L-DOPA/GPR143 and possible subsequent PEDF, VEGF and exosome signalling, are discussed next.
In albinism, the fovea is characterised by the absence of a foveal avascular zone. Interestingly, PEDF (part of the retinal pigmentation pathway) is known to be a potent inhibitor of angiogenesis (Kozulin et al., 2009b;Xi, 2020). In the developing retina, PEDF is not only expressed by the RPE, but also by RGCs (Karakousis et al., 2001). It is especially expressed by RGCs located at the emerging fovea when compared to other portions of the retina (Kozulin et al., 2010), illustrating the importance of PEDF signalling in the macular area. Therefore, we hypothesise that, if PEDF expression is decreased in albino retinas compared to healthy controls, this lack of PEDF could contribute to underdevelopment of the foveal avascular zone, and, ultimately, foveal hypoplasia in these patients (Kruijt et al., 2022).
In contrast with PEDF, VEGF promotes angiogenesis, and a balance of these growth factors is essential for maintaining the avascular areas of the retina (Ohno-Matsui et al., 2001). Indeed, similarly to PEDF, VEGF is also expressed by the RGCs in the central retina before the formation of the fovea (Sandercoe et al., 2003). This suggests an early expression of these two factors in the immediate area of the yet-to-develop fovea plays a crucial role in the development of the important foveal avascular zone. Currently, there is no indication for a potential role for exosomes or GPM6A in foveal development. Finally, it should be noted that, in conditions other than albinism, such as aniridia caused by mutations in the PAX6 gene, foveal hypoplasia can occur in the absence of optic nerve misrouting (Neveu et al., 2005).

Potential common mechanisms
Persons with albinism have a decreased foveal cone density (Wilson et al., 1988) and a lack of a rod-free zone (Fulton et al., 1978;Woertz et al., 2020b). Albino animals show lower rod numbers in general (Donatien and Jeffery, 2002;Grant et al., 2001) as well as a decreased number of ipsilateral projecting RGCs. This can (at least in part) be attributed to delayed neurogenesis in the neural retina (Baker and Reese, 1993;Dräger, 1985;Ilia and Jeffery, 2000). This could take place during the first wave of neural retina cell development since that includes, in the healthy retina, simultaneous genesis of RGCs and photoreceptors. This delay could be caused by a (genetically determined) lack of L-DOPA, since that molecule not only activates GPR143, but is also involved in the cell-cycle exit (Ilia and Jeffery, 2000;Tibber et al., 2006). Thus, theoretically, a lack of L-DOPA signalling could result in a delayed genesis or differentiation of specific neural retinal cells affected in albinism. However, in the case of FHONDA, pigmentation is normal and we hypothesise that L-DOPA and GPR143 signalling are unaffected. Although this hypothesis has not been studied specifically, FHONDA patients do present with foveal hypoplasia and optic nerve misrouting. In addition, SLC38A8 appears not to be present in the RPE, but in the neural retina (Poulter et al., 2013).
Thus, the most likely explanation is that a common pathological mechanism for foveal hypoplasia and misrouting in albinism exists, downstream of RPE/GPR143 mediated signalling and FHONDA action in the neural retina, and which probably originates in the developing early neural retina. At the same time, it is not clear whether candidate signalling cues identified, such as L-DOPA, EphA6, PEDF/VEGF balance, and GPM6A (also) affect the expression or function of SLC38A8 (FHONDA) or even more downstream mechanisms and restricted phenotypes directly.
The potential action of L-DOPA/GPR143, GPM6A is discussed above (section 8.2). Three (additional) possible molecular mechanisms governing both foveal development and chiasmal misrouting are discussed next. Two proposed mechanisms originate in the neural retina probably downstream of SLC38A8, one of them is RPE based.
First, Poulter et al. postulated a general hypothesis how SLC38A8 mutations may lead to the FHONDA phenotype (Poulter et al., 2013): in summary, glutamine transporters, such as SLC38A8, play a role in the recycling of (glutamate and GABA) neurotransmitter metabolites in the CNS, including possibly also in the retina. Indeed, synaptic activity does play a role in the developing retina and refinement of retinal projections including those of RGCs (Koch et al., 2011). Additionally, neurotransmitters may be involved in retinal progenitor proliferation regulation before synapses are even present (Martins and Pearson, 2008). Taken together, these phenomena may affect the temporal timeline of retinal cell generation or differentiation, which is thought to be crucial to the proper proportion of crossing axons at the optic chiasm (Martins and Pearson, 2008) and thus may contribute to the albinism-related FHONDA phenotype.
The second hypothesis regarding the SLC38A8-related mechanism causing both foveal hypoplasia and chiasmal misrouting was recently postulated by Kruijt and co-workers (Kruijt et al., 2022): SLC38A8 may be involved in PEDF and EphA6 signalling in the developing retina. Interestingly, PEDF and EphA6 are involved in formation of the foveal avascular zone (Kozulin et al., 2009a), which is essential for formation of the foveal pit (Provis et al., 2013). High levels of EphA6 are present in the macula during development and foveal cone specialisation (Kozulin et al., 2009a(Kozulin et al., , 2009b(Kozulin et al., , 2010. In addition, the temporal retina, where ipsilateral retinal projections originate, exhibits high expression of EphA6 (Lambot et al., 2005;Provis et al., 2013). Aberrant regulation of EphA6, possibly in the case of albinism and FHONDA, could therefore potentially explain the combination of foveal hypoplasia and chiasmal misrouting.
Finally, altered PEDF/VEGF balance in the developing (albino) retina could also be involved in both foveal development and regulation of RGC projections. Altered PEDF and VEGF signalling can account for the absence of a foveal avascular zone (Ohno-Matsui et al., 2001;Provis et al., 2013). The foveal avascular zone is hypothesised to be essential for the formation of the foveal pit (Provis et al., 2013). At the same time, PEDF is involved in RGC neurite development and the neuroprotection of RGCs after damage to the optic nerve (Vigneswara et al., 2013). In addition, VEGF acts as an attractant for contralateral RGCs in the optic chiasm through the Nrp1 receptor (Erskine et al., 2011(Erskine et al., , 2017. Possibly, VEGF secreted by the RPE can have a similar effect in the neural retina. However, this needs to be confirmed. Obviously, the precise role of each of the (combination of) many factors controlling these processes needs to be further elucidated.
11. The pigmentation pathway in the skin: why is the OA1 phenotype restricted to the retina?
As described above (section 5), the OA1 albinism phenotype is caused by mutations in GPR143, and its pathology is largely restricted to the visual system (Lewis, 1993a). The GPR143 receptor is not only implicated in OA1, but also one of the key features of the proposed retinal pigmentation pathway potentially resulting in both abnormal growth factor secretion as well as vesicle formation and release (Fig. 3) (Falk et al., 2012;Locke et al., 2014;Lopez et al., 2008;McKay, 2019). OA1 patient RPE is hypopigmented because (macro-)melanosome numbers are low and not evenly dispersed compared to healthy RPE (Cortese et al., 2005;Garner and Jay, 1980;Wong et al., 1983). Intriguingly, however, Gpr143 is also expressed in skin melanocytes in mice (Bassi et al., 1996). In OA1 patients, skin macro-melanosomes are also present, but the colour of the skin is (nearly) normal (Garner and Jay, 1980). These observations raise the question: what are the pigmentation signalling differences between retina and skin, and what role does GPR143 play in skin development, pigmentation or homeostasis? Two clear differences between retina and skin are that mature skin melanosomes, as opposed to those in the RPE, traffic out of the melanocytes into skin keratinocytes, upon which local intracellular melanosome formation largely ends. Skin melanocytes are also constantly producing melanosomes, whereas RPE melanocytes only produce these during development (Boulton, 2014;Ohbayashi and Fukuda, 2020). Again, we discuss three potentially relevant features of the proposed pigmentation pathway (Fig. 3): melanosome formation, PEDF/VEGF secretion and exosome release. Macro-melanosome formation in the RPE occurs through the GPR143 signalling feedback loop (section 9.2) (Cortese et al., 2005;Schiaffino and Tacchetti, 2005). Possibly, GPR143 cannot (fully) exert its action in skin, most likely due to subtle molecular differences between RPE and skin melanocytes: Further investigation is warranted whether macro-melanosomes can only form when multiple melanosomes are co-maturing as is the case in RPE, but not in skin as the melanosomes are trafficked out of the cell.
Paracrine signalling (PEDF, VEGF and exosome release) may also be different between (OA1) RPE and skin. The role of PEDF, VEGF and exosome release in the retina is discussed above (section 8.1-3). In the skin, it is not clear whether GPR143 signalling modulates skin physiology and disease through control of PEDF and VEGF secretion at different stages in life. PEDF is produced by skin melanocytes and contributes to skin homeostasis (Zhang et al., 2009). This growth factor does have a role in physiology of hair follicles where it affects cell motility and possibly inhibits angiogenesis . PEDF has also been implicated in melanoma tumour growth (Chi et al., 2006). Next to PEDF, VEGF also plays a role in skin homeostasis. In this context, VEGF has been implicated in angiogenesis, melanoma formation and wound healing (DiPietro, 2016). Interestingly, VEGF receptor expression increases after UVB exposure and promotes melanogenesis and melanocyte proliferation (Zhu et al., 2020). Like PEDF, VEGF has a role in melanoma development: higher VEGF expression in melanocytes is indicative of neoplastic changes and the development of melanoma (Einspahr et al., 2007;Gajanin et al., 2011;Stefanou et al., 2004), which may underlie the aforementioned link between albinism and skin cancer. Unfortunately, VEGF release from healthy skin melanocytes has been understudied.
Finally, potential exosome release from the RPE (in albinism) via GPR143 signalling is discussed extensively above (section 8.3). In the skin, it is presently not clear whether GPR143 signalling in melanocytes regulates exosome release. Healthy skin melanocytes do release exosomes which play a role in skin physiology: Upon UVB stimulation, exosome secretion from skin melanocytes increases (Shen et al., 2020). Exosomes released from melanocytes probably do have an influence on skin homeostasis. Finally, a potential role for GPM6A in the development or homeostasis of the skin is presently unclear.
Taken all available data together, we hypothesise that a molecular pigmentation pathway in the skin is present, that resembles the retinal pigmentation pathway postulated above. There are a number of molecular and cellular differences between the two aforementioned pathways that warrant further in-depth investigation. One of these differences seems to be that in skin, melanosomes are trafficked out of the cell, whereas in the RPE they remain inside the cell. Another difference seems to be that, in skin, GPR143 feedback signalling to form macro-melanosomes may not be fully active, while GPR143 downstream signalling, potentially related to PEDF, VEGF secretion and exosome secretion is. In-depth further research on both albinism retina and skin (models) regarding these issues will yield further detailed knowledge about the aetiology of albinism and the development of potential future therapies.

A summary of concepts and findings
In this report, we proposed one central pathway for retinal pigmentation implicated in the pathogenesis of human albinism. In the context of this pathway, we connected disease gene, gene function, and cellular pathology to phenotypes. This suggests a correlation between increasingly specific cellular and molecular pathology and increasingly restricted phenotypes (Fig. 3): We formulated that defects affecting the genesis or function of multiple cellular organelles, including melanosomes, cause syndromic forms of albinism (HPS, CHS). Specific defects in melanosome function and pigment cell differentiation cause specific forms of OCA (OCA1-8). Further, we incorporated that GPR143 mutations uniquely causes OA1, resulting in macro-melanosome formation and a restricted ocular pigmentation phenotype. Defective GPR143 signalling may affect paracrine signalling in the RPE resulting in an abnormal (developmental) neural retinal phenotype. Finally, we highlighted the involvement of SLC38A8, in mutated form leading to a very restricted neural retinal "albinism" phenotype (FHONDA), without pigment abnormalities. The retinal pigmentation pathway, based on albinism gene involvement, beginning upstream and continuing downstream, can be defined as: HPS1-9, CHS; OCA1-8; GPR143/OA1; SLC38A8/FHONDA (Fig. 3).
In terms of biochemical action and physiological signalling, the proposed retinal pigmentation pathway can be described as follows: The amino acid L-tyrosine is converted to melanin in RPE melanosomes under release of L-DOPA. Subsequently, L-DOPA from the RPE and L-DOPA/dopamine from the neural retina activates or inhibits the subsequent signalling from the apical GPR143 RPE receptor. Next, GPR143 signalling may have multiple effects: via a regulatory feedback loop it has an effect on maturation of the (macro-)melanosome, and downstream it possibly modulates PEDF, VEGF and exosome secretion from the RPE. Less local tyrosine-melanin conversion leads to less local L-DOPA production and subsequently, less GPR143 activation. A lack of functional GPR143, for example, due to mutational events, may have a similar effect in the proposed retinal pigmentation pathway; and both can potentially result in altered paracrine signalling to the neural retina. The resulting alterations in neurogenesis in the neural retina may contribute to the ocular pathology: foveal hypoplasia and chiasmal misrouting. We have described the FHONDA syndrome and how the corresponding protein may act downstream of GPR143 signalling in the albinism pathway to produce such a remarkably albinism-like phenotype.
Next, we suggested three features of this pathway that go beyond direct albinism disease gene involvement: first, we noted that the regulation of the activity of the (retinal) pathway proposed is under circadian control. Second, we bioinformatically compared (GPR143mediated) exosome content release from the apical RPE and proteins known to be involved in neural development. That analysis yielded one particular new signalling protein of interest: GPM6A. We suggest that this protein may contribute, at least in part, to neural retinal development and, perhaps, albinism pathology. We also explore how the retinal pigmentation pathway could partly explain foveal hypoplasia and chiasmal misrouting. RGC genesis and the timing of RGC axon development may be key to proper crossing of the RGC axons at the optic chiasm and the formation of the foveal avascular zone for foveal development. Both could be modulated via the retinal pigmentation pathway. Finally, we compared retinal pigmentation in skin and eye, as far as data is available.

Future perspectives I: research
In this section, we will discuss possible directions of new research into the pathophysiological mechanisms of albinism. The conceptualization of one retinal pigmentation pathway together with new technological possibilities, opens up many new avenues for systematic future research. These topics include at least: further clinical (sub-)classification in relation to genotype, the role of specific (candidate) regulatory or signalling molecules (L-DOPA, PEDF, VEGF, GMP6A) along the proposed pathway and exosome signalling in the developing (albinism) retina, the similarities and differences between molecular and cellular mechanisms underlying foveal hypoplasia and chiasmal misrouting and, even, the molecular mechanisms of nystagmus in albinism. New technological possibilities include advanced in vitro modelling of albinism using stem cell-derived (knock-out) RPE cultures and retinal organoids, and the development and characterization of new animal models. These issues are described in summary below.
There is a clear need in albinism and pigmentation research for further detailed clinical and molecular genetic investigations. It is not only essential to determine further phenotype-genotype relationships and to develop early DNA diagnostic tests, for example in new-born screenings. Most importantly, emerging new therapies can only enter clinical trials when the natural history of disease subtypes is established together with timely treatment opportunities.
As described in sections 3-6, there are, so far, 22 disease genes implicated in (the retinal pathology) of albinism (including FHONDA/ SLC38A8). Between the different genetic causes there is considerable variability in the phenotypic severity, except for very limited variability in the more consistently severe FHONDA phenotype. Further characterization of albinism (sub-) phenotypes and their correlation to genotypes enhances understanding of patient-specific pathologies and treatments. In addition, it is important to mention that additional genetic causes of albinism still have to be identified, with approximately 25% of patients remaining genetically unresolved (Kruijt et al., 2018). Some candidate genes have been proposed in the literature. A downstream effector of GPR143, GNAI3 (discussed briefly in section 5.2), has been raised as a potential additional candidate gene for OA (Young et al., 2016). Interestingly, animal models with mutations in BLOC-1 subunits (BLOC1S4, -5 and SNAPIN) that have not been implicated in HPS in humans, as well as a protein that interacts with BLOC-3 (RAB38), show HPS-like phenotypes (Bowman et al., 2019;Loftus et al., 2002;Montoliu and Marks, 2017;Oiso et al., 2004;Zhang et al., 2014). Finally, there are three more genes (VPS33A, RABGGTA, and SLC7A11) involved in LRO trafficking of associated with pigmentation phenotypes in mice (Chintala et al., 2005;Detter et al., 2000;Suzuki et al., 2003). Consequently, the aforementioned genes are excellent candidates to elucidate the yet unresolved molecular pathology in albinism patients.
The currently available models of human retinal development and albinism have a number of limitations. As discussed in section 10 above, these include the absence of a true fovea in most animal models used, and significant differences in contralateral and ipsilateral (non-) crossing RGC axons at the optic chiasm. With the development of patient-derived induced pluripotent stem cells and methods to derive RPE cells and 3D retinal organoids, possibilities have opened up to study cellular and molecular processes in human-representative in vitro model systems. This has already been applied to a number of (genetic) ocular diseases, including hereditary glaucoma, retinitis pigmentosa, and Xlinked retinoschisis (Gao et al., 2020;Huang et al., 2019;VanderWall et al., 2020). In some of these disorders, these organoid models recapitulate the clinical phenotype remarkably well. RPE cultures derived from patient stem cells can also be useful to characterise the details of the molecular and cellular pathology of albinism, and test new treatments. Several (undifferentiated) albinism patient-derived iPSC lines (OCA1, OCA2, OA1, CHS, HPS1/2) have currently been developed (Baulier et al., 2018;Maguire et al., 2016aMaguire et al., , 2016bSerra-Vinardell et al., 2020). OCA1 and/or OCA2 patient-derived iPSC RPE lines have been produced by Schaub, George and Bakker and co-workers (George et al., 2022;Schaub et al., 2020) (this report; Fig. 13), with typical monolayers of cobblestone shaped cells, absence of pigment and mature RPE barrier function. These models await further detailed characterization, but are essential for systematic (therapeutic) investigations of human albinism. No reports have been published yet on full retinal organoids derived from albinism patient iPSC lines. Of note, even the most advanced in vitro 3D retinal organoid models (Cowan et al., 2020;Kim et al., 2019;Wagstaff et al., 2021b), have not yet been able to recapitulate RGC projections, foveal development and nystagmus. Indeed, this will be a challenge for future albinism research.
The conceptualization of one canonical retinal pigmentation pathway provides the opportunity for further systematic investigation. Subject for further study is the subcellular location(s) of the GPR143 receptor in the in vivo human RPE. Further, the regulation and signalling of the proposed retinal pigmentation pathway is, most likely, complex, and partly needs to be elucidated. More specifically, further research in this area includes the exact role of circadian rhythms, effects of L-DOPA signalling, growth factor action, SLC38A8 function and exosome release, and the potential role(s) of EphA6 and GPM6A. Further downstream, the intrinsic neural properties and neural guidance cues (specified in section 10) that determine proper retinal and optic nerve development are also an exciting area of ongoing research. A subject of future research may be the release of exosomes from the RPE, not only in human albinism, but also in other retinal disease entities. Indeed, the role of exosomes in various physiological contexts recently gained attention: exosomes are not just a way of cellular waste secretion but also an important signalling vehicle between cells (Colombo et al., 2014;van der Pol et al., 2012). Interestingly, L-DOPA treatment of cultured primary porcine RPE cells (likely, but not per se exclusively, via GPR143 activation) halt their apical RPE exosome release (Locke et al., 2014). If we translate this finding to current concepts in human albinism, where L-DOPA is not sufficiently produced or GPR143 itself is not functional, we hypothesise that subsequent exosome release may increase. Another specific question to be answered is: what do these albinism driven exosomes carry compared to healthy controls and does the (abnormal) biomolecular content play a role in (defective) development of the neural retina? Our analysis of the available proteome datasets of apically released exosomes from RPE cultures and genes related to neural retina development yielded one interesting protein: GPM6A, that is an excellent candidate to modify neural retina phenotype of albinism (as described in section 8.2). Furthermore, in addition to proteins, exosomes also carry signalling lipids, RNAs and small molecules (Colombo et al., 2014;van Fig. 13. Rows: WT hESC (A1-2) and OCA1 iPSC-derived RPE (B1-2) cultures. Column, left (A1-B1) corresponds with culture time point day 25, which coincides with the start of pigmentation in the WT cultures. Column, right (A2-B2), day 60: full pigmentation can be seen in the WT cells (A2), but not in the OCA1 line (B2). Both lines exhibit the typical hexagonal RPE morphology (Bakker et al., unpublished data). Data presented here corroborate those previously described (George et al., 2022;Schaub et al., 2020).
der Pol et al., 2012). All these molecules can have their own effects in health and disease. Taken together, functional analysis of exosome release is a promising new frontier in albinism research.

Previous and ongoing clinical trials
Before discussing potential new experimental therapies, it is of interest to make a short inventory of recent and ongoing clinical trials in the context of the retinal pigmentation pathway: https://www.clinical trials.gov/ct2/results?cond=Albinism. In summary, the resulting entries are oral L-DOPA treatment, dietary supplementation of zeaxanthin and lutein, and, finally, nitisinone treatment. These are mostly based on repurposing drugs; i.e. existing drug FDA/EMA approved for another indication, and subsequently approved in clinical trials for albinism. These trials are summarised below.
A phase 2 clinical trial, administering L-DOPA orally to albinism patients has been performed (ClinicalTrials.gov Identifier: NCT01176435) (Summers et al., 2014). L-DOPA supplementation is a potentially promising potential therapy since, in most forms of albinism, with ocular albinism and FHONDA as the exceptions, local L-DOPA production is probably limited, resulting in pathological effects in the neural retina via the proposed retinal pigmentation pathway. In this clinical trial, the age of the participants ranged from 3.5 to 57 years old (mean age: 14.5). Participants included OCA1, OCA2 and HPS1 patients. The result of the trial showed that there was no significant effect on VA after 20 weeks of oral L-DOPA treatment (Summers et al., 2014). One possible explanation for the lack of significant results was the recruitment of albinism patients with different gene mutations, whose response to L-DOPA are not necessarily comparable. Another explanation could be that people with more mature visual systems were included, and once the development of RGCs and the fovea is complete, it is unlikely that a prominent effect would be seen following L-DOPA treatment. A stratification of future clinical trials according to albinism genotypes could provide more precise results.
Given the proposed retinal pigmentation pathway (Fig. 3), the results of this L-DOPA treatment trial may, in retrospect, be explained as follows. As albinism mechanistically seems to be a very early onset disorder, treatment at later ages may have no effect. In this trial, L-DOPA supplementation, which in principle can activate the GPR143 receptor to rescue the albinism phenotype, was given orally to albinism patients over the age of 3.5. At that age the initial formation of the foveal pit and retinal projections to the chiasm are already largely completed. The foveal pit is forming already at 25 weeks of gestation, and foveal cone specialisation and packing starts from birth and continues at least until four years of age. The fovea is still developing up to 13 years of age (Hendrickson et al., 2012). A potential large treatment effect in individuals with a considerably matured macula is unlikely. Nonetheless, because cone packing and specialisation correlates with VA, aiding this process with L-DOPA supplementation could still be beneficial (Provis et al., 2013). Finally, it should be noted that L-DOPA release from the neural retina, and possibly the RPE, is, at least postnatally, under circadian control (section 9.1) (Bagchi et al., 2020). Thus, a potential L-DOPA supplementation treatment should not only start early in life, but also take circadian timing into account. Prenatal or perinatal treatment could perhaps still be an option when clinical genetic diagnosis is already clear in the family, taking ethical and safety considerations into account.
Another relevant (pre-)clinical albinism study using oral L-DOPA supplementation in the form of Levodopa is currently carried out by Helena Lee, Andrew Lotery and co-workers in Southampton, UK (http s://gtr.ukri.org/projects?ref=MR%2FR007640%2F1; the OLIVIA study). Levodopa is a drug that is currently successfully used to treat (three-month-old) infants with infantile dystonia, which is now being repurposed for human albinism. Indeed, in contrast with similar previous studies and trials, the OLIVIA study includes infants from three months onward. The purpose of the OLIVIA study is threefold: (1) prove that Levodopa supplementation at early age (up to postnatal day 15) can rescue retinal development, morphology and visual function in a mouse model of human albinism. This was reported to be successful ; (2) Analyse the response of the developing retina after use of different doses of L-DOPA at multiple stages of development; (3) Test the feasibility of Levodopa intervention and determine inclusion criteria for a future randomised controlled clinical trial in children with albinism. As the clinical phase of the study is planned to end in 2024, no additional public data is yet available.
A completely different approach was to test oral nitisinone supplementation in OCA1 patients with residual pigmentation, in a combined phase 1/2 clinical trial (ClinicalTrials.gov Identifier: NCT01838655) (Adams et al., 2019). Nitisinone (an FDA-approved drug for hereditary tyrosinemia type 1) raises plasma levels of tyrosine through inhibition of tyrosine catabolizing enzymes. This approach may be particularly effective in forms of albinism where a residual function of TYR remains. Obviously, some residual TYR function is essential for L-DOPA production in the RPE (Fig. 2). This hypothesis was tested by preclinical supplementation of nitisinone in an OCA1 mouse model (with residual Tyr activity) which increased pigmentation in coat, iris and RPE (Onojafe et al., 2011). Similar experimental treatment of an OCA3 mouse model with nitisinone yielded a significant positive pigmentation effect, but in the iris only (Onojafe et al., 2018). Exposing other genetically modified OCA1 mouse models to nitisinone (Onojafe et al., 2011(Onojafe et al., , 2018 also did not result in significant re-pigmentation (Lluis Montoliu, unpublished results). It is not clear whether treatment of the aforementioned mice had an effect on RPE L-DOPA production and subsequent neural retina development or optic nerve decussation.
The participants in the human clinical nitisinone supplementation trial indeed had some residual TYR activity, and the increase in the substrate tyrosine could thus possibly raise L-DOPA and melanin production closer to normal levels. No adverse safety issues of the trial were reported. At the same time, no statistically significant difference in skin and eye pigmentation or VA was found. In some subjects there appeared to be a slight increase of skin pigmentation. Consequently, these results were in line with the aforementioned pre-clinical studies. In view of the retinal pigmentation pathway proposed, in some severe cases of albinism nitisinone treatment would be most effective if supplemented at a relatively young age when the affected developmental processes are still happening in the retina. A challenge in this context is that the VA would have to be assessed very early, possibly through grading of foveal hypoplasia, as this correlates with VA. In this stage, the fovea is still in development and higher tyrosine levels inducing higher L-DOPA production could possibly aid in foveal maturation and improve VA. Again, while this treatment could in principle have some effect in a subset of patients, treatment at early age raises ethical and safety questions.
Yet another albinism trial has been listed by investigators of the Johns Hopkins University (ClinicalTrials.gov Identifier: NCT02200263). They performed a dietary supplementation trial with the macular carotenoid pigments zeaxanthin and lutein in a group of genetically undefined albinism patients aged 12 and older. This can be considered a low-risk trial, given that these carotenoids generally are taken up from a general dietary source of eggs, fruits and vegetables. The function of these non-melanosomal macular pigments is to absorb excess light and to increase the protection of photoreceptors from oxidative damage. The overall reason for this clinical trial was to investigate a potentially (improved) correlation between fundus pigmentation and visual function. The potential effect was expected to be optical, reducing light scatter and chromatic aberration. The goal was not to rescue the physical abnormalities of the retina, such as foveal hypoplasia or chiasmal misrouting. So far, to the best of our knowledge, no results of this trial have been published in any form.
To summarise all published therapeutic trials, the L-DOPA and nitisinone studies both attempt to restore the L-DOPA levels in the retina, albeit in different ways. Most studies included adult subjects, i.e. with little ongoing foveal maturation. A significant effect of the treatment in humans was not observed. No major safety issues of the treatments were identified in the published trials. A next step could be perinatal administration, which, obviously, raises considerable ethical and safety issues that need to be addressed first. Zeaxanthin and lutein administration to albinism patients has not been reported to elicit an effect.

Development of potential new therapies
On the basis of the proposed retinal pigmentation pathway, there may be a few new therapeutic options to consider. Treatment of (albinism) eye defects has many advantages over many other organ systems. These include potential local application and non-invasive monitoring of the treatment, the fact that the eye is a closed system, and that one eye can sometimes be used as an untreated control. Obviously, early treatment in albinism is essential but experimentation and invasive treatments on young children are subject to ethical and safety concerns. If possible, local treatment is preferred over systemic treatment to limit potentially harmful side effects. Therapeutic options for albinism may potentially also be applicable to other diseases involving melanosome pigment defects or early onset retinal disease.
New potential albinism treatment options include gene therapy, and small molecule or growth factor administration (Ruan et al., 2017). Gene therapy application in the eye is promising. Adeno-associated virus (AAV) gene therapy treatment for patients with RPE65-related retinal dystrophies was recently approved by FDA and EMA (Russell et al., 2017). Many new preclinical ocular studies and trials, including modern CRISPR/Cas9 approaches, are well under way. As genetic defects in patients can now be analysed better, faster and cheaper than in the past, gene therapy replacing the defective disease gene could potentially treat albinism (Dunbar et al., 2018). As with previous clinical trials for albinism (section 12.3.1), a challenge would be to administer the therapy in an early stage of pathology since intact retinal cells are needed to accommodate gene therapy. So far, gene therapy is not considered suitable for infants and young children, and therefore may also be too late for foveal development to be rescued AAV gene replacement therapy.
Small molecule intervention affecting the regulation and activity of the retinal pigmentation pathway could also be considered. As described above, L-DOPA and nitisinone treatment are examples of small molecules that can potentially have an effect on (the development of) albinism pathology. Several types of albinism appear to be associated with impaired proteins regulating the internal pH of melanosomes affecting TYR activity. Therefore, one could also imagine the discovery and application of small molecules that could also counteract the pH distortions and, thus, restore pigmentation levels. Further (pre-) clinical research into other small molecules regulating the activity of (parts of the) retinal pigmentation pathway should be considered as an experimental therapeutic option.

A potential role for the retinal pigmentation pathway in other retinal diseases, including AMD
In addition to albinism, L-DOPA/GPR143 signalling (Figueroa and McKay, 2019) and the proposed retinal pigmentation pathway (section 2) could play a role in other retinal pathologies. Alterations in melanin pigmentation and L-DOPA could have an effect on downstream RPE signalling, contributing to pathological outcomes. The activity of the proposed retinal pigmentation pathway may be not exclusively genetically driven, but the mechanism itself can also be affected by environmental factors, genetic background, diet or drugs, metabolic differences, disease and ageing, resulting in melanosomal pigmentation differences. The retinal pigmentation pathway could also (partly) play a role in non-albinism pathologies in various stages of life. It is outside of the scope of this report to address all potential retinal pathologies in which melanosomal pigment changes play a role, but to show the reasoning, we briefly describe the examples of AMD and ROP below.
AMD is a disease affecting 4% of the population over 60 and almost a third of individuals over the age of 75 (Klein et al., 2011;Wong et al., 2014). AMD is characterised by central vision loss initially characterised by macular pigment alterations and the formation of subretinal deposits, called drusen. AMD consists of a common dry form (retinal atrophy) and a more rare wet form (neovascularization). AMD patients experience a severe loss of quality of life and lose their independent lifestyle (Klein et al., 2011). There is no (effective) cure for at least 90% of patients . The disease is caused by (combinations of) environmental factors, such as smoking; and genetic risk factors, most importantly genetic variation in genes that determine the activity of the complement system (de Jong et al., 2021). Mechanistically, abnormalities of the complement system, the lipid metabolism, and extracellular matrix have been implicated in the disease (Fritsche et al., 2016).
Oxidative stress is probably a major driver of the AMD (Abokyi et al., 2020), affects RPE health, and basolaterally secreted oxidatively modified biomolecules may invoke a complement response, leading to further local pathology. Much attention has been given to the photoprotective role of specific macular pigments in AMD (Chew et al., 2013(Chew et al., , 2014. Also, a decline in the optical density of non-melanosomal macular pigment is part of normal ageing, and significantly less macular pigments are observed in AMD compared to healthy individuals (Beatty et al., 2001). However, the potential role of melanosomal pigments in the RPE (and choroid) in AMD is relatively understudied. Here, we focus on melanosomal pigment abnormalities accompanying AMD and how the proposed retinal pigmentation pathway may be implicated in part of the disease: Does (reduced) melanosomal pigmentation have a direct (declining) protective antioxidant effect or is there an indirect effect due to altered signalling through the proposed retinal pigmentation pathway, or both? Below, we will first discuss the antioxidant activity (defect) of melanin in the RPE in AMD, and subsequently the potential indirect effects via the retinal pigmentation pathway.
The ageing RPE (and choroid) is severely burdened by oxidative stress and resulting free radicals (Mettu et al., 2012;Miceli et al., 1994). Apart from other (macular) pigments, melanin pigmentation in the RPE has also been directly implicated in protecting the AMD macula from oxidative damage (Bergen et al., 2019;de Jong et al., 2021). Furthermore, epidemiological evidence suggests that less pigmented individuals may have increased risk for oxidative stress-related diseases. Indeed, in a prospective cohort study with subjects of many different ethnicities within the United States, Caucasian participants were 5-fold more likely to suffer from AMD than participants of African descent (Klein et al., 2011. Participants of Hispanic and Asian descent showed an intermediate risk. Controlling for other known risk factors like body mass index, hypertension, diabetes, smoking and alcohol consumption did not explain the differences (Klein et al., 2011. Albinism as a risk factor for developing AMD would support this hypothesis, but whether albinism patients have an (slightly) increased risk for developing AMD is unknown. The possible late-onset clinical diagnosis of AMD in albinism patients may go unnoticed since these patients are rare and have a low VA already early in life.
We next consider the potential role of the retinal pigmentation pathway in AMD pathology. Even in the presence of protective pigments, oxidative stress harms many molecular compounds and microstructures of the RPE cell, including the melanosomes. Thus, we can hypothesise that the RPE and individual features of the retinal pigmentation pathway may be directly affected by oxidative stress. Alternatively, this stress may primarily affect TYR activity and melanin production, and, through L-DOPA, indirectly, downstream signalling and functionalities, such as GPR143 signalling and potentially PEDF and VEGF secretion and exosome release (Falk et al., 2012;Locke et al., 2014). Indeed, it may affect the resilience of the ageing retina and development of AMD (McKay, 2019).
To the best of our knowledge, a direct effect of subretinal L-DOPA/ dopamine levels on AMD pathology has not been investigated. Nonetheless, if GPR143 signalling protects from AMD damage during ageing, L-DOPA supplementation treatment may exert protective effects. Interestingly, oral L-DOPA treatment for Parkinson patients reduces, in parallel, risk for AMD or delays onset of the disease (Brilliant et al., 2016). Further, in a pilot study, patients with wet AMD (who received no anti-VEGF treatments) were treated with L-DOPA, which increased VA and delayed the necessity of anti-VEGF treatment (Figueroa et al., 2021). The possible role of these growth factors (PEDF and VEGF) as well as exosomes (including GPM6A) will be discussed next.
According to the proposed pigmentation pathway, decreased pigmentation (in areas of the RPE) may lead to less L-DOPA, less GPR143 activity and increased VEGF secretion (Falk et al., 2012;Holekamp et al., 2002). Increased local VEGF secretion enhances angiogenesis as seen in neovascular AMD. Similarly, decreased pigmentation, L-DOPA and GPR143 activity may lead to less PEDF secretion, which also has been implicated in AMD (Barnstable and Tombran-Tink, 2004;Farnoodian et al., 2017;Holekamp et al., 2002). GPR143 activation/signalling may possibly increase the amount of exosome secretion (Locke et al., 2014). Interestingly, there are a number of independent relevant leads in the literature implicating basolateral RPE exosomes in AMD: First, exosomes from stressed ARPE-19 cells may enhance the sensitivity of the receiving cells for VEGF signalling, possibly facilitating pathological events (Atienzar-Aroca et al., 2016;Fukushima et al., 2020). Second, RPE exosomes, released upon oxidative stress, contain a collection of phosphoproteins that can activate inflammation, autophagy, apoptosis, and degeneration pathways in receiving RPE (and other cells) (Biasutto et al., 2013;Gao et al., 2015;Ke et al., 2020;Shah et al., 2018;Wang et al., 2009b;Yang et al., 2020). This may facilitate spread of the AMD pathology. Third, a role of RPE exosomes in AMD may come from trafficking of (complement) proteins, or lipids that can protect both RPE and photoreceptor cells from oxidative stress and prevent apoptosis (Anderson et al., 2010;Boon et al., 2008;Klingeborn et al., 2017;Sreekumar et al., 2010;Wang et al., 2009a, b). Fourth, exosomes could contribute to the formation of drusen, a hallmark of the AMD pathology (Bergen et al., 2019;Boon et al., 2013;Flores-Bellver et al., 2021).
Finally, the potential future treatments discussed in section 12.3.2, based on manipulation of L-DOPA or other compounds part of the retinal pigmentation pathway could possibly also be of interest for treatment of other disorders, such as AMD. Indeed, L-DOPA administration for AMD patients could be beneficial: a retrospective study of patients that took L-DOPA (for other indications than AMD) showed that treated individuals have a smaller risk for developing AMD or develop symptoms later (Figueroa et al., 2021). A potential role of the exosome protein GPM6A has not been studied in AMD.
To illustrate further that the proposed retinal pigmentation pathway does not have to be driven by genetic factors (alone), we describe the example of ROP next. ROP is an important cause of infant blindness. Risk factors to develop ROP are low birth weight and/or premature birth. In ROP, the blood vessel system in the retina is not fully developed yet, frequently resulting in fragile and leaky blood vessels, retinal scarring and detachment. When children are born prematurely, ROP can occur in two major phases. In the first phase, hyperoxia causes VEGF levels to decrease, which leads to a halt in the development of the retinal vasculature (Cavallaro et al., 2014;Hellström et al., 2013). In the second phase, the rising metabolism of the developing retina and resulting hypoxia lead to a rise in VEGF levels and abnormal production of blood vessels in the vitreous (Cavallaro et al., 2014). Both phases may lead to children with ROP, presenting subtle changes to their foveae (Krumova et al., 2019;Martínez-Córdoba et al., 2021). How does the proposed retinal pigment pathway potentially play a role in this? Does the level of activity of the retinal pigmentation pathway, which possibly modulates local VEGF, PEDF and exosome secretion, affect ROP development? Indeed, evidence from the literature suggests so: first, children born to mothers with dark pigmented skin are less likely to develop ROP (Husain et al., 2013). Next, it is clear that VEGF secretion plays a central role in ROP pathology (Cavallaro et al., 2014;Zhu et al., 2015). Third, PEDF is considered an inhibitor of angiogenesis and could protect against abnormal vessel growth (Xi, 2020). Indeed, PEDF administered to the eyes of an ROP mouse model reduced neovascularization (Zhang et al., 2016). Finally, while research concerning the role of exosomes in ROP is lacking, they possibly play a mechanistic role in ROP: previous research, albeit in AMD context, showed that VEGFR and VEGFR mRNA containing exosomes could exacerbate aberrant vasculature formation (Atienzar-Aroca et al., 2016).

Conclusion
We propose here a retinal pigmentation pathway linking all (known) albinism causing genes, their function and their (ocular) phenotype together into one cohesive framework. We have explored the possible processes up and downstream that could influence neural retinal development and the albinism phenotype. Upstream, the role for TYR, L-DOPA and GPR143 in the pathology has been reasonably documented. Downstream of GPR143, SLC38A8 (FHONDA), together with a range of (candidate) signalling molecules that play an essential role in retinal development, optic nerve guidance and albinism. Taken together, this report opens new and systematic avenues for research into albinism and signalling from RPE to neural retina during retinal development. We suggest that the proposed retinal pigmentation pathway opens new avenues for development of new therapeutics in human albinism, but may also have an impact on other retinal pathologies involving melanosomal pigment abnormalities.