Neural and Müller glial adaptation of the retina to photoreceptor degeneration

Abstract

The majority of inherited retinal degenerative diseases and dry age-related macular degeneration are characterized by decay of the outer retina and photoreceptors, which leads to progressive loss of vision. The inner retina, including second- and third-order retinal neurons, also shows aberrant structural changes at all stages of degeneration. Müller glia, the major glial cells maintain retinal homeostasis, activating and rearranging immediately in response to photoreceptor stress. These phenomena are collectively known as retinal remodeling and are anatomically well described, but their impact on visual function is less well characterized. Retinal remodeling has traditionally been considered a detrimental chain of events that decreases visual function. However, emerging evidence from functional assays suggests that remodeling could also be a part of a survival mechanism wherein the inner retina responds plastically to outer retinal degeneration. The visual system´s first synapses between the photoreceptors and bipolar cells undergo rewiring and functionally compensate to maintain normal signal output to the brain. Distinct classes of retinal ganglion cells remain even after the massive loss of photoreceptors. Müller glia possess the regenerative potential for retinal recovery and possibly exert adaptive transcriptional changes in response to neuronal loss. These types of homeostatic changes could potentially explain the well-maintained visual function observed in patients with inherited retinal degenerative diseases who display prominent anatomic retinal pathology. This review will focus on our current understanding of retinal neuronal and Müller glial adaptation for the potential preservation of retinal activity during photoreceptor degeneration. Targeting retinal self-compensatory responses could help generate universal strategies to delay sensory disease progression.

Introduction

The immature retina undergoes dynamic changes in synaptic activity and connectivity during early development. The opening of the eye stimulates the maturation of retinal ganglion cells (RGCs). Light deprivation affects the stratification of RGC dendrites and results in permanent damage to visual responses (Tian and Copenhagen, 2001, 2003; Di Marco et al., 2009; Strettoi et al., 2022). In addition, mice with a loss of rod bipolar cells (RBCs) early in development were found to adjust synaptogenesis to preserve retinal function in dim light (Johnson et al., 2017). Similarly, the loss of a major cone bipolar cell (CBC) type can be fully compensated by other types of CBCs such that visual contrast and temporal frequency tuning is fully preserved by adulthood (Tien et al., 2017). The role of homeostatic plasticity in maintaining retinal function during development is further evidenced by the formation of synaptic connections between RBCs and cones in mice with neural retinal leucine zipper deficiency (Strettoi et al., 2004). This hypothesis is further supported by the preservation of retinal circuitry in mice with disproportionately large retinal neural cells induced by overexpressed antiapoptotic gene Bcl-2 (Strettoi and Volpini, 2002). However, it is unclear if matured retinas demonstrate plasticity to maintain retinal function in degenerative diseases.

Photoreceptor degeneration causes visual dysfunction and ultimately blindness in retinal disorders across all age groups. Synaptic connectivity in the inner retina and Müller glia processes undergo robust structural changes in response to photoreceptor loss. However, it is poorly known how these structural changes manifest in retinal function and vision. Recent studies suggest that mammalian retinas possess the capability of undergoing adaptive changes to preserve neural function during photoreceptor degeneration in a mouse model of retinitis pigmentosa (RP) (Leinonen et al., 2020; Fu et al., 2021b; Tomita et al., 2021), a rare genetic disorder characterized by photoreceptor loss (> 150 causal genes identified so far) (Newton and Megaw, 2020). This phenomenon corresponds to clinical observations that normal vision can be maintained for years in patients with inherited retinal degenerative diseases. Current ophthalmological technologies allow minimally invasive, longitudinal investigation of retinal function and morphology. Electroretinography (ERG) precisely records the electrical activation of the retina, producing data with signals corresponding to specific cellular subtypes. The eye´s transparency allows unrestricted examination of retinal anatomy using optical coherence tomography and fundoscopy. The retina is also a very practical target tissue for molecular biology and neuroscience research techniques (Figure 1). These aspects contribute to the retina being one of the best characterized neuronal systems in our bodies. Deciphering how the retina adjusts to photoreceptor defects could enhance our understanding of sensory diseases in general. In this review, we focus on the current knowledge of the impacts of retinal plasticity on neuronal function. Literature was searched between March and June 2022 using keywords “Retinal AND degeneration AND vision AND remodeling OR plasticity” and “Müller AND glial AND reprogramming OR remodeling”, respectively.

Figure 1.

Schematics of retinal structure and established technologies for retina research.

(A) The transparency of the eye allows direct visual access to the retina, which is the only neural tissue that is exposed to visual inspection noninvasively. (B) The layered and strictly hierarchical architecture of the retina is arguably the best characterized neural system in our bodies. (C) Electroretinogram (ERG) is a minimally invasive method to inspect sensory neuron (photoreceptor) activation (a-wave) and synaptic transmission to interneurons (b-wave). (D) Optical coherence tomography (OCT) allows segmentation of the retina noninvasively. (E) Scanning laser ophthalmoscopy (SLO) allows high-resolution images of vasculature. (F) Several advanced retina imaging methods using fluorescence, including 2-photon excitation, are established. (G) Dissected and perfused live retina is highly amenable to electrophysiology by e.g. multielectrode arrays (MEA) and patch-clamp, and (H) to assessment of metabolism. (I) Retina’s distinctly layered anatomy renders histological inspection and immunohistochemistry very practical. RGCL: Retinal ganglion cell layer; IPL: inner plexiform layer; INL: inner nuclear layer; OPL: outer plexiform layer; ONL: outer nuclear layer. (J) Single-cell transcriptomics maps gene profile at specific cells. (K) Established methods exist for preparation of primary cell cultures of several main cell classes in the retina. Graphs A, B, E, F, G, H, and K were adapted/reprinted from “Eye and Retina schematics”, “Diabetic Retinopathy Hallmarks”, “Rodent Fundoscopy”, “Retina MEA”, “Retinal ganglion cell patch clamp”, “Metabolic Assays -Using Seahorse Analyzers” and “Primary Cell Culture Preparation” with BioRender.com (2022). Retrieved from https://app.biorender.com/biorender-templates (license #: ZR241JTGY4, BU241JTVMS, NW241JU576, AQ241JUC41, LR241JUJWJ, JB241JVDIA, RD241JUX5O). Graphs C, D, I, and J present unpublished data from the Leinonen Retina Laboratory.

Retinal Remodeling and Rewiring and Functional Consequences during Retinal Degeneration

RP, primarily characterized by rod cell death, loss of night vision, and visual field reduction, accounts for roughly half of inherited retinal degenerative diseases. Clinically prominent pathology in the central retina can be delayed by several years when secondary cone photoreceptor death ensues (Hartong et al., 2006). In the early stages of RP, the retinal neural connectivity may undergo remodeling in conjunction with progressive rod degeneration (Cuenca et al., 2014; D’Orazi et al., 2014; Pfeiffer et al., 2020; Strettoi et al., 2022).

Retinal remodeling is characterized into Phases 0 to 3 (Pfeiffer et al., 2020). Phase 0 represents the healthy state of the retina. Phase 1 consists of initial photoreceptor stress, rod degeneration, glial activation, as well as emerging glial and neural remodeling characterized by sprouting of rod, RBC, and horizontal cell (HC) processes. Specifically, rod axons begin to sprout beyond the outer plexiform layer (OPL) into the inner retina and can reach the inner limiting membrane. RBCs and rod-contacting HCs retract their dendrites from rods and their dendrites may find new contacts with cone pedicles. The ON pathway, which activates at light onset, is particularly susceptible to remodeling, whereas the OFF pathway, responsible for light decrements, appears less susceptible to these changes. Concurrent with dendritic retraction from the OPL, metabotropic glutamate receptor-6 is downregulated in ON-RBC dendrites and can mislocalize to the RBC soma and axons (Gargini et al., 2007; Barhoum et al., 2008). A large proportion of RBCs and ON-CBCs can become “OFF-BC-like” as they start to aberrantly express ionotropic glutamate receptors during progressive photoreceptor degeneration (Marc et al., 2007; Jones et al., 2011). In Phase 2, cone death occurs. BCs become completely deafferented. HCs can extend dendrites backwards into the inner plexiform layer, and even amacrine cells and RGCs initiate sprouting (Strettoi and Pignatelli, 2000; Strettoi et al., 2002; Jones et al., 2003; Pfeiffer et al., 2020). Phase 3 represents the most advanced stage when all photoreceptors have died, leading to neurite outgrowth by all retinal neuron classes and global retinal cell death.

Unlike anatomical descriptions of remodeling, the functionality of the rewired retina, particularly in early disease stages, is less well understood. Emerging evidence suggests that early remodeling could be a manifestation of neuroplastic events (homeostatic plasticity) trying to retain normal light responses and visual function. Adaptational changes in retinal function in response to partial photoreceptor degeneration have been observed in multiple disease paradigms such as the light-induced retinal degeneration (LIRD) model (Richards et al., 2006; Montalban-Soler et al., 2012; Rubin et al., 2022), the photoreceptor ablation model (Care et al., 2019, 2020), and the progressive retinal degeneration model (Leinonen et al., 2020).

Retina can adapt to partial photoreceptor loss

Inner retina function recovers after partial light-induced retinal degeneration

Increasing evidence of inner neural retinal plasticity in mature retinas has been suggested in the past two decades. When C57BL/6J and Balb/C cross-bred mice are subjected to a continuous 20-day bright white light regimen (150–175 lux) to induce partial LIRD, there is a 50% loss in both the outer nuclear layer (ONL) thickness and scotopic ERG a-wave amplitude (Richards et al., 2006). Interestingly, the b-wave amplitude recovers to a higher 60% level. The negative polarity ERG a-wave represents the sum activation of photoreceptors, which is primarily rod-driven in scotopic conditions (Penn and Hagins, 1969). The positive b-wave occurs when the signal is transmitted from the photoreceptors to bipolar cells (Stockton and Slaughter, 1989). In theory, if diminished photoreceptor function directly propagates to inner retinal function, both ERG a- and b-wave amplitudes should decline by the same degree. Furthermore, the amplitude of fast frequency oscillatory potentials, which are believed to be mainly generated at the level of amacrine and ganglion cell interactions (Dong et al., 2004), initially drops significantly at 0–60 days post LIRD, but recovers to near normal levels by 90 days post LIRD (Richards et al., 2006). These observations indicate functional adaptation in the inner retinal circuitry during long-term LIRD. Similar observations are also reported in a 24-hour LIRD paradigm with a brighter light regimen (3000 lux) in Balb/C mice, in which a permanent 30% loss of a-wave amplitude is reported. However, these mice also display b-waves that gradually recovered to near normal levels by 7 days post LIRD (Montalban-Soler et al., 2012). The scotopic threshold responses representing ganglion cell function (Saszik et al., 2002) recovered completely (Montalban-Soler et al., 2012). In Sprague-Dawley albino rats subjected to partial LIRD, the ERG b-wave also recovered to over 80% of its original amplitude (Rubin et al., 2022). Moreover, inner neural retinal adaption may occur in large animals subjected to LIRD. In LIRD-induced RHO^T4R mutant dogs, there is a light-dose-dependent thinning of ONL (Iwabe et al., 2016). The visual behavior of these dogs at scotopic, mesopic, or photopic testing conditions (ranging from 0.003 to 65 lux) is normal as measured at 2 and 33 weeks post LIRD in an obstacle-avoidance behavioral vision testing. Taken together, following an acute phase of suppressed inner retina function after partial light damage, the downstream retinal circuitry seems to recover despite permanent sensory cell dysfunction.

Synaptic connectivity plasticity and cell migration restoration after controlled photoreceptor injury

In addition to LIRD, laser photocoagulation has also been used to injure the outer retina structures while leaving the inner retina without directly applied damage (Sher et al., 2013; Beier et al., 2017). In rabbit retinas with laser-induced injury specifically in RPE/photoreceptors, the lesion site heals as evidenced by the migration of adjacent photoreceptors to occupy the lesion site. The migrating photoreceptors establish new functional synaptic connections with local bipolar cells. The visual sensitivity of RGCs at the lesion site recovers to baseline levels after 2 months (Sher et al., 2013). Importantly, typical remodeling events include deafferentation of RBCs, followed by sprouting and formation of ectopic synapses in the ONL, which typically occur immediately following injury (Beier et al., 2017). However, as healthy photoreceptors continuously migrated into the lesion site, RBCs abandon aberrant connections to favor properly formed local synapses at the OPL. In the retinas of ground squirrels subjected to photocoagulation (Beier et al., 2018), there is a remarkable capability of the adult retina for selective circuit repair, as expanded short-wavelength sensitive cone (S-cone) connecting CBCs dendritic trees bypass the closely related medium-wavelength sensitive cones to selectivity synapse with their correct target, the S-cones. The adult retina’s propensity for synaptic plasticity is further corroborated in the Cngb1^neo/neo mouse RP model, which lacks a crucial CNGB1 protein for phototransduction (Wang et al., 2019). CNGB1 defects lead to diminished ERG a- and b-wave amplitudes, attenuated b-wave sensitivity, as well as severely mislocalized and malformed synaptic contacts between rods and RBCs. When CNGB1 expression is recovered in young adult Cngb1^neo/neo mice, photoreceptor degeneration is halted, rod-RBC synapse organization is corrected, and light sensitivity at RBCs and RGCs is significantly recovered, highlighting the capability of the adult retina for restorative synaptic plasticity.

The inner retina functionally compensates for partial photoreceptor loss

To further investigate how the adult retina functionally responds to partial rod or cone death, a selective cell ablation model using the cre-recombinase mediated diphtheria toxin receptor insertion technique has been applied (Care et al., 2019, 2020). Following the loss of half of the rod population in otherwise healthy adult mice, three possible functional outcomes have been hypothesized: (1) input loss from rods could directly propagate through the circuit leading to a matching defect in retinal output; (2) functional defects could be exacerbated in downstream circuitry leading to an even larger decrease in retinal output as predicted from magnitude of rod loss; or, (3) input loss could be compensated in downstream circuitry so that retina output is stronger than expected from the extent of rod loss (Care et al., 2020). Full-field ERGs and patch-clamp recordings at RBCs and RGCs have shown that hypothesis 3 is most likely. First, mean spiking activity recorded at A_ON-S RGCs to flashes of light is reduced by only 20%, although the average loss of rods in the experiment is 60%. Furthermore, A_ON-S RGC sensitivity to dim light flashes appears to increase compared to the values of normal controls. Secondly, scotopic ERG b-wave amplitude is much less reduced than the a-wave amplitude (Care et al., 2020), similar to what has been shown after partial rod death due to LIRD (Richards et al., 2006; Montalban-Soler et al., 2012; Rubin et al., 2022). To elucidate the locus of the apparent functional compensation, Care et al. (2020) have also patched and recorded individual RBCs. Voltage responses from RBCs remain intact despite marked rod loss, which further confirms compensation occurring at the rod-RBC synapses. Compensatory changes and maintained visual function are also observed after the loss of half of the cone population in adult mice (Care et al., 2019; Lee et al., 2022). These latter reports are challenged by the findings that cone pathway function recovers to normal levels when half of the cones are ablated in juvenile mice at postnatal day (P)10, but not when the same procedure is performed in matured mice retinas at P30 (Shen et al., 2020). This highlights the well-known dogma of the lower capability for neuroplasticity in adulthood compared to juvenility, which is also recapitulated in retinal tissue. Therefore, while retinal remodeling is capable of adjusting to retinal degeneration in adult mice after postnatal development, the extent of this plasticity is variable.

High inner retina activation and visual responses can coincide with progressive retinal degeneration

The hypothesis of whether functional compensation in the retina could occur during progressive retinal degeneration has also been tested (Leinonen et al., 2020). ONL thickness and rod-driven ERG responses are lost by around 75% at 5 months of age in heterozygous P23H mice (Figure 2; Leinonen et al., 2020). Despite the robust ongoing rod degenerative disease occurring alongside global retinal inflammation and oxidative stress (Leinonen et al., 2020; Fu et al., 2021b; Ortega et al., 2022), visual contrast sensitivity remains well-maintained up until 5 months of age (Leinonen et al., 2020). This is true even in P23H mice bred on a Gnat2^–/– background that eliminates cone phototransduction and renders the mice reliant on their rod-function. Mass RBC responses derived from ex vivo ERG recordings showed increased relative sensitivity (Leinonen et al., 2020), which mirrors prior findings (Care et al., 2020). The functional compensatory changes in P23H mouse retinas are accompanied by transcriptomic profile shifts that indicate robust cellular restructuring and neural plasticity with “cell adhesion molecules”, “axon guidance” and “glutamatergic synapse” being among the top 10 most upregulated KEGG pathways. Importantly, the decorrelation of scotopic ERG a- and b-waves, i.e. increased amplitude ratio of b to a, has also been observed in P23H rats (Machida et al., 2000; Aleman et al., 2001). However, visual performance in RP can decay nonlinearly as exemplified by abrupt loss of RBC responses and contrast sensitivity between 5 and 6 months in GNAT2^–/–/P23H mice (Figure 2; Leinonen et al., 2020). This needs to be taken into account when designing therapeutic interventions.

Figure 2.

Near normal rod-dependent visual contrast sensitivity remains for several months in retinitis pigmentosa mice.

(A) Representative optical coherence tomography (OCT) images showing degeneration progression in P23H mice. (B) Mean outer nuclear layer (ONL, photoreceptor nuclei layer) thickness as measured at 500 µm from the optic nerve head at dorsal, ventral, nasal, and temporal retinal quadrants. (C) Representative ex vivo ERG responses for dim flash (17 photons/µm²) in GNAT2^–/– control and GNAT2^–/–/P23H retinitis pigmentosa mice at baseline and after DL-AP4 perfusion to isolate rod-specific response. (D) Rod bipolar cells (RBC) and rod response amplitudes. RBC response was acquired offline by digitally subtracting the rod-specific response from the basal response. (E) RBC/rod response ratio. (F, G) Visual contrast thresholds as measured by optomotor responses (OMR) in 2–3-month-old (F) and 5–6-month-old (G) mice. Note the abrupt drop in ERGs and OMR performance between 5 and 6 months of age in GNAT2^–/–/P23H mice. Adapted from Leinonen et al. (2020).

Behavioral vision in a genetic retinal degenerative diseases model has also been tested closer to the terminal disease stage. Peripherin-2 deficient Rds mice never generate fully formed photoreceptor outer segments and express only approximately 3% of normal opsin content, which should lead to a similar extent of photon catch loss. However, ERG b-wave amplitudes are recorded at approximately 15% of normal levels, suggesting compensatory signal gain between photoreceptor and bipolar cells in these mice (Thompson et al., 2014). Although optokinetic tracking and visual water task behavior are severely compromised in Rds mice as expected, they still exhibit distinct and meaningful pattern vision despite minimal rod and cone function. In addition, RBCs display normal sensitivity to local application of exogenous glutamate in patch-clamp experiments in 2-month-old Rd10 retinas (Barhoum et al., 2008). This finding is remarkable as 2-month-old Rd10 mouse retinas are completely devoid of rods and show notable metabotropic glutamate receptor-6 downregulation in RBCs.

RGCs remain relatively stable during progressive retinal degenerative diseases, but several aspects of high-fidelity vision decline

In heterozygous P23H rats, receptive field strength declines immediately upon ONL degeneration (Sekirnjak et al., 2009, 2011). The rod-driven light responses in RGCs as well as receptive field size start to decline relatively late, only after P200, and RGCs spontaneous firing peaks at the same time. In Royal College of Surgeons rats at a disease stage when nearly all rods have died but most cones remain, RGC receptive field distortion is primarily linked with anatomic holes in cone mosaic and cone structural changes, rather than inner retina remodeling (Yu et al., 2017). Importantly, all the functionally diverse RGCs characterized in the study persist after rod death. However, although direction-selective RGCs remain in Royal College of Surgeons rats, their direction tuning broadens and direction selectivity decreases.

Moreover, some electrophysiology studies focusing on the midbrain targeting the superior colliculus (SC), dorsal lateral geniculate nucleus, or primary visual cortex (V1) function have suggested that RGC’s operation mode during retinal degenerative disease directly propagates to these brain targets (Fransen et al., 2015; Procyk et al., 2019; Leinonen et al., 2022). Recordings from the P23H rat superior colliculus demonstrate the loss of light responsivity in ON cells that practically coincide with input loss (Fransen et al., 2015). In contrast, OFF responses become supernormal and progressively elevate above wild-type levels. Recordings of the dorsal lateral geniculate nucleus in 3–5-week-old Rd1 mice, at a disease stage with complete loss of rods and partial loss of cones, demonstrate diminished ON cell responses and slightly declined receptive field sizes (Procyk et al., 2019). However, cone contrast sensitivity remains intact at this relatively advanced disease stage. In 2-month-old S334ter-3 rats at a disease stage when most rods are dead but cones remain, even when responses to pattern stimuli recorded at V1 have declined at medium-to-high contrasts, the contrast sensitivity as measured by the C50 parameter shows no deterioration (Chen et al., 2020). Similarly, excellent pattern contrast sensitivity in V1 single-unit recordings in young adult P23H mice has been shown, although several other receptive field properties are altered compared to wild-type mice (Leinonen et al., 2022). In summary, RGC physiology inevitably changes during photoreceptor degenerative disease and functional consequences propagate into the visual areas of the brain. The adult human visual cortex may also be sufficiently plastic to adapt to altered visual inputs (Lunghi et al., 2019; Castaldi et al., 2020). Still, many aspects of visual function remain unexpectedly stable, with contrast sensitivity particularly remaining strikingly resistant to retinal degeneration in animal models.

Does hyperexcitability in the inner retina counteract beneficial effects of rewiring?

A major issue that could discount the functional benefits of the apparent retinal adaptation to photoreceptor loss is the simultaneously increasing spontaneous neural activity that can worsen neural signal-to-noise ratio and mask light-responses. Characteristic of the severely degenerated retina is spontaneous, rhythmic oscillatory waves which physiological meaning has not yet been fully characterized. Although these oscillatory waves are generally recorded at the level of RGCs, they are not intrinsic to RGCs, but rather originate presynaptically (Borowska et al., 2011). They appear at two distinct, dominant frequencies. The less characterized are slow, sub-3 Hz oscillations that are believed to originate from spontaneous Ca²⁺ transients in remnant cones as a direct consequence of synaptic remodeling and decreased negative feedback by HCs (Haq et al., 2014). The fast approximately 10 Hz oscillations arise from CBC-AII amacrine cell network activity, which is independent of input from the degenerating outer retina (Borowska et al., 2011) and is believed to be dependent on voltage-gated Na⁺ channels and gap junctional coupling (Trenholm et al., 2012). The increased spontaneous oscillations likely incur detrimental effects on vision as their pharmacologic suppression improves light responses and stimulation efficiency recorded at RGCs (Toychiev et al., 2013; Barrett et al., 2015; Gehlen et al., 2020). Moreover, inner retinal neurons can also increase spontaneous spiking (become hyperexcitable) during photoreceptor degeneration. This can occur early in disease progression and appears to propagate into the visual areas of the brain (Drager and Hubel, 1978; Leinonen et al., 2022). After investigation of RGC’s responses during RP progression in Rd1 and Rd10 mouse retinas, increased spontaneous RGC firing in parallel with decreased light responses was found (Telias et al., 2019). The light responses at RGCs are normal before degeneration onset in P14 Rd10 mice, but at P28, RGC’s spontaneous activity increases by 2–3-fold and light-responses decrease by half. The hyperactivity in Rd10 mouse RGCs is intrinsic as blocking all synaptic drive to RGCs does not decrease hyperactivity. Based on an extensive series of pharmacological and genetic experiments that either increased or decreased retinoic acid signaling, the researchers have shown causative evidence that increased retinoic acid signaling in the retina could explain RGC hyperexcitability during progressive retinal degenerative diseases (Telias et al., 2019) and after laser-induced local photoreceptor ablation (Denlinger et al., 2020). Suppression of retinoic acid signaling remarkably improves luminance-detection and pattern vision in Rd10 mice (Telias et al., 2019, 2022), revealing a novel drug target to improve vision in RP.

An intriguing hypothesis is that increased retinal electrical activity upon retinal degeneration is part of a survival mechanism, as loss of input is known to lead to neuronal death whereas increased activity enhances longevity (Corredor and Goldberg, 2009). Therefore, long-term effects of interventions aimed at either increasing or decreasing the electrical activity need to be carefully assessed. It will also be important for future studies to elucidate whether increased synaptic gain at the rod-RBC synapse and increased inner retina hyperexcitability in RP retinas share the same mechanistic origin, as it could have significant implications for prospective therapies.

Müller Glia-Derived Retina Adaptation during Photoreceptor Degeneration

Metabolic adaptation

Müller glial cells span all retinal layers, providing structural and trophic support for neighboring neurons. In response to injury, Müller glia are activated and releases antioxidants and neurotrophic factors to preserve retinal function from further damage at early stages (Bringmann et al., 2009). However, prolonged gliosis is detrimental as the ability of Müller glia to support retinal neurons becomes disrupted. Müller glia may also produce lactate, lipoproteins, and glutamine, which are then shuttled to photoreceptors for synapse formation and energy production (Fu et al., 2021a). In healthy retinas, Müller glia have a very homogenous metabolic signature, which becomes highly variable in the levels of glutamine, glutamate, taurine, and glutamine synthetase among neighboring Müller glia in the region of the degenerating retina (Pfeiffer et al., 2016).

Recent single-cell transcriptomics shows decreased expression of genes involved in metabolism in both rods and cones in P23H mice during early photoreceptor loss. These analyses also show increased expression of genes involved in mitochondrial localization and microtubule-based transfer in Müller glia (Tomita et al., 2021), suggesting possible transfer of mitochondria from Müller glia to photoreceptors or increased energy production in Müller glia. Although the mechanisms of mitochondrial release are unclear, brain astrocytes shuttle mitochondria to injured neurons after stroke (Hayakawa et al., 2016). Decreased photoreceptor metabolism accelerates photoreceptor loss; therefore, improving glucose supply and uptake could promote the photoreceptor survival in degenerating mouse retinas (Punzo et al., 2009; Ait-Ali et al., 2015). Therefore, the question arises of whether Müller glia can compensate for photoreceptor energy shortage during early retinal degeneration.

Our current knowledge of Müller glia metabolic fuel sources is still quite limited. Most of the current literature relies on in vitro studies, making it difficult to fully characterize Müller glial metabolic properties, as the normal cell interactions between Müller glia with their connecting cells is lost. Furthermore, Müller glial cells in vitro are positive for GFAP, which reflects gliosis and fails to capture their diverse pathophysiology in the retina (Fu et al., 2021a). A recent report has shown that loss of hexokinase 2 (catalyzes the first step in glycolysis) and phosphoglycerate dehydrogenase (an endogenous serine synthetic gene), but not pyruvate dehydrogenase E1 alpha 1 (catalyzes the conversion of pyruvate into acetyl-CoA) and lactate dehydrogenase A (catalyzes the interconversion of pyruvate and L-lactate) in Müller glia leads to photoreceptor degeneration and reduced retinal neuronal function (Shen et al., 2021), suggesting that Müller glia may use glucose to produce serine to supply photoreceptors. Consequently, photoreceptors convert glucose to lactate to fuel Müller glia (Kanow et al., 2017). The contribution of cellular metabolic changes to cell fate decisions warrants extensive investigation. Illustrating the importance of this concept, targeting metabolic modulation via nutrient supplementation is a feasible approach to treat retinal degeneration (Rowe et al., 2021).

Remodeling

Müller glial activation facilitates the adaptive retinal remodeling as the hypertrophied glial cell processes may guide neuronal migration and aberrant neuronal cell process sprouting to maintain visual function (Bringmann et al., 2009; Fu et al., 2021b). Following the loss of photoreceptors, glial scars are formed between the remnant neural retina and RPE/choroid to fill the retinal breaks. The initial purpose of glial scars is to protect the neuroretina from further damage. However, glial scars also account for the failure of tissue repair and therapeutic intervention.

In fibroblast growth factor 21 (FGF21)-treated P23H mice, not only is there improved retinal neural function, but there is also a notable increase in the expression of genes involved in axon and synapse development in Müller glia and amacrine cells identified with single-cell transcriptomics (Fu et al., 2021b), suggesting glial remodeling may preserve retinal function during photoreceptor degeneration. However, the preservation of retinal neuronal responses lasts for six weeks and diminishes at nine weeks of treatment. This could be due to the inhibitory impacts of FGF21 on RPE fatty acid oxidation, which may potentially cause metabolic switches in RPE and limit the fuel supply to photoreceptors. Further exploration of cellular interaction and metabolic alterations is needed. Better understanding of the molecular mechanisms may help prolong the beneficial effects of glial remodeling during photoreceptor loss.

In addition, Müller glia produce photoreceptor components such as 11-cis-retinol in the chicken retina and rhodopsin protein in mouse retina (Goel and Dhingra, 2012; Kaylor et al., 2014). In P23H mouse retinas, there is co-expression of rod markers (Rho and Pde6b), cone markers (Opn1mw and Opn1sw), and Müller glial cell markers (Rlbp1 and Slc3a1) in a cluster of cells identified with single-cell transcriptomics. Decreased expression of genes involved in phototransduction, inner and outer segment, photoreceptor cell cilium, and photoreceptor development are found in both rods and cones in P23H mouse retinas (Tomita et al., 2021). Müller glia may enhance the pathways involved in photoreceptor maintenance to compensate for those lost in rods and cones.

Collectively, Müller glia may support photoreceptor health by providing the necessary proteins lost during retinal degeneration, or that Müller glia may undergo neurogenesis to replace degenerated photoreceptors. Further validation at translational and function levels is needed to test this hypothesis. However, Müller glia reprogramming, proliferation, and differentiation are well documented in zebrafish, but may not occur spontaneously in mammalian retinas (Salman et al., 2021).

Reprogramming

Zebrafish possess the ability to regenerate a damaged retina and restore vision (Goldman, 2014; Lahne et al., 2020). In response to retinal injury, multiple extrinsic signaling pathways could trigger Müller glial cell cycle-reentry (Figure 3A). Growth factors and cytokines such as insulin, FGF2 (Wan et al., 2014), heparin-binding epidermal-like growth factor (HBEGF) (Wan et al., 2012), midkine-a (Nagashima et al., 2020), and tumor necrosis factor-alpha (Nelson et al., 2013; Iribarne et al., 2019) activate the regenerative potential of Müller glia. Signaling pathways such as glycogen synthase kinase 3β/β-catenin, Mapk/Erk, PI3K/Akt and Jak/signal transducer and activator of transcription 3 stimulates Müller glial reprogramming and retinal progenitor formation (Wan et al., 2014). The downstream transcriptional factor achaete-scute complex-like 1a (Ascl1a) is a key regulator of genes involved in Müller glia-derived progenitor production and zebrafish retina regeneration (Fausett et al., 2008; Ramachandran et al., 2010, 2011). Ascl1a positively regulates Müller glial insulinoma-associated 1a, which in turn induces the expression of the cyclin-cdk inhibitor p57^kip2 and suppresses cell-cycle genes (such as ccna2, cdk1, and cdk2), thus stimulating cell cycle exit and neural differentiation (Ramachandran et al., 2012). Insulinoma-associated 1a, via modulating the cone-rod homeobox (Crx, responsible for photoreceptor differentiation) and Nr2e3 (a photoreceptor-specific nuclear receptor responsible for rod photoreceptor specification), is required for photoreceptor differentiation in the developing zebrafish retinas (Forbes-Osborne et al., 2013). Meanwhile, suppression of the Notch signaling (Campbell et al., 2021) and transforming growth factor β (TGFβ) signaling (Lenkowski et al., 2013), which drive Müller glial differentiation and quiescence, promotes progenitor proliferation and retinal regeneration. Furthermore, inhibition of GABA or glutamate receptors also stimulates Müller glial proliferation in uninjured retinas (Rao et al., 2017; Kent et al., 2021). Although the role of these factors might be species-specific, some knowledge gained from zebrafish has promoted the study of induction of Müller glial reprogramming, proliferation, and differentiation in mammalian retinas.

Figure 3.

Müller glial responses in retinal degeneration.

(A) Müller glial regeneration in zebrafish. Upon injury, Müller glia undergo reprogramming and enter the cell cycle for proliferation, stimulated by β-catenin, Mapk/Erk, Jak/signal transducer and activator of transcription 3 (Stat3), PI3K/Akt. These signaling pathways further induce Ascl expression to generate Müller glial progenitor cells, which regenerate the injured retina. Growth factors and cytokines (such as FGF2, HBEGF, insulin, Midkine-a, TNFα) as well as Wnts positively stimulate the process. TGF-β, Notch, and neurotransmitter (GABA, glutamate) signaling negatively regulates the process. Adapted and updated from Salman et al. (2021). (B) In mammalian retina, the regeneration potential of Müller glia is limited. The activated Müller glial may undergo metabolic adaptation and remodeling to preserve neural retinal function. Interestingly, EGF and HBEGF stimulate while TGF-β suppresses Müller glial proliferation in mammalian retinas. Manipulation of β-catenin, Nfia/b/x, Hippo, and YAP pathways stimulates the generation of Müller glial progenitor cells, which can be further differentiated into photoreceptors upon the subsequent activation of Otx2, Crx, and Nrl following the stimulation of the Wnt pathway by targeting β-catenin. The graph was adapted/reprinted from “retinal cell (Müller glia)”, “cone photoreceptor”, “rod photoreceptor”, “retinal cell (ganglion 1)”, “bipolar neuron”, “retinal cell (amacrine)”, “retinal cell (horizontal)” by BioRender.com (2022). Retrieved from https://app.biorender.com/biorender-templates (license #:HT240EKBMT). Ascl: Achaete-scute complex-like 1; Crx: cone-rod homeobox; EGF: epidermal-like growth factor; FGF2: fibroblast growth factor 2; GABA: gamma-aminobutyric acid; HBEGF: heparin-binding epidermal-like growth factor; Insm1a: insulinoma-associated 1a; Jak/Stat3: Janus kinase-signal transducer and activator of transcription 3; Mapk/Erk: mitogen-activated protein kinase/extracellular signal-regulated kinase; Nfia/b/x: nuclear factor I; Nrl: neural retina leucine zipper; Otx2: orthodenticle homeobox 2; PI3K/Akt: phosphatidylinositol-3-kinase/protein kinase B; TGF-β: transforming growth factor-β; TNFα: tumor necrosis factor α; YAP: yes-associated protein.

Unlike those in zebrafish, the reactive Müller glial cells in mice will not proliferate and instead return towards quiescence (Figure 3B). However, Müller glia in mammalian retinas still hold regenerative potential. As shown in zebrafish, modulating growth factors such as EGF and TGFβ also controls Müller glial progenitor cell generation in mammalian retinas. In isolated mouse retinal explants, EGF treatment via activation of Erk1/2 and PI3K/Akt pathways induces Müller glial proliferation (Ueki and Reh, 2013). HBEGF pretreatment also stimulates Müller glial-derived progenitor cell formation in N-methyl-D-aspartate-damaged mouse retinas (Todd et al., 2015). TGFβ inhibitors further potentiate EGF-stimulated Müller glial proliferation in rat retinas in vivo (Close et al., 2005). A combination of factors including FGF2, insulin-like growth factor 1, taurine, retinoid acid induces photoreceptor differentiation of human Müller stem cells in vitro; however, TGFβ suppresses the process (Angbohang et al., 2016). Moreover, insulin treatment via increasing glucose use prolongs cone photoreceptor survival in mice with inherited retinal degeneration (Punzo et al., 2009). Intravitreal injection of both insulin and basic fibroblast growth factor induces Müller glial migration from the inner nuclear layer to photoreceptor layer in degenerating mouse retinas (Goel and Dhingra, 2021). However, migrated Müller glial cells in photoreceptor layer express stem cell marker Chx10 but do not further differentiate into rods and cones (Goel and Dhingra, 2021). These reports suggest distinct impacts of growth factors and neurotransmitters on Müller glial proliferation and retinal regeneration in mammalian retinas.

Targeting signaling pathways is an intriguing intervention to trigger Müller glial reprogramming in mammalian retinas. Overexpression of Ascl1 in Müller glia in young mice stimulates the regeneration of amacrine and bipolar cells, as well as photoreceptors after N-methyl-D-aspartate or light injury (Ueki et al., 2015). However, this regenerative potential is dampened in adult mice. Müller glia have been induced to differentiate into rod photoreceptors and restore vision in retinal degenerative mice by firstly stimulating Wnt signaling (β-catenin) in 4-week-old mice and then transcriptional factors essential for rod induction (Crx, Otx2, and Nrl) 2 weeks later (Yao et al., 2018). In addition, overactivation of YAP (yes-associated protein), which interacts with EGF receptor signaling, induces Müller glia to reprogram into highly proliferative cells in adult mouse retinas (Hamon et al., 2019). Alternatively, the deletion of Müller glial-specific Hippo pathway components, which suppress YAP signaling, also results in a highly proliferative, progenitor-like cellular state (Rueda et al., 2019). Loss of Müller glial nuclear factor I factors a, b, x (Nifa/b/x), which maintains glial quiescent status, leads to Ascl1 up-regulation, Müller glial reprogramming, and ultimately bipolar- and amacrine-like cell generation in adult mice after treatment with N-methyl-D-aspartate, FGF2 and insulin (Hoang et al., 2020). This process can be suppressed by YAP inhibition (Hoang et al., 2020). Targeting common signaling pathways may generate generic therapeutic potential for mammalian Müller glial reprogramming and retinal regeneration.

Interestingly, a recent report shows that the hybrid resulting from fusion of Müller glia with adult stem cells can integrate within human retinal organoid and differentiate towards a neural fate (Bonilla-Pons et al., 2022). Treatment of a combo of small molecules (dbcAMP, Forskolin, ISX9, CHIR99021, I-BET151, and Y-27632) in Müller glia in vitro drives the direct reprogramming into bipolar-like cells (Yang et al., 2022). Further tests in vivo are needed to confirm these findings. Taken together, these findings suggest that the stimulation of retinal regeneration in mammals is possible. However, the efficiency needs to be further improved in developed retinas. The long-term impacts also need to be carefully evaluated, because the balance between the differentiative state and proliferative state is interfered with, which could compromise the supportive role of Müller glia in retinal neural and vascular function.

Conclusion

With these results taken together, the mature mammalian retina possesses some degree of plasticity that may help maintain vision during retinal degenerative diseases. Regarding the therapeutic potential of retinal glial cells, activated Müller glia benefit inner retinal neurons via producing trophic factors (Vecino et al., 2016). Müller glia could also enhance the pathways affected in photoreceptors as well as potentially improve photoreceptor energy supply in degenerative retinas. Supplementation of natural metabolic modulators that induce Müller glial remodeling helps preserve neural retinal function (Fu et al., 2021b). Better understanding the interaction between Müller glia and retinal neurons, as well as the underlying mechanisms, would greatly accelerate the field. Further studies regarding how to improve neuroplasticity during blinding diseases are warranted. Investigations targeting the cellular and molecular contributions to the compensatory neural signals are key to better understanding the disease pathogenesis. Further advancement of proteomics and metabolomics at a single cell level will further validate the results obtained from single-cell transcriptomics. The field could also be further advanced by the application of spatial transcriptomics that map the cell gene profile at specific locations within the retina. The application of the microfluidics 3D system, which better models the cell-cell interactions in in vivo conditions, would benefit drug testing. Continuous efforts will be necessary to explore this field as it could lead to novel therapeutics that tackle the unmet medical need caused by retinal degeneration.

Additional file: Open peer review report 1 ^{(85.6KB, pdf)} .