Candidate pathways for retina to scleral signaling in refractive eye growth

Abstract

The global prevalence of myopia, or nearsightedness, has increased at an alarming rate over the last few decades. An eye is myopic if incoming light focuses prior to reaching the retinal photoreceptors, which indicates a mismatch in its shape and optical power. This mismatch commonly results from excessive axial elongation. Important drivers of the myopia epidemic include environmental factors, genetic factors, and their interactions, e.g., genetic factors influencing the effects of environmental factors. One factor often hypothesized to be a driver of the myopia epidemic is environmental light, which has changed drastically and rapidly on a global scale.

In support of this, it is well established that eye size is regulated by a homeostatic process that incorporates visual cues (emmetropization). This process allows the eye to detect and minimize refractive errors quite accurately and locally over time by modulating the rate of elongation of the eye via remodeling its outermost coat, the sclera. Critically, emmetropization is not dependent on post-retinal processing. Thus, visual cues appear to influence axial elongation through a retina-to-sclera, or retinoscleral, signaling cascade, capable of transmitting information from the innermost layer of the eye to the outermost layer.

Despite significant global research interest, the specifics of retinoscleral signaling pathways remain elusive. While a few pharmacological treatments have proven to be effective in slowing axial elongation (most notably topical atropine), the mechanisms behind these treatments are still not fully understood. Additionally, several retinal neuromodulators, neurotransmitters, and other small molecules have been found to influence axial length and/or refractive error or be influenced by myopigenic cues, yet little progress has been made explaining how the signal that originates in the retina crosses the highly vascular choroid to affect the sclera.

Here, we compile and synthesize the evidence surrounding three of the major candidate pathways receiving significant research attention — dopamine, retinoic acid, and adenosine. All three candidates have both correlational and causal evidence backing their involvement in axial elongation and have been implicated by multiple independent research groups across diverse species. Two hypothesized mechanisms are presented for how a retina-originating signal crosses the choroid — via 1) all-trans retinoic acid or 2) choroidal blood flow influencing scleral oxygenation. Evidence of crosstalk between the pathways is discussed in the context of these two mechanisms.

1. Introduction

Myopia, commonly known as “nearsightedness”, describes a mismatch between an eye’s optical power and its geometry where the eye is too long for its optics. This mismatch results in light focusing in front of the retina instead of directly on it (i.e., a refractive error) and is typically the result of excessive axial elongation of the eye rather than improper development of optical power (Flitcroft et al., 2019). While myopic refractive errors are commonly treated by correcting the optical power of the eye via spectacles or contact lenses, these optical interventions do not slow axial elongation or myopia progression (Wildsoet et al., 2019). In the United States, myopia prevalence increased from about 25% in 1970 to 40% in 2000 (Vitale, 2009), and in Singapore, myopia reportedly affects more than 80% of young adults (Morgan et al., 2012). Predictions estimate that about 50% of the global population will be myopic by 2050 unless more effective preventative interventions or treatments of the underlying biological cause are identified (Holden et al., 2016). With excessive elongation being a risk factor for many additional sight-threatening disorders (Grytz et al., 2020), this epidemic requires additional study of the mechanisms of myopigenesis to develop interventions capable of resolving the underlying biological mismatch.

The rapid rise in prevalence is thought to be real, not due to increased identification, and driven primarily by environmental factors rather than drifting population genetics (Dolgin, 2015; Goldschmidt and Jacobsen, 2014; Vitale, 2009). While polygenic risk scores obtained from genome-wide association studies (GWAS) can be modestly predictive for severe myopia, they do not perform well for mild and moderate myopia (Ghorbani Mojarrad et al., 2020), which make up the vast majority of cases. Instead, the pathogenesis of myopia is thought to be environmentally driven by “myopigenic” cues, a hypothesis dating back to at least the late 16^th century, when near work was first proposed as causing myopia (de Jong, 2018). In the last century, the types of environmental visual cues presented to an eye during adolescence have been found to be highly influential over its mature size, with the homeostatic process linking vision and eye size being termed emmetropization (Troilo et al., 2019). Thus, the global increase in myopia throughout the 20^th century occurring during a period of increasing urbanization and widespread adoption of artificial electric lighting is likely not coincidental. Indoor environments are known to contain very different visual cues than outdoor, natural environments, e.g., different luminance, spatial, and spectral properties (Flitcroft et al., 2020; French et al., 2013).

The rise in time spent indoors and in urban environments is a likely driver of the myopia epidemic (Flitcroft, 2013), supported by the often-repeated finding that increasing time spent outdoors displays protective effects against myopigenesis. A study of monozygotic twins in the UK (n=64 pairs), where each pair had a ≥2 diopters (D) difference in refractive error between them, found that the less myopic of the pair tended to spend more time outdoors or doing outdoor sports, while the more myopic individuals more often lived in an urban area, performed more near work, and had greater professional status (Ramessur et al., 2015). Other studies have shown that children who spent more time outdoors (assessed via questionnaire or objective light measurement) showed modest, but significant, slowing of myopia progression (Jin et al., 2015; Read et al., 2014; Rose et al., 2008; Wu et al., 2013). The protective effect also appears to be quite variable. A study of monozygotic twins in China (n=490 pairs) found that near work and time outdoors only explained roughly 3% of the discordance in refractive error between twins (Ding et al., 2018). These discrepancies could possibly be explained by low sensitivity of questionnaire data for exposure measurements; however, it is also possible that variation in outdoor visual environments (outdoors in a city around buildings versus outdoors in a forest) and/or genetic factors that influence susceptibility to myopigenic cues could influence the amount of protection conferred.

Significant progress has been made in not only identifying myopigenic/protective visual cues, but also interventions capable of slowing myopigenesis, including optical, behavioral, and pharmacological methods (Smith and Walline, 2015; Wildsoet et al., 2019). Despite this, our understanding of the signaling mechanisms underlying myopigenesis is still limited. This is exemplified with the clinical use of topical atropine; atropine has been the most common pharmacological intervention for myopia control for over a century now, yet little consensus has been reached regarding its mechanism of action in the eye (Galvis et al., 2016; Mathis et al., 2020; Upadhyay and Beuerman, 2020).

Nevertheless, much has still been learned about general characteristics and requirements of myopigenic signaling. The mature eye size is influenced by visual cues detected by the retina. Remodeling of the sclera, the white outermost layer of the posterior eye, is causal to modulating eye size and occurs during myopigenesis (Boote et al., 2020; Grytz, 2018). The signaling does not appear to require post-retinal processing of the visual stimulus and thus is largely contained to the posterior eye wall (McFadden and Wildsoet, 2020; Norton et al., 1994; Troilo et al., 1987; Troilo and Wallman, 1991; Wildsoet, 2003; Wildsoet and Pettigrew, 1988; Wildsoet and Wallman, 1995). Myopigenic visual cues presented to the retina therefore result in a signal, likely chemical in nature (e.g., signaling molecules), capable of 1) propagating across the retinal pigment epithelium (RPE) and vascular choroid and 2) influencing remodeling processes in the sclera. How environmental visual cues interact with genetics to induce myopigenesis in some people but not others and what molecules are involved in the retina-to-sclera (retinoscleral) signaling pathways are currently open questions and the subject of active research (de Jong, 2018; Dirani et al., 2009; Huang et al., 2015; Jones-Jordan et al., 2012; Rose et al., 2008).

This review aims to summarize and contextualize recent findings related to a subset of signaling molecules and pathways (dopamine, retinoic acid, and adenosine), as opposed to serving as a fully comprehensive review of every molecule involved in retinoscleral signaling (for a broader review, see (Summers et al., 2021)). These signaling pathways were selected as the most likely to directly regulate myogenesis due to each having been correlated with and causally linked to axial length and myopia numerous times across many different research groups and species. Additionally, each has been suggested to act in multiple tissues throughout the retinoscleral cascade.

2. Emmetropization and homeostasis of eye size

The size of a mature eye is determined by a combination of growth and elongation, both of which appear to involve the sclera. Here, growth is defined as an increase in scleral tissue volume, which leads to an overall increase in ocular dimensions. In contrast, axial elongation is a process that modulates the axial length of the eye without increasing scleral tissue volume, which in mammals is believed to be determined primarily by remodeling of the sclera (Lim et al., 2011). This remodeling includes both the reorganization of existing tissue components (primarily collagen) and altered synthesis/degradation of others (collagen and proteoglycans/glycosaminoglycans). During remodeling, the sclera is more extensible, permitting elongation with either no change or a small decrease in overall volume (Backhouse and Gentle, 2018) and leading to a thinner sclera (for recent reviews, see (Boote et al., 2020; Grytz, 2018)).

In the human prenatal period, eye size increases rapidly via growth, closely matching the general growth curve of the body (Mutti et al., 2005; Watanabe et al., 1999). Upon birth, infants typically have mismatches in the power and geometry of their eyes (i.e., refractive errors), biased towards the eyes being too short for their optical power, a state termed hyperopia (Chakraborty et al., 2020; Mutti et al., 2005; Watanabe et al., 1999). Growth of the eye may only occur for about two years (Shen et al., 2016); however, eyes continue to elongate for many years, initially quite rapidly but slowing exponentially, reaching mature size around the teenage years (Fledelius et al., 2014; Fledelius and Christensen, 1996). By the age of 6, significant emmetropization of the population has generally occurred (characterized by a significant decrease in variance of refractive errors) (Chakraborty et al., 2020); however, eyes that will continue on to be emmetropic later in life are still on average hyperopic by about 1 diopter (Figure 1A). The remaining hyperopic refractive errors are corrected over the next decade while the eye continues to elongate and the optical components mature (Gordon and Donzis, 1985; Hagen et al., 2019; Jones et al., 2005; Mutti et al., 2005). Thus, throughout the first nearly two decades of life, eye development occurs in a complex yet highly coordinated manner, evidenced by the fact that only a minority (albeit a rapidly growing one) develop significant refractive errors.

Figure 1:

(A-C) Real distributions of refractive errors in various populations. Vertical red dashed line indicates no refractive error. A) Infant populations show approximately Gaussian distributions of refractive error, the majority being hyperopic on average (dashed line). By age 6, emmetropization has reduced the variance. (B,C) As populations age, refractive errors become increasingly leptokurtic and can be well described by fitting two separate Gaussian distributions, representing properly emmetropized (red) and dysregulated (blue) populations. D) Hypothetical paths of refractive development. Dysregulated populations could be explained by a combination of failure to initially emmetropize (A), or failure maintain emmetropia (B,C). Modified from Flitcroft, 2013 and Chakraborty et al, 2020.

The homeostatic process responsible for this coordinated development is called emmetropization, and it appears to function via controlling the rate of axial elongation. Emmetropic eyes can properly coordinate their rate of axial elongation such that the length of the eye matches the optical power and results in little to no refractive error. In most adult human populations, emmetropes are overrepresented, leading to a non-Gaussian leptokurtic distribution of refractive errors (Figure 1B,C) (Flitcroft, 2013). While refractive errors follow a mostly Gaussian distribution at birth and through early adolescence (Flitcroft, 2013), the number of myopes continue to increase. These statistical features, among others, reflect a tightly controlled homeostatic mechanism that is now understood to incorporate visual signals, primarily, blur.

Interestingly, many features that differ between indoor and outdoor visual environments have been demonstrated to be disruptive to achieving or maintaining emmetropia (for in-depth review see (Lingham et al., 2020; Norton, 2016; Rucker, 2019; Schaeffel and Feldkaemper, 2015)). Sunlight is significantly brighter than most indoor environments, and this increased luminance has been shown to be protective in chicks (Ashby et al., 2009; Ashby and Schaeffel, 2010; Chen et al., 2017), tree shrews (Norton and Siegwart, 2013), guinea pigs (Li et al., 2014), and macaques (Smith et al., 2013, 2012) compared to typical indoor lighting. Additionally, emmetropic children have been observed to spend more time in these bright light levels (Read et al., 2014). However, a recent re-analysis of the study by Read et al. found that children with myopia spent less time in both bright and dim light (Landis et al., 2018), and both luminance conditions were demonstrated to be protective in mice (Landis et al., 2021). Additionally, when chicks were exposed to sunlight daily, they had shorter myopic axial elongation (Ashby et al., 2009), but when they were reared outdoors, acute myopic protection did not last (Stone et al., 2016).

Sunlight also contains a broader spectrum of wavelengths than indoor light, including a larger proportion of shorter wavelengths (Krutmann et al., 2014; Thorne et al., 2009). Visual stimulation with violet light slowed myopia progression in human clinical studies (Torii et al., 2017), and short-wavelength light protected mice (Jiang et al., 2021; Strickland et al., 2020), guinea pigs (Yu et al., 2021), and chicks (Torii et al., 2017; Wang et al., 2018) from induced myopia. However, in tree shrews and rhesus monkeys, long-wavelength red light was protective, suggesting potential species differences (Gawne et al., 2017; Smith et al., 2015; Ward et al., 2018).

The protective effect of the outdoors could also be related to the spatial frequency content of the environment since outdoor environments contain more mid-level (~10cpd) and higher spatial frequencies than artificial environments (Flitcroft et al., 2020). Manipulating the visual experience to degrade (Howlett and McFadden, 2006; Norton and Rada, 1995; Pardue et al., 2013; Shen et al., 2005; Troilo and Nickla, 2005; Wallman et al., 1978) or reduce mid-high spatial frequencies leads to axial elongation in chicks (Bartmann and Schaeffel, 1994; Tran et al., 2008), guinea pigs (Bowrey et al., 2015), mice (Pardue et al., 2008), and monkeys (Smith and Hung, 2000), and evidence exists that extends this effect to humans. Adolescents with cataracts, which significantly reduce sensitivity to high spatial frequencies (Lewis and Maurer, 2005), often develop severe myopia (He et al., 2017). More directly, it has recently been shown that movies carefully low-pass filtered to match spatial frequency spectra of imposed optical defocus can acutely influence axial length in emmetropic and myopic humans (Swiatczak and Schaeffel, 2021). Thus, axial blur, luminance, chromaticity, and spatial frequency content have all been demonstrated to be myopigenic cues. Most animal studies of myopia impose one or more of these visual cues on the animal, commonly axial blur via powered lenses (lens induced myopia, LIM) or diffuser goggles (form-deprivation myopia, FDM).

Myopic eyes have failed to either properly emmetropize or maintain emmetropia (Figure 1D), and it is now well established that these failures cannot be traced back to a single source. The variable age of myopia onset in adolescents and the much larger variance in the distribution of refractive errors in myopic populations (Flitcroft, 2013) (Figure 1) supports many distinct origins of a disrupted homeostatic process. Visual environments contain many types of myopigenic visual cues, some of which may drive myopigenesis to differing degrees. However, the retinal circuitry used to sense these cues is highly complex, involving many distinct cell types and signaling pathways and thus has many points at which one’s genetics could drive myopia or influence susceptibility to myopigenic cues.

While many genetic abnormalities have been demonstrated to disrupt emmetropization and confer a degree of heritability to myopia (with estimates ranging from 15% to 98%) (Tedja et al., 2019), the types and durations of visual cues presented to the eye appear especially critical to emmetropization and its maintenance. Thus, it is important to study the mechanisms by which these myopigenic cues signal to increase axial elongation.

3. Retinoscleral signaling

In mammals, myopigenic visual cues appear to influence eye size primarily by increasing the rate of axial elongation through altering scleral remodeling (Troilo et al., 2019), as opposed to the avian model in which the cartilaginous layer of the sclera grows. The influence of visual cues on the sclera occurs quite locally — myopigenic visual cues can influence eye size unilaterally (if only one eye is presented the stimulus), without a functioning connection to the visual cortex (Norton et al., 1994; Troilo et al., 1987; Wildsoet, 2003; Wildsoet and Pettigrew, 1988), and regionally within an eye (if presented a spatially nonuniform myopigenic stimulus, e.g., hemi-diffusers) (Diether and Schaeffel, 1997; Norton and Siegwart, 1995; Smith et al., 2014, 2010, 2009; Wallman et al., 1987). Thus, the retina senses and encodes visual cues and initiates a signaling cascade that propagates to regions of the sclera located posterior to the region of the retina that sense the cue, ultimately leading to scleral remodeling and modulation of eye size.

While this general framework for retinoscleral signaling is well supported, the specifics of the process are still unknown. Many types of retinal signaling molecules and pathways have been associated with myopigenesis, and the signaling dynamics of some have been studied in detail (most notably, dopamine). However, efforts to incorporate retinal research into a pathway spanning the entire ocular wall have been much less successful. Additionally, different myopigenic cues may have differing retinal mechanisms, evidenced by retinal interventions that protect against lens-induced myopia but not form-deprivation (or vice versa) (Bitzer et al., 2000; Dong et al., 2011; Wang et al., 2014). However, there is no evidence to date showing different myopigenic cues result in different outcomes in scleral remodeling. Thus, various types of retinal signaling may converge at or before the choroid, resulting in a graded signal that combines various myopigenic cues and determines if the eye will increase, maintain, or slow its axial elongation (sometimes referred to as a GO, STAY, STOP signal (Guo et al., 2019)).

In the following sections, we focus on the roles and influence of dopamine, retinoic acid, and adenosine in myopigenesis (Figure 2). While these represent only a subset of the many signaling molecules reported in the literature, we focus on these three due to 1) robust reports of causal influence on and correlations with refractive development, 2) localization of the signaling molecule/receptor in multiple ocular tissues, 3) evidence from a variety of species (especially mammalian), and 4) corroboration by multiple independent research groups. While little direct evidence currently exists linking these three pathways in the eye, we suggest that these pathways likely represent key components of the signaling cascade that warrant future study. Additionally, two possible points of signal convergence are discussed alongside supporting evidence.

Figure 2:

Cartoon representation of the human posterior eye wall and the path of a hypothetical retinoscleral signal. Arrows below the diagram signify where each molecule has been implicated as a potential signaling molecule. The proportions of each layer are approximately to scale. NFL: nerve fiber layer, GCL: ganglion cell layer, IPL: inner plexiform layer, INL: inner nuclear layer, OPL: outer plexiform layer, ONL: outer nuclear layer, RPE: retinal pigment epithelium.

3.1. Dopamine

The role of retinal dopamine (DA) in myopigenesis has been an active area of research for the last 30 years since the first report that DA may link visual signals with myopic axial elongation (Stone et al., 1989). In this report, Stone and colleagues discovered that levels of DA and its metabolite, 3,4-dihydroxyphenylacetic acid (DOPAC), were significantly reduced in retinas of chickens with experimental myopia (FDM). This study suggested that retinal dopaminergic (DAergic) signaling could be a principal modulator of myopic changes in the retina. Since then, extensive effort has been spent investigating DAergic signaling (Table 1), mostly through measurements of DA and its metabolites, in retinas of myopic eyes or those exposed to myopigenic stimulus (for a recent in-depth review, see (Zhou et al., 2017)). In this section, we will present evidence for DA as the starting point for the retinoscleral signaling cascade that leads to myopic axial elongation.

Table 1: Potential of dopamine to be a retinoscleral signal in refractive eye growth.

Column C.1: Treatments given to animals. C.2: General description of main effect of treatment. Columns C.3–6: Additional outcomes (C.3) studied with corresponding treatments. In A) Endogenous, C.4–6 are locations of measurements, and (+/−) represent whether receptors are expressed. In B) Exogenous and C) Genetic, C.4–6 are different locations of the treatment. Cells are shaded to highlight trends. Orange/blue shading indicate an increase/decrease in signaling or elongation. Gray indicates inconclusive or no effect. Table cells are split when studies came to different conclusions. AL: Axial Length, Conc: Concentration, Expr: Expression, Synth: Synthesis, FD: Form-Deprivation, NL: Negative Lens, PL: Positive Lens, Rec: Recovery from myopia, DA: Dopamine, DOPAC: 3,4-Dihydroxyphenylacetic acid, DAC: Dopaminergic amacrine cell. See Supplemental Table 1 for associated references.

3.1.a. Dopamine signaling pathway

Retinal DA is produced and released by a subset of DAergic amacrine cells (DACs) that are sparsely tiled across the retina (Witkovsky, 2004). Despite their low numbers, DACs have long dendrites that can span millimeters of the retina and can extend from the inner plexiform layer to the outer plexiform layer and ganglion cell layer (Witkovsky, 2004). DA can act through local synaptic release from DAC varicosities and dendritic spines (Dacey, 1990) or extra-synaptically through paracrine diffusion (Puopolo et al., 2001) across the retina to activate multiple types of postsynaptic and extra synaptic receptors. DA receptors are located on almost all retinal cells and fall into two families, D1-like receptors (D1 and D5) and D2-like receptors (D2, D3, and D4), which respectively increase and decrease cyclic adenosine monophosphate (cAMP) (Witkovsky, 2004). D1 receptors are located on bipolar, horizontal, amacrine, and ganglion cells, D2 receptors on DACs, D4 receptors on rod and cone photoreceptors, and D1, D2, D4, and D5 receptors on the RPE (Baba et al., 2017; Witkovsky, 2004). After DA binds to its receptor, it is metabolized to DOPAC (Cohen et al., 1983) which reflects DA release and turnover (Megaw et al., 2001).

DA release and synthesis is light-dependent, increasing with continuous diffuse and flickering light stimulation (Bauer et al., 1980; Boelen et al., 1998; Cohen et al., 2012; Godley and Wurtman, 1988; Iuvone et al., 1978; Kramer, 1971; Proll et al., 1982) and regulated by circadian rhythms (Doyle et al., 2002). Consequently, DACs are activated by light onset visual signals through three potential pathways: 1) rod photoreceptor-mediated input through the rod-rod bipolar-AII amacrine-ON cone bipolar circuit 2) direct cone photoreceptor-ON cone bipolar cell connection, and 3) melanopsin retinal ganglion cell (mRGC) axon collateral input (Boelen et al., 1998; Dumitrescu et al., 2009; Prigge et al., 2016; Zhang et al., 2012, 2008; Zhao et al., 2017). Therefore, DACs can release DA from stimulation of multiple ON pathways (Qiao et al., 2016) that are active under different environmental conditions. The actions of DA on retinal signaling is widespread and occurs at multiple levels from photoreceptors to ganglion cells (Roy and Field, 2019). Therefore, the extensive functions that DA has in retinal signaling supports its potential role in the development of myopia due to altered visual input, as was first noted by Stone et al. (Stone et al., 1989).

3.1.b. Evidence for dopamine signaling in experimental myopia

DA is generally understood to be a ‘stop’ signal in myopic axial elongation (Feldkaemper and Schaeffel, 2013). In chickens, guinea pigs, and monkeys, DA and DOPAC levels are reduced after form-deprivation (Dong et al., 2011; I. Papastergiou et al., 1998; Iuvone et al., 1989; Stone et al., 1989; Sun et al., 2018) and lens defocus (Guo et al., 1995; Ohngemach et al., 1997). Interestingly, the reduction in DOPAC was specific to only the region of the retina deprived of form vision (Ohngemach et al., 1997; Stone et al., 2006), matching the property of regional specificity observed in visually-mediated axial elongation. However, altered DA metabolism associated with myopigenic stimuli may not be conserved across species, as form-deprivation did not alter retinal DA levels in mice (Chakraborty et al., 2015, 2014; Park et al., 2014; Wu et al., 2015). However, mice, as well as rabbits and guinea pigs, show inhibited FDM (Gao et al., 2006; Junfeng et al., 2010; Landis et al., 2020; Mao et al., 2016; Mao and Liu, 2017) following injections of L-DOPA, the precursor to DA, or other DAergic activators. These findings suggest that while direct measurements of DA levels in mice were not different in FDM, activation of the whole DAergic system could still slow myopic eye growth.

Furthermore, in multiple species, DA receptor agonists prevent FDM, but do not appear to be as effective for LIM (Ashby et al., 2007; Dong et al., 2011; Iuvone et al., 1991; McCarthy et al., 2007; Rohrer et al., 1993; Schmid and Wildsoet, 2004; Stone et al., 1989; Yan et al., 2015), and it is still unclear the reason for this difference. One explanation could be that the visual cues that cause FDM and LIM initiate two distinct pathways that are differently influenced by dopamine; however, this cannot yet be concluded. Direct comparisons between the two models may be confounded by other unmatched features of the visual cues, e.g., the spatial frequency spectra or optical aberrations introduced by the lenses.

Surprisingly, reduced axial elongation associated with increased DA is only effective when visual input is disrupted (e.g., form deprivation, lens defocus) and not during normal vision (Dong et al., 2011; Junfeng et al., 2010; Landis et al., 2020; Rohrer et al., 1993; Yan et al., 2015). These findings suggest that the activation of DA pathways is important but not sufficient to affect eye growth. One hypothesis is that visual environments deficient in certain features change activation of pathways that drive DAergic signaling (e.g., ON pathways). Recent work shows that stimuli which overstimulate ON pathways lead to anti-myopigenic outcomes, e.g., choroid thickening (Aleman et al., 2018) and increased vitreal DA (Wang et al., 2019). However, the direct connections between environmental stimulation of ON pathways, DA, and myopia susceptibility are still not fully elucidated.

Unlike activation of DA pathways, depletion of DA pathways prevents FDM but has no effect on LIM in chickens and fish (Kröger et al., 1999; Li et al., 1992; Schaeffel et al., 1995, 1994) (Table 1). In contrast, decreasing DA with 6-hydroxydopamine in mice results in a myopic refractive shift with normal vision and increased FDM (Wu et al., 2016). Additionally, when retinal tyrosine hydroxylase, the rate-limiting enzyme in DA synthesis, is conditionally knocked out, mice have reduced DA levels and develop spontaneous myopia in normal conditions, while responding like wild-type mice to FDM (Bergen et al., 2016).

Interestingly, the dopamine receptor pathways involved in the protective effects of dopamine and bright light in myopia depends on the animal model (Zhou et al., 2017). For example, in chickens, D2 receptor activation inhibited FDM and LIM (Nickla et al., 2010; Rohrer et al., 1993; Schmid and Wildsoet, 2004; Stone et al., 1989) while blocking D2 receptors inhibited the protective effect of light on FDM and LIM (Ashby and Schaeffel, 2010; McCarthy et al., 2007). Additionally, blocking D2, but not D1, receptors attenuated the inhibitory effects of levodopa and dopamine on FDM and LIM (Thomson et al., 2020a, 2020c). Likewise, activation of primarily D2 receptor pathways reduced FDM in tree shrews (Ward et al., 2017). However, in the guinea pig, D1 receptor activation inhibited myopic refractive development and FDM while D2 receptor activation tended to do the opposite; however, this modulation varied with the DA receptor agonist (Zhang et al., 2018). In the mouse, it appears to be more complicated. Blocking D1 receptors attenuated the protective effect of bright light exposure during FDM (Chen et al., 2017). Additionally, activation of all DA receptors in wild-type mice and D2, but not D1, receptor knockout mice inhibited FDM (Huang et al., 2018) supporting a D1 receptor mechanism for myopigenesis in the mouse model. However in other studies, D2 receptor agonists and antagonists altered FDM in a dose-dependent manner and the protective effects of a D2 antagonist disappeared in D2 receptor knockout mice (Huang et al., 2022, 2020, 2014). Lastly, a D1 receptor mechanism is further supported in human genetic studies; a GWAS utilizing data from the Consortium for Refractive Error and Myopia (CREAM) cohort and the 23andMe, Inc. participants reported that a single nucleotide polymorphism (SNP) located near the Dopamine Receptor 1 gene (DRD1), a type of D1-like receptor, was significantly associated with refractive error (untransformed spherical equivalent) (Tedja et al., 2018).

3.1.c. Influence of dopamine signaling on ocular tissues

There is strong evidence showing a clear but complicated relationship between DAergic pathways and myopia. Retinal DA appears to have a profound influence over axial length under certain conditions. For example, in chicks, topical or intravitreal application of DA, DA receptor agonists, or levodopa reduced axial length in form-deprived or negative lens-treated eyes compared to control eyes (Nickla et al., 2019; Thomson et al., 2021, 2020b, 2020a, 2020c). Moreover, in mice, intravitreal injections of the DA agonist apomorphine attenuated axial length changes after form-deprivation (Huang et al., 2018). Additionally, D1 receptor antagonist blocked the protective effect of bright light on axial length changes in FDM (Chen et al., 2017). Likewise, intravitreal injections of high doses of D2 DA receptor agonists or antagonists reduced vitreous chamber elongation in the form-deprived tree shrew eye (Ward et al., 2017). However, modulation of DAergic signaling by DA, DA receptor agonists or antagonists, and levodopa did not affect axial length in otherwise untreated eyes (Thomson et al., 2020a, 2020c; Zhang et al., 2018), further supporting the notion that DA effects on axial elongation and myopia require altered visual input.

Despite the significant effects of DA on axial length changes, DA has not been associated with scleral signaling, and there is no evidence that DA itself reaches the sclera to affect structural remodeling. However, choroidal thickness is well established to change with myopia and correlated with axial elongation (for a recent review see (Liu et al., 2021), and there is strong evidence that dopamine also affects choroidal thickness. In chickens, choroidal thickness is regulated through D2 receptors, where D2 receptor agonists increase thickness after brief periods of unaltered vision following LIM and D2 receptor antagonists inhibit this increase after FDM (Nickla et al., 2010; Nickla and Totonelly, 2011). However, in rabbits, dopamine-induced vasodilation of the choroid was mediated through D1 receptors (Reitsamer et al., 2004). Based on the previous evidence, it is likely that the interaction between DA and choroidal thickness occurs outside of the neural retina and not within the choroid. While tyrosine hydroxylase has been localized to the choroid (Klooster et al., 1996), no evidence to date shows significant choroidal expression of DA receptors. This leaves the RPE as a tissue potentially able to translate dopaminergic signaling to choroidal changes, that ultimately may influence axial elongation.

The RPE is a single layer of pigmented cells whose apical surface is embedded between photoreceptor outer segments and whose basal surface is lined with Bruch’s membrane, separating it from the vascular choroid. The RPE carries out diverse and important set of roles for visual function, but most relevant is its role in ionic transport (Strauss, 2005), which 1) may affect fluid flow into the choroid/choroidal thickness (Zhang and Wildsoet, 2015) and 2) is altered during myopigenesis. Both the gene for chloride channels and expression of chloride channels and transporters were down-regulated after lens defocus (Seko et al., 2000; Zhang et al., 2011). Moreover, during recovery from FDM in chickens, sodium and chloride ions decrease in the RPE and increase in the choroid, which correlates temporally with a thickening of the choroid (Liang et al., 2004). Ionic flow across the retina, particularly with regards to chloride, sodium, and potassium, have been implicated in and are affected by experimental myopia (Crewther et al., 2008, 2006) and hypothesized to regulate fluid flow that could change the size of various ocular layers. Since changes in cell size and transporters can affect fluid balance across cells, regulation at the RPE provides an intriguing potential link between the retina and the classical changes in choroidal thickness associated with myopia reported more than 20 years ago (Wallman et al., 1995; Wildsoet and Wallman, 1995).

DA is a promising signal that could link the RPE to changes in the choroid due to RPE expression of D1, D2, D4, and D5 DA receptors in chicks and mice, likely on both the apical and basolateral membranes (Baba et al., 2017; Rymer and Wildsoet, 2005). When DA is applied to the basolateral surface, chicken RPE cells depolarize, due to the opening of chloride channels (Gallemore and Steinberg, 1990). This response is dependent on DA concentration and likely involves interaction effects with other receptors, possibly adrenergic receptors. For example, with low DA concentrations, DA receptors and B-adrenergic receptors are active, reducing the probability of chloride channels opening. However, when DA concentrations are high, alpha-adrenergic receptors become active which may also contribute to chloride channel opening (Rymer and Wildsoet, 2005). These findings on chloride channel activity in vitro are supported by in vivo ERG recordings using intravitreal injections of DA or DA receptor antagonists and measuring the chloride channel-mediated ERG component (Gallemore and Steinberg, 1990; Sato et al., 1987; Wioland et al., 1990). Regardless of the specific pathway, these data suggest that DAergic activity has the potential to regulate ionic flow and thus fluid transfer across the RPE into the choroid. However, whether or not this effect is large enough to significantly impact choroidal thickness and myopigenesis has not been directly shown.

3.1.d. Summary

In summary, DA has long been known to be causally involved in myopigenesis and is likely the starting point linking at least some types of myopigenic cues to scleral remodeling. However, its direct influence via receptor binding appears to be limited to the retina and RPE. It is possible that the RPE serves as the link between myopigenic stimuli, DA, and choroidal thickness by regulating choroidal swelling via dopamine receptors. Additional research is necessary to elucidate signaling occurring at the RPE and how it may propagate and influence the choroid.

3.2. Retinoic acid

Retinoic acid (RA) is a metabolite of vitamin A and a critical signaling molecule involved in numerous autocrine and paracrine developmental and physiological processes (Cvekl and Wang, 2009), regulating the transcription of over 100 genes (Balmer and Blomhoff, 2002). With advances in retinoid measurement technologies, evidence suggests that all-trans RA (atRA) is important in postnatally regulating the growth of many organ systems, including the eye (Summers, 2019). It is a promising candidate in signaling myopic axial elongation for many reasons: it is present in all eye wall tissues, bidirectionally modulated by the direction of blur (Mao et al., 2012; McFadden et al., 2004), and is known to stimulate catabolic remodeling processes (Bonassar et al., 1997; Mertz and Wallman, 2000). However, despite significant evidence linking RA to myopia development, several discrepancies remain, complicated by the breadth of its biological roles and the spatial and temporal variability in its action.

3.2.a. RA signaling pathway

Retinal RA is thought to be synthesized primarily by RPE cells as a byproduct of phototransduction (Weiler et al., 1998); however, subpopulations of amacrine (Milam et al., 1997; Saari et al., 1995) and Müller cells (Edwards et al., 1992; Milam et al., 1990) have been shown to either synthesize or express enzymes that synthesize RA. Currently, the canonical pathway by which RA appears to exert its effects involves the all-trans form (Balmer and Blomhoff, 2002). The synthesis of atRA occurs in two steps: all-trans retinaldehyde is synthesized from all-trans retinol (vitamin A) via alcohol dehydrogenases (ADHs), which is subsequently oxidized into atRA via cytosolic retinaldehyde dehydrogenases (RALDH1/2/3). A dimer of a nuclear RA receptor (RAR) and retinoid-X receptor (RXR) binds atRA, subsequently binding DNA and influencing transcription (Balmer and Blomhoff, 2002). After synthesis, atRA can be bound by one of many retinoid binding proteins (RBPs) to facilitate intra- or extracellular transport or metabolized by CYP26 (Summers, 2019).

In vertebrates, retinal atRA concentration and synthesis have been demonstrated to be visually mediated. It is increased in response to both myopigenic cues (Mao et al., 2012; McFadden et al., 2004; Seko et al., 1998; Troilo et al., 2006) and greater luminance (McCaffery et al., 1996), although these increases may occur through different pathways. Together, this implies that the regulation of many retinal processes by light may occur at the transcriptional level via atRA, including light adaptation of horizontal cells (Pottek and Weiler, 2000; Weiler et al., 2000, 1999, 1998) and the transcription of arrestin, a critical phototransduction protein (Wagner et al., 1997).

3.2.b. Evidence for RA signaling in experimental myopia

Retinal atRA has been demonstrated to be increased after FDM and LIM in the guinea pig (Huang et al., 2011; Mao et al., 2012; McFadden et al., 2004) and chicken (Bitzer et al., 2000; Seko et al., 1998). It has also been observed to decrease with positive defocus (either via powered lenses or recovery from myopia) (McFadden et al., 2004). The significance of this bidirectional modulation of atRA in the retina is still not known. While recent genomics studies suggest that the sign of blur may be encoded through two distinct sets of genes (Tkatchenko et al., 2018), this does not imply distinct signaling pathways between the retina and the RPE.

Recently, atRA has been demonstrated to modulate the protective effect of short wavelength light against myopigenesis. A study in guinea pigs exposed to different wavelengths of light showed a retinal atRA dependence on chromatic content, with the retinal atRA concentrations decreasing in the shorter wavelength (blue) light. Additionally, guinea pigs given unilateral LIM and reared in white light had significantly more retinal atRA in the myopic eyes than the contralateral eyes. However, when reared in blue light, there was no difference between the treated and control eyes, and both eyes had less atRA than those reared in white light (Yu et al., 2021). Finally, inhibition of RA-synthesizing enzymes in the retina has also been demonstrated as protective against myopia development (Bitzer et al., 2000; Yu et al., 2021), suggesting that the influence of atRA on axial elongation may begin in the retina.

As has been found in the retina, choroidal atRA concentrations appear to be bidirectionally influenced by the direction of blur imposed on the retina (Table 2). This choroidal atRA is likely choroidally-derived, as it has been demonstrated that the choroid synthesizes a large amount of atRA, more than the retina and liver (Mertz and Wallman, 2000). Three sets of co-culture experiments have important implications regarding atRA and retinoscleral signaling. When eyes from chickens were exposed to hyperopic defocus/diffusers or myopic defocus (conditions that influence choroidal thickness) and the choroids subsequently dissected and cultured, atRA production was bidirectionally altered (Mertz and Wallman, 2000) (similar bidirectional modulation of atRA synthesis, although opposite trends, were found in marmosets (Troilo et al., 2006)). Second, a more complex experiment was carried out which exposed chicken eyes to various visual cues and co-cultured different combinations of tissues, assessing scleral remodeling when cultured with choroids from different visual conditions (form-deprived or recovering from myopia) (Marzani and Wallman, 1997). Scleral remodeling (quantified by glycosaminoglycan incorporation) was influenced by the visual condition from which the choroid came. Finally, Mertz and Wallman also showed that when sclera and choroid were cultured together, the choroid would rapidly release atRA into medium, and the sclera had the tendency to concentrate significantly more atRA than would be predicted by diffusion (the scleral tissue accumulated over 50% of the atRA released by the choroid, despite only making up 1% of the volume in culture) (Mertz and Wallman, 2000). Together, these studies make a compelling case that atRA can be produced in the choroid, has the tendency to be transported to the sclera, and can influence scleral remodeling. However, the exact source of the atRA is still unclear. A recent study aiming to characterize cells positive for RALDH2, an atRA synthesis enzyme, in the choroid of both humans and chicks found that choroidal atRA concentrations were partially controlled by proliferation of RALDH2+ cells (Summers et al., 2020). Yet, protein levels of RALDH2 were only found to be altered in chicks recovering from myopia, not in those with myopia (Rada et al., 2012; Summers et al., 2020).

Table 2: Potential of atRA to be a retinoscleral signal in refractive eye growth.

Column C.1: Treatments given to animals. C.2: General description of main effect of treatment. Columns C.3–6: Additional outcomes (C.3) studied with corresponding treatments. In A) Endogenous, C.4–6 are locations of measurements, and (+/−) represent whether receptors are expressed. In B) Exogenous, C.4–6 are different locations of the treatment. Cells are shaded to highlight trends. Orange/blue shading indicate an increase/decrease in signaling or elongation. Gray indicates inconclusive or no effect. Table cells are split when studies came to different conclusions. Bolded text indicates non-mammalian species used in study. AL: Axial Length, Conc: Concentration, Expr: Expression, Prolif: Proliferation, Synth: Synthesis, FD: Form-Deprivation, NL: Negative Lens, PL: Positive Lens, Rec: Recovery from myopia. See Supplemental Table 2 for associated references.

The association of retinoids with specific RBPs that influence transport may be another means by which atRA could reach the sclera. Recently, Summers and colleagues showed that apolipoprotein A-1 (apoA-1) may act as a RBP in the eye that can traverse the choroid. apoA-1 is produced by the chick choroid and is also upregulated by atRA (Summers et al., 2016). In the same study, it was found that despite cultured sclera not synthesizing apoA-1 at detectable levels, cultured sclera releases significant amounts of apoA-1 into the medium, implying transport from choroid to sclera (Summers et al., 2016) and possibly explaining the previously discussed finding from Mertz and Wallman showing the tendency for the sclera to concentrate choroidally-derived atRA. Together, these results suggest a directionality to the transport of atRA and may overcome one of the primary limiting factors of retinoscleral signaling.

In humans, multiple GWAS studies of refractive error and myopia have also identified SNPs in genes related to atRA signaling. For example, multiple studies have reported associations between refractive error and the retinol dehydrogenase 5 (RDH5) gene (Kiefer et al., 2013; Tedja et al., 2018; Verhoeven et al., 2013), which encodes the enzyme 11-cis retinol dehydrogenase and is expressed in the RPE. GWAS findings have also implicated the RPE-retinal G protein-coupled receptor (RGR) (Kiefer et al., 2013; Tedja et al., 2018), expressed in the RPE and Müller glia (Pandey et al., 1994), and the atRA receptor-related orphan receptor beta (RORB) (Tedja et al., 2018), expressed in retinal tissue (Jia et al., 2009) and a component of the circadian clock system (André et al., 1998).

While retinoscleral signaling by atRA appears to be well supported, many nuances remain to be studied. Perhaps the greatest discrepancy is in the opposite effects of axial elongation and atRA in mammals and chickens. In both, retinal atRA increases with myopigenic stimuli while choroidal and scleral atRA concentrations are negatively correlated with axial elongation in the chick but positively correlated in mammals (Table 2). A likely factor contributing to this are the different scleral compositions of the two species: mammals have a fibrous sclera and chicks have both a cartilaginous and fibrous sclera. The different resident cell types of the two tissues have been demonstrated to respond differently to atRA in the presence of TGF-β, arresting the proliferation of fibroblasts but not chondrocytes (Y. Seko et al., 1996). Additionally, co-culturing the choroid with fibrous sclera and cartilaginous sclera also lead to opposite outcomes on scleral remodeling (Marzani and Wallman, 1997). Thus, it is possible that there are significant differences in the physiological function of atRA as a retinoscleral signal. For example, in guinea pigs, RALDH2 was not found to be expressed in the choroid (Mao et al., 2012). Additional study of the function of atRA in mammalian models is required to determine the extent of species differences.

3.2.c. Influence of RA on ocular tissues

Unlike DA, atRA displays the ability to affect scleral remodeling processes that underly myopic axial elongation. While atRA is not produced in the sclera, scleral fibroblasts have atRA receptors and remodeling processes are influenced by atRA (Li et al., 2010; Troilo et al., 2006). In marmosets, RA concentration in retina, choroid, and sclera were found to be positively correlated with rate of axial elongation and negatively correlated to the rate of proteoglycan synthesis (Troilo et al., 2006). Systemic treatment of guinea pigs with atRA also led to axial elongation and an altered scleral microstructure with similarities to the outcome of myopigenic visual cues (McFadden et al., 2004), and treatment of mice led to myopic refractive errors, axial elongation, and altered scleral biomechanics (Brown et al., 2021).

However, exogenous atRA treatment has also been observed to lead to significant ocular growth without refractive errors in both chickens and guinea pigs, proportionally growing the eye and causing the lens of these animals to thicken instead of just causing axial elongation (McFadden et al., 2006, 2004). One possible explanation could be a dual role for atRA in signaling proportional eye growth and axial elongation, with younger animals more susceptible to growth rather than elongation.

3.2.d. Summary

In summary, atRA is one of the few chemical signals with evidence supporting the capacity to cross the choroid. However, its role in myopigenesis is still unclear. While exogenous atRA causes myopia to develop in some cases, in others it leads to excessive proportional growth of the eye. Additional mammalian research is required to make more general conclusions due to previous work being avian-specific, such as the role of apoA-1 as an RBP. However, with multiple genes related to atRA signaling and myopia implicated in GWAS studies, this is a promising pathway for further investigation.

3.3. Adenosine

Adenosine (Ado) is one of the most widely occurring organic compounds. It is a purine and is a basic building block of life as one of four nucleosides of DNA and RNA with a role in energy transport. Ado’s role as an extracellular signaling molecule was discovered in the early 20^th century (Drury and Szent-Györgyi, 1929). It has since been observed to modulate many cellular processes and act as a neuromodulator, exerting its actions through four types of membrane-bound, G-protein coupled receptors (AdoRs) - AdoRA₁, AdoRA_2A, AdoRA_2B, AdoRA₃ (Eltzschig, 2013). Similar to DA, AdoRs have opposing effects on cAMP production, with AdoA₁ and AdoRA₃ decreasing cAMP and AdoRA_2A, AdoRA_2B increasing cAMP by inhibiting or activating adenyl cyclase (AC), respectively (Spinozzi et al., 2021).

3.3.a. Adenosine signaling pathway

The exact mechanisms by which Ado reaches extracellular targets are complex and multivariate. Cells and interstitial fluids in tissues have basal concentrations of Ado in the nanomolar range (Fredholm, 2007). Many types of cells, including neurons and glial cells, have been demonstrated to release some amount of Ado from intracellular stores into the extracellular space (Brundege and Dunwiddie, 1998; Cotrina et al., 1998; Pascual et al., 2005), which appears to be increased by glutamate binding to N-methyl-D-aspartate (NMDA) receptors (Brambilla et al., 2005).

In general, extracellular Ado tends to dramatically increase with tissue activity, hypoxia, and other pathological states and biological stressors, increasing to the micromolar range (Fredholm, 2014; Haskó et al., 2018). However, this increase appears to be due to leakage or controlled release of adenosine triphosphate, which is subsequently degraded into adenosine monophosphate and then Ado (Antonioli et al., 2013; Eltzschig, 2013). Additionally, the extent to which these changes affect signaling is not clear, as Ado signaling is highly dynamic. Ado is rapidly metabolized to inosine and hypoxanthine, exhibiting a half-life of only ~1.5 seconds (Spinozzi et al., 2021), and equilibrative nucleoside transporters (ENTs) rapidly equilibrate imbalances in intra- and extracellular Ado (Lovatt et al., 2012).

In many cases, AdoRs are significantly activated by basal levels of Ado (Fredholm, 2007) and significantly blocked by commonly consumed amounts of caffeine, a nonselective inhibitor of AdoRs (Fredholm et al., 1999). However, the sensitivity of the cell or tissue to the ligands are highly dependent on the expression of the receptor, which has also been noted to vary significantly with diseased states and presence of stressors, most notably of which are related to the immune system and hypoxia. Hypoxic conditions have been shown to increase expression of AdoRA_2B (Kong et al., 2006) without affecting AdoRA_2A (Fredholm et al., 2007) and may influence trafficking of the receptors to the membrane (Arslan et al., 2002).

It is critical to note that these effects are very much cell type-dependent, and only a small fraction of the work on Ado signaling has occurred in ocular tissues. In the retina, it has been demonstrated that extracellular Ado signaling is controlled by light intensity and circadian rhythms, both of which have been implicated in myopigenesis (Ribelayga and Mangel, 2005).

3.3.b. Evidence for adenosine signaling in experimental myopia

The first connection of purinergic Ado signaling to myopigenesis was made with the discovery that 7-methylxanthine (7-MX), a metabolite of caffeine and a nonselective inhibitor of AdoRs, influenced scleral collagen and proteoglycan content in rabbits, opposite to that occurring with myopigenesis (Trier et al., 1999). Following this finding, many additional studies have also found a causal role of Ado on refractive state, axial elongation, and scleral remodeling (see 3.3.c and Table 3).

Table 3: Potential of adenosine to be a retinoscleral signal in refractive eye growth.

Column C.1: Treatments given to animals. C.2: General description of main effect of treatment. Columns C.3–7: Additional outcomes (C.3) studied with corresponding treatments. In A) Endogenous, C.4–7 are locations of measurements, and (+/−) represent whether receptors are expressed. In B) Exogenous and C) Genetic, C.4–7 are different locations of the treatment. Cells are shaded to highlight trends. Orange/blue shading indicate an increase/decrease in signaling or elongation. Gray indicates inconclusive or no effect. Table cells are split when studies came to different conclusions. AL: Axial Length, Expr: Expression, FD: Form-Deprivation, NL: Negative Lens, PL: Positive Lens, Rec: Recovery from myopia. See Supplemental Table 3 for associated references.

Despite significant experimental evidence, most details of the role of purinergic signaling in myopigenesis are still unclear. All four subtypes of AdoRs have been found to be expressed in all layers of the posterior eye wall of rhesus monkeys and guinea pigs (Beach et al., 2018; Cui et al., 2010) and in human scleral fibroblasts (Cui et al., 2008). A small epigenetic study of youth with high myopia (n=18 cases, 18 controls) reported hypermethylation of the AdoA_2A receptor gene (ADORA2A) in peripheral blood samples (Vishweswaraiah et al., 2019), suggesting Ado signaling may be influenced by myopia. Additionally, when guinea pigs were deprived of form vision, protein levels of AdoA₁ / AdoA_2B in the retina were elevated (Cui et al., 2010).

The poor ability of other methylated xanthines like 7-MX to pass the blood-brain/blood-retina barrier (Hung et al., 2018; Shi and Daly, 1999) would suggest that the target tissue of these systemic treatments is not likely to be the retina. Neither AdoA_2A gene expression (in tree shrew) nor protein levels (in guinea pigs) are affected in choroid or sclera (Cui et al., 2010; He et al., 2014). However, scleral cAMP, known to influence collagen remodeling, has been demonstrated to be influenced by visual cues (Tao et al., 2013), and pharmacologically activating AC led to myopic refractive errors, and inhibiting AC led to attenuation of myopigenesis in guinea pigs (Tao et al., 2013). Yet, it is not clear if altered collagen metabolism is critical to myopigenesis, or if it occurs in response to the elongated eye, since significant myopia can develop in tree shrews with little change in collagen remodeling (McBrien et al., 2001).

3.3.c. Influence on ocular tissues

7-MX treatment has been demonstrated to reduce the progression of myopia in children (Trier et al., 2008) and in rabbits, chickens, guinea pigs, and macaques subjected to myopigenic visual stimuli (Cui et al., 2011; Hung et al., 2018; Nie et al., 2012; Wang et al., 2014). Despite these findings, oral 7-MX had only a minor protective effect with lens defocus and no effect on form-deprivation in chickens and tree shrews (Khanal et al., 2020; Liu et al., 2020; Wang et al., 2014) (Table 3). Together, these data may indicate significant species differences in myopigenic signaling and/or bioavailability of the 7-MX. Additionally, AdoR inhibition often only leads to partial protection against myopia development, which may suggest Ado influences myopia development either through a separate parallel and opposing pathway to myopigenic signaling or a pathway that contributes to retinoscleral signaling. In contrast to the pharmacological findings, transgenic mice lacking AdoA_2A developed relative myopia, both in terms of refractive error and axial length, compared to littermate wildtype controls (Zhou et al., 2010).

3.3.d. Summary

Ado is a highly dynamic and ubiquitous type of signaling with clear connections to scleral remodeling. While present in all tissues of the posterior eye wall, there is little evidence suggesting it connects any two layers and relatively little evidence showing altered signaling with myopigenesis, both in animal and human GWAS studies. However, thus far scleral Ado has not been ruled out as integral to myopigenesis, if influenced by another pathway capable of transmitting information through the choroid.

3.4. Other signals

It is known that retinoscleral signaling involved in emmetropization and myopigenesis involves many distinct signaling molecules and pathways. In addition to the signaling molecule candidates featured above, a few other signaling pathways are of particular interest due to their bidirectional modulation with visual stimulus (retinal glucagon) and successful clinical use as a pharmacological treatment for slowing myopia (atropine, a nonselective metabotropic muscarinic acetylcholine receptor antagonist). For in depth review of these and other chemical signaling molecules, see the recent review (Troilo et al., 2019).

Beyond these commonly studied signaling molecules, the thickness of the choroid has long been hypothesized to be involved in directing eye growth. It is bidirectionally modulated by visual cues, thinning with myopigenic cues and thickening in response to positive lenses or recovery from myopia. The consequences of these changes have been hypothesized to influence the scleral oxygen content and thus potentially function as a signal (Wu et al., 2018).

3.4.1. Hypoxia-inducible factor 1-alpha (Hif-1α)

It is well established that the choroid rapidly thins in response to myopigenic cues. It has been proposed that this thinning may be the result of decreased blood flow, and thus may decrease the oxygenation of the sclera (Liu et al., 2021). Retinoscleral signaling via oxygenation appears to be potentially viable, requiring that: 1) visual cues presented to the retina are able to influence the choroid, 2) decreased blood flow in the choroid underlies the choroidal thinning and results in graded degrees of scleral hypoxia, and 3) hypoxia in the sclera affects remodeling processes. Additionally, if oxygenation of the sclera explains myopigenesis, choroidal thickening should either elicit an opposite response in the sclera, or an additional pathway is required to signal slowing of axial elongation.

In support of this hypothesis, recent studies have found scleral Hif-1α to be correlated to myopigenesis. Hif-1α protein levels in the sclera, but not the retina, were associated with FDM in mice and guinea pigs and LIM in guinea pigs (Pan et al., 2021; Wu et al., 2018). Additionally, analysis of human gene databases revealed a moderate association with scleral Hif-1α signaling pathway in individuals with high myopia (Wu et al., 2018; Zhao et al., 2020). However, in tree shrews, scleral Hif-1α mRNA expression was not altered with form-deprivation, lens defocus, or recovery from lens-defocus (Guo et al., 2019, 2013). And in contrast to all, one study reported scleral Hif-1α mRNA expression decreased with lens defocus in guinea pigs (Guo et al., 2013).

Furthermore, experimentally manipulating hypoxia signaling appears to influence refractive state. Simultaneous treatment of guinea pigs with lens defocus and anti-hypoxia drugs resulted in reduced Hif-1α levels and partially suppressed myopic refractive error development and axial elongation (Wu et al., 2018). Additionally, two treatments used to experimentally reduce choroidal blood perfusion (and presumably decrease oxygenation of the sclera) were found to also induce myopia in guinea pigs (Zhou et al., 2021). These studies suggest a relationship between hypoxia, Hif-1α and development of myopia; however, the strength of this connection and whether scleral hypoxia is solely responsible for trans-choroidal propagation of myopigenic signaling is not clear. Additionally, whether the signal can encode both signs of defocus is not clear; the choroid has been shown to thicken with positive defocus and recovery from myopia, but how this translates to oxygen diffusion to the sclera and Hif-1α is not known.

4. Pathway crosstalk

Each of the primary signaling molecules discussed in the previous section are implicated to some degree in myopigenic retinoscleral signaling (Figure 2). Currently, the only retina-derived signal with any evidence to reach and act on the sclera is atRA, and more evidence suggests atRA acting on the sclera may be derived from the choroid. Thus, DA and other retinal signaling molecules known to affect myopigenesis acting anterior to the choroid, suggest that the choroid is a likely point of crosstalk between various pathways. Two mechanisms by which information can traverse the choroid have been presented: 1) changes to choroidal thickness/blood flow 2) a RA-based signal propagated across the choroid, possibly involving RBPs.

Many visual cues and retinal signaling molecules are known to influence choroidal thickness. Retina-originating signals (DA, Ado, atRA, glucagon, insulin, vasoactive intestinal peptide, monamines) could interact in the retina. While the genetic evidence for DA and Ado in human cohorts is limited, one of the most robust and repeated GWAS findings are the associations between refractive error and the gap junction delta-2 protein (GJD2) gene (Cheng et al., 2013; Fan et al., 2016; Hysi et al., 2020; Miyake et al., 2015; Solouki et al., 2010; Tedja et al., 2018; Verhoeven et al., 2013), which encodes connexin-36. As connexin-36 is dephosphorylated by DA and phosphorylated by Ado (Li et al., 2013), connexin-36 may be a conduit for these neuromodulators to affect eye growth. While not yet evaluated in the context of myopia, DA has also been found to act as an epigenetic histone marker in a process termed “dopaminylation” (Lepack et al., 2020), which may provide a future explanation linking the influence of the visual environment on genetic contribution to eye growth.

It is likely that many retina-originating signals are relayed by the RPE, due to the presence of many receptor types. With many distinct myopigenic cues similarly affecting choroidal thickness, and choroidal thickness possibly influenced by the RPE, many distinct retinal signals could be summed and influence choroidal thickness in a graded manner at the RPE. If choroidal thickness does in fact influence the oxygen environment of the sclera, further study into the role of the RPE on choroidal thickness could help elucidate these connections.

apoA-1 expression is known to be at least partially regulated by peroxisome proliferator-activated receptors (PPARs) (Gervois et al., 2000), which may be downregulated under hypoxic conditions. Intravitreal injections of the PPARα agonist GW6747, upregulated retinal and scleral apoA-I expression and suppressed experimental myopia in chickens (Bertrand et al., 2006). This increased protein expression of apoA-1 in the sclera could be from the choroid, since the sclera has not been shown to synthesize apoA-1 (Summers et al., 2016). However, GW6747 had no effect on choroidal apoA-1 mRNA expression (Summers et al., 2016).

Evidence suggests that hypoxic conditions may inhibit PPAR subtypes, possibly linking the effects of hypoxia to apoA-1 and atRA. When guinea pigs were treated with PPAR α/γ antagonists, Hif-1α increased and the animals developed myopic refractive errors and axial elongation (Pan et al., 2018). The opposite effect was found for animals treated with agonists (Pan et al., 2018). This bidirectionality of the influence of PPAR would further support the possibility of hypoxia as a cue for signaling increased/decreased axial elongation. However, PPAR activation has not yet been demonstrated to occur with choroidal thickening/hyperoxia or recovery from myopia. It is also not clear if scleral hypoxia and PPAR expression would influence apoA-1 that has presumably been transported from the choroid.

5. Conclusions and Future Directions

All of the major signaling pathways identified in this paper contribute in some way to myopigenesis and retinoscleral signaling. While DAergic signaling appears to be primarily restricted to the retina and RPE, the consequences of the signaling are known to influence both choroidal thickness and axial length. Ado is ubiquitous through the posterior eye wall and influences scleral remodeling; however, whether it is involved in signaling between tissues is not clear. RA is a promising candidate, ubiquitous to all posterior eye wall tissues, bidirectionally modulated with positive/negative defocus and has a plausible hypothesized transport mechanism by which it could cross the choroid. However, much of the work surrounding this molecule has been performed in the chicken model, which displays significant differences in scleral composition compared to mammalian species and necessitates further study in mammalian models. While many pathways play a role in myopigenesis, signaling across the choroid appears to be a bottleneck before which crosstalk likely occurs. Identified here are two main mechanisms by which information originating in the retina may reach the sclera: via an atRA-driven mechanism and/or oxygenation of the sclera. Future studies should be performed in mammalian models in order to better understand which findings are general mechanisms of myopigenesis. Additionally, directing research effort towards the use of inducible knockout mouse models could help to elucidate additional details of some of these difficult to study signaling mechanisms, such as: the role of adenosine in the retina, how dopamine influences the RPE, whether myopigenesis can occur in apoA-1 deficient animals, and how the various pathways influence choroidal apoA-1.

Abbreviations/acronyms

GWAS

Genome-wide association study

D

Diopters

RPE

Retinal Pigement Epithelium

LIM

Lens induced myopia

FDM

Form-deprivation myopia

DA

Dopamine

DOPAC

3,4-dihydroxyphenylacetic acid

DAergic

dopaminergic

DAC

Dopaminergic amacrine cell

cAMP

cyclic adenosine monophosphate

mRGC

melanopsin retinal ganglion cell

CREAM

Consortium for Refractive Error and Myopia

SNP

single nucleotide polymorphism

DRD1

Dopamine Receptor 1 gene

RA

Retinoic Acid

atRA

all-trans retinoic acid

ADH

alcohol dehydrogenase

RALDH

Retinaldehyde Dehydrogenase

RAR

retinoic acid receptor

RXR

Retinoid-X receptor

RBP

retinoid binding protein

apoA-1

apolipoprotein A-1

RDH5

retinol dehydrogenase 5

RGR

retinal G protein-coupled receptor

RORB

receptor-related orphan receptor beta

Ado

Adenosine

AdoR

Adenosine receptor

AC

adenyl cyclase

NMDA

N-methyl-D-aspartate

ENT

equilibrative nucleoside transporters

7-Mx

7-methylxanthine

Hif-1α

Hypoxia-inducible factor 1-alpha

GJD2

gap junction delta-2 protein