The endoplasmic reticulum: Homeostasis and crosstalk in retinal health and disease

Abstract

The endoplasmic reticulum (ER) is the largest intracellular organelle carrying out a broad range of important cellular functions including protein biosynthesis, folding, and trafficking, lipid and sterol biosynthesis, carbohydrate metabolism, and calcium storage and gated release. In addition, the ER makes close contact with multiple intracellular organelles such as mitochondria and the plasma membrane to actively regulate the biogenesis, remodeling, and function of these organelles. Therefore, maintaining a homeostatic and functional ER is critical for the survival and function of cells. This vital process is implemented through well-orchestrated signaling pathways of the unfolded protein response (UPR). The UPR is activated when misfolded or unfolded proteins accumulate in the ER, a condition known as ER stress, and functions to restore ER homeostasis thus promoting cell survival. However, prolonged activation or dysregulation of the UPR can lead to cell death and other detrimental events such as inflammation and oxidative stress; these processes are implicated in the pathogenesis of many human diseases including retinal disorders. In this review manuscript, we discuss the unique features of the ER and ER stress signaling in the retina and retinal neurons and describe recent advances in the research to uncover the role of ER stress signaling in neurodegenerative retinal diseases including age-related macular degeneration, inherited retinal degeneration, achromatopsia and cone diseases, and diabetic retinopathy. In some chapters, we highlight the complex interactions between the ER and other intracellular organelles focusing on mitochondria and illustrate how ER stress signaling regulates common cellular stress pathways such as autophagy. We also touch upon the integrated stress response in retinal degeneration and diabetic retinopathy. Finally, we provide an update on the current development of pharmacological agents targeting the UPR response and discuss some unresolved questions and knowledge gaps to be addressed by future research.

1. Introduction

Seeing is believing. To the majority of people, vision is the foremost valued sense among others: auditory (hearing), olfactory (smell), tactile/haptic (touch), and gustatory (taste). Coincidently, vision is the most extensively studied sensory modality, in part due to the great complexity of the biological system responsible for generating and processing visual information in the eye and the brain (Hutmacher, 2019). Yet, many fundamental questions in the formation and regulation of visual signaling and the mechanisms of vision-threatening human diseases are far from being fully understood. The initial steps of the generation of vision are carried out in the retina, a thin sensory neural tissue located in the back of the eye. The retina consists of five major types of neurons, including photoreceptors, bipolar cells, amacrine cells, horizontal cells, and retinal ganglion cells (RGCs), which can be further divided into over 100 subtypes (West et al., 2022). Photoreceptor cells are the primary neuronal types that receive and convert light into an electrical signal, which is sequentially transmitted to the interneurons (e.g. bipolar cells) and RGCs. RGCs are the final output neurons of the retina, whose axons form the optic nerve that physically connects the eye to the brain and conveys the visual signals to the visual cortex for final processing. In vertebrates, the retina and the optic nerve originate as outgrowths of the developing brain (Amini et al., 2018). As such, the retina is considered an extension of the brain responsible for the generation of the visual signal and is a bona fide part of the central nervous system (CNS).

Apart from neurons, the retina contains large numbers of glial cells, including Müller cells, astrocytes, and microglia. These cells, like in other tissues of the CNS, have important functions including providing anatomical and functional support to the neural retina, producing neurotrophic factors and other cytokines, and closely interacting with retinal neurons in response to environmental stresses and injuries (Fletcher et al., 2008). In addition, the retina is supplied by dual vascular systems derived from the central retinal artery and choroidal blood vessels, which efficiently deliver nutrients and oxygen to the inner and outer retinal tissue, respectively. The outer retina, consisting of cell bodies, inner segments (IS), and outer segments (OS) of photoreceptors, is an avascular zone. Photoreceptor cells obtain glucose and oxygen diffused from the choriocapillaris through the retinal pigment epithelium (RPE), a single layer of hexagonal epithelial cells lying between photoreceptors and the choroid. RPE cells are essential for the maintenance of neural retinal structure and function through many multiple activities, including diurnal phagocytosis of the aged and damaged photoreceptor OS, participating in a visual cycle that is important for vision formation, and acting as an integral part of the ecosystem for retinal metabolism (Hurley, 2021). They also form the outer blood-retinal barrier (BRB), which, together with the inner BRB formed by retinal endothelial cells, provides a stable microenvironment for retinal neurons (Simó et al., 2010). Dysfunction and loss of the RPE can lead to photoreceptor cell death and degeneration and consequently, vision impairment.

Proteins are the fundamental building blocks of all complex organisms and are involved in almost every aspect of biological processes. Maintaining a healthy and homeostatic proteome is critical for neuronal survival and function. In addition, the retina has several unique features that require a higher level of protein synthesis and quality control than any other CNS counterparts. For example, photoreceptor cells, which are the most abundant retinal neurons, have a very high protein turnover rate to maintain the phototransduction machinery (Pearring et al., 2013). Other processes, including the formation and remodeling of neural synapses, production and release of neural transmitters, and regulation of signaling pathways to adapt to environmental and metabolic changes in the retina, all involve a large variety of proteins with distinct structures and functions. The highly complex nature of the retinal neuronal networks and the tightly controlled interactions between retinal neurons and their supporting systems – the RPE, glial cells, and vascular cells, further renders the retina susceptible to perturbations in protein homeostasis. Disruption of protein homeostasis results in increased ER stress and activation of the ER stress signaling (Kaufman, 1999). In this review, we will summarize the recent progress on the role of ER homeostasis in maintaining retinal neuronal survival and function and discuss the implication of ER stress signaling in the pathogenesis of a broad range of retinal diseases including inherited retinal degeneration (IRD), achromatopsia, age-related macular degeneration (AMD), and diabetic retinopathy (DR). We will highlight how the ER interacts with other intracellular organelles such as mitochondria and how ER stress signaling crosstalk with other stress response pathways such as autophagy. We will discuss the implication of integrated stress response (ISR), an important cellular response pathway activated by a variety of stress conditions targeting protein synthesis, in retinal diseases. Finally, we will review the current development of pharmacological agents targeting ER stress signaling and discuss the needs of future research to tackle some of the unresolved questions for improving our understanding of ER stress signaling in retinal health and disease.

2. Structure and function of the ER in the retina and retinal neurons

The endoplasmic reticulum (ER) is a large, membrane-bound organelle responsible for a diverse range of important functions including the biosynthesis, folding, quality control, and trafficking of membrane proteins and secretory proteins, lipid and steroid biosynthesis, and carbohydrate metabolism, and intracellular calcium storage and gated release (Schwarz and Blower, 2016). Although the ER was among many other intracellular organelles first identified by light microscopy in the late 19th century, the term of ER was not given until the 1950s following the discovery of the ER ultrastructure – a “lace-like” structure in the perinuclear region of the cytoplasm under electron microscope [reviewed in (Sree et al., 2021)]. As the largest endomembrane system, the ER comprises about 10% of the total cell volume with its membrane comprising about half of the total membrane in a eukaryotic cell (Voeltz et al., 2002). The structure and morphology of the ER are highly heterogeneous forming specific domains to carry out distinct functions. The flat sheet-like rough ER with the presence of ribosomes attached to the outer membrane that gives rise to a studded appearance is the major compartment for protein biosynthesis, protein folding, and post-translational modifications, such as disulfide bond formation and N-linked glycosylation. The smooth ER, on the other hand, is tubular and primarily responsible for sterol biosynthesis, lipid droplet formation, carbohydrate metabolism, detoxification, and intracellular calcium storage. The distribution and content of the ER and the ratio between the rough and smooth ER vary significantly across cell types associated with their specialized functions. For example, secretory cells such as pancreatic acinar cells possess abundant stacks of sheet-like rough ER while hepatocytes have a large smooth ER network responsible for carbohydrate/lipid metabolism and waste detoxification (Goyal and Blackstone, 2013). Recent research taking advantage of live-cell microscopy and in situ cryo-electron tomography identified specialized ER subdomains, namely ribosome-associated vesicles (RAV), a dynamic sub-compartment of the rough ER in cultured cells across various cell types including secretory cells, fibroblasts, endothelial cells, and neural cells, and in neurons of human brain tissue (Carter et al., 2020; Ning et al., 2023). The formation of RAV is believed to be associated with an increased demand of protein synthesis in secretory cells and local translation in the dendrites of developing neurons to meet the protein requirement for synaptic plasticity and remodeling (Carter et al., 2020). In addition, RAVs were found to directly contact with the mitochondria, suggesting that these newly identified rough ER sub-compartments may participate in the regulation of calcium signaling and mitochondrial function (Carter et al., 2020).

In neurons, the ER has unique morphological features to befit the structural and functional requirement to fulfill the neural activity (Sree et al., 2021). While the ER exists in all compartments of a neuron, the morphology and size of ER sheets and tubules are substantially different in the area of cell body or soma, dendrites, and axons. In neuronal soma, the ER consists of predominantly highly packed rough ER sheets, which were initially identified as Nissl bodies or Nissl substance (Palay and Palade 1955). The fission or fragmentation of the Nissl substance, also known as chromatolysis, which indicates the disruption of the protein machinery, can lead to apoptosis and demyelination in neurons with axonal injury and is observed in motor neurons in amyotrophic lateral sclerosis patients (Sree et al., 2021). In addition, the ER makes extensive contact with the plasma membrane in the neuronal soma and this contact is reduced following excitation (Sree et al., 2021). In dendrites, the proximal somato-dendritic regions are enriched in rough ER. The distal dendritic regions and dendritic spines, which are tiny protrusions that form functional contacts with the axons of neighboring neurons, contain predominantly tubular ER. Similarly, the ER in neuronal axons is mostly tubular with intermittent small cisternae in the synaptic varicosities. The narrow ER tubules run parallel along the axon and form physical contact with several organelles including the mitochondria and microtubules, possibly contributing to the transportation of organelles in the axon elongation (Fig. 1) (Khan, 2022; Sree et al., 2021). Mutations of genes that are involved in the regulation of axonal ER morphology and/or the dynamics of ER membrane contacts with other organelles can lead to dysregulated calcium homeostasis and mitochondrial dysfunction contributing to axonal degeneration in several human diseases including hereditary spastic paraplegia (HSP) (Öztürk et al., 2020).

Fig. 1.

ER localization in retinal neurons. A) Morphology and distribution of ER in the soma, axon, presynaptic terminals of a neuron. 1) Rough ER is distributed around the nuclear envelope in the soma and in somato-dendritic regions. 2) Smooth ER is localized predominantly to distal dendritic regions and axons. 3) The ER forms physical contacts with mitochondria, microtubules, endosomes, and lysosomes to support axon growth and organelle transportation. 4) In presynaptic terminals, the ER contributes to neurotransmitter production and regulation of calcium signaling. B) Distribution of ER in rod and cone photoreceptors.

In retinal neurons, the ER is most well characterized in the photoreceptors (Križaj, 2012; Mercurio and Holtzman, 1982). As discussed earlier, photoreceptors are highly compartmentalized cells consisting of elongated OS, connecting cilium, IS, cell body, and synaptic terminal. OS is a highly special structure consisting of tightly stacked membrane discs for phototransduction and does not contain an ER network. The IS enriched in rough ER sacs, where proteins, such as rhodopsin, are synthesized and transported through the connecting cilium to the OS discs (Mercurio and Holtzman, 1982). Photoreceptors also contain abundant smooth ER tubules, which continuously span the IS and cell body and extend into the synaptic terminal (Fig. 1). The smooth ER plays a critical role in the regulation of photoreceptor function, in particular controlling the amplitude, response speed, and sensitivity of photoreceptor signals in synaptic transmission (Križaj, 2012; Suryanarayanan and Slaughter, 2006). In addition, the ER in the synaptic terminal may also contribute to local protein synthesis for establishing and modulating the synaptic proteome and neurotransmitter production. Similar extensive distributions of the ER in other types of retinal neurons have been reported, although the specific features of the ER in individual neuronal cell types are sparsely studied. Using subretinal injection and electroporation of Emerald-Sec61β under control of the mGluR6 promoter that specifically labels the ER in ON-Bipolar cells, Agosto et al. demonstrated that the ER network is distributed throughout the cell body and extends into axons and dendrites, but not in the dendritic tips of bipolar cells (Agosto et al., 2018). In cultured Xenopus laevis RGCs, the axonal ER is undergoing dynamic remodeling through lysosome-driven ER tubule elongation that supports axonal growth (Lu et al., 2020). Disruption of ER-lysosome contacts leads to ER fragmentation and axon growth defects, suggesting a potentially important role of targeting ER dynamics in axonal regeneration for the treatment of neurodegenerative disease.

3. The unfolded protein response (UPR)

Given the important functions of the ER involved in governing protein, lipid, and carbohydrate biosynthesis and metabolism and intracellular calcium signaling, disturbance of the homeostatic status of the ER environment is detrimental to cell survival and function. Perturbation of ER homeostasis results in an accumulation of unfolded proteins or misfolded proteins in the ER lumen, a condition known as ER stress. ER stress can be induced by genetic factors that impair protein glyco-sylation or protein folding and environmental factors such as aging, oxidative stress or hypoxia, glucose deprivation or hyperglycemia, calcium dysregulation, and disturbed autophagy in retinal cells [reviewed in (Chen et al., 2023; McLaughlin et al., 2022) ]. To eliminate the misfolded or unfolded proteins, cells activate a sophisticated adaptive mechanism namely the unfolded protein response (UPR) via three trans-ER-membrane proteins: inositol-requiring protein-1 (IRE1), protein kinase RNA- (PKR-) like endoplasmic reticulum kinase (PERK), and activating transcription factor-6 (ATF6). The activation of these ER stress-sensing proteins appears to be controlled in a temporal manner eliciting a well-programmed cellular response to eliminate or adapt to ER stress and also allow the functional recovery of the ER. Failure of the UPR in restoring the ER homeostasis can lead to cell death and dysfunction contributing to the development of human diseases.

3.1. The IRE1 pathway

The IRE1 pathway is the most conserved UPR branch existing from yeast to humans. There are two isoforms of IRE 1 proteins, namely IRE1α and IRE1β, encoded by ERN1 and ERN2 in humans, respectively. IRE1α is the most abundant isoform expressed in all types of cells and tissues whereas the expression of IRE1β is restricted to epithelial cells in the gastrointestinal and respiratory tracts (Bertolotti et al., 2001; Iwawaki et al., 2009). Structurally, IRE1 consists of an N-terminal luminal domain, which acts as an ER stress sensor, a transmembrane domain, and a cytoplasmic domain that contains regions possessing protein kinase function and endoribonuclease function [reviewed in (Siwecka et al., 2021)]. A simplified process of activation of the IRE1/XBP1 pathway is illustrated in Fig. 2. In resting cells, IRE1 is kept inactive by binding to GRP78 which maintains IRE1 in a monomeric state. Upon ER stress, GRP78 dissociates from IRE1 and binds to misfolded proteins facilitating their refolding. The dissociation allows IRE1 dimerization or oligomerization in cells, which brings the cytoplasmic kinase domains in close proximity in a face-to-face orientation that stimulates autophosphorylation (Prischi et al., 2014). The autophosphorylation then triggers conformational changes that activate the endoribonuclease domain of the enzyme (Li et al., 2010). The activated endoribonuclease domain initiates an unconventional splicing process that converts the mRNA of unspliced X-box binding protein 1 (XBP1u) into the mRNA of spliced XBP1 (XBP1s) (Yoshida et al., 2001a). This process is so-called unconventional splicing because it occurs in the cytoplasm as opposed to the “conventional nuclear splicing” and does not require de novo synthesis of the protein to be spliced (Uemura et al., 2009). It is important to note that the activation of the RNase domain and the kinase domain can occur independently. Complete activation of the RNase function requires a dimerization/oligomerization-dependent intermolecular autophosphorylation (Siwecka et al., 2021). Preventing dimerization/oligomerization or phosphorylation can lead to an inhibition of IRE1-mediated UPR activation.

Fig. 2.

The IRE1/XBP1 signaling pathway. In resting cells, ER stress sensors including IRE1 bind to Bip/GRP78, which keeps them in inactive state. 1) Upon ER stress, Bip dissociates from IRE1 and binds to accumulated unfolded or misfolded proteins. 2) IRE1 is activated by dimerization and autophosphorylation. Increased kinase activity of IRE1 promotes JNK and IKK activation resulting in inflammation and apoptosis. 3–4) The endoribonucease domain of IRE1 is activated resulting in an unconventional splicing of XBP1 mRNA (3) and regulated IRE1-dependent decay of mRNA (RIDD) (4). The resulting spliced XBP1 (XBP1s) encodes an active transcription factor that upregulates ER chaperones and genes encoding ER-associated degradation (ERAD) proteins.

The splicing product, XBP1s, encodes a transcription factor that regulates a large array of UPR-related genes, including ER chaperones such as BiP, p58IPK, ERdj4, PDI-P5, and HEDJ, genes responsible for ER-associated degradation (ERAD) machinery such as HRD1, EDEM, Derlin-2, and Derlin-3, and genes encoding proteins involved in lipid synthesis and ER biogenesis (Lee et al., 2003; Sriburi et al., 2004). Ablation of XBP1 in pancreatic β cells, hepatocytes, and antibody-producing plasma cells results in a loss of their secretory functions, suggesting an essential role of XBP1 in secretory cells (Gass et al., 2004). The unspliced XBP1, on the other hand, is believed not to act as a transcription factor because it lacks a transcription activation domain (Iwakoshi et al., 2003). However, recent studies identified that XBP1u also possesses several important functions in regulation of the UPR, autophagy, and other cellular processes. For example, XBP1u negatively regulates the UPR by targeting XBP1s and activated ATF6 for proteasomal degradation (Chen et al., 2014b; Yoshida et al., 2009). It also binds to other proteins such as β-catenin and FoxO4 (Forkhead box protein O 4) in vascular cells and inhibits aneurysm formation and vascular calcification (Yang et al., 2022; Zhao et al., 2017). Interestingly, a recent study shows that expression of XBP1u, but not XBP1s, increases the Gal4-CREB reporter activity and rate-limiting gluconeogenic gene expression in cultured hepatocytes, suggesting a novel role of XBP1u in the regulation of gluconeogenesis in the liver (Peng et al., 2022a).

Although XBP1 mRNA is the foremost studied substrate of IRE’s RNase activity, IRE1 also takes advantage of this mechanism to remove selected mRNAs and micro-RNAs, a process known as regulated IRE1-dependent decay of mRNA (RIDD) [reviewed in (Maurel et al., 2014)]. Most RIDD-targeted mRNAs contain XBP1u-like stem-loop endomotifs that can be cleaved by IRE1. Intriguingly, recent research identified that in ER-stressed cells IRE1α can also remove mRNAs that do not harbor canonical endomotifs; this process was named “RIDD lacking endomotif (RIDDLE)” (Le Thomas et al., 2021). Functionally, RIDD can promote ER homeostasis and cell survival. For example, RIDD activity is increased during ER stress to partially deplete mRNAs for protein translation thus reducing the client protein load of the ER. It can also induce apoptosis and cell death when being constitutively activated by a sustained ER stress (Maurel et al., 2014). In addition to regulation of ER stress-related cell fate, RIDD and RIDDLE also target genes involved in other important cellular functions. For example, DGAT2 mRNA, encoding the rate-limiting enzyme in TAG biosynthesis, can be cleaved by IRE1. Inhibition of IRE1α results in DGAT2-dependent accumulation of TAGs in lipid droplets and sensitizes cells to nutritional stress (Almanza et al., 2022). In addition, several mRNAs including IRF4, PRDM1, IKZF1, KLF13, NOTCH1, ATR, DICER, RICTOR, CDK12, FAM168B, and CENPF were identified as potential RIDD targets, contributing to cell survival and proliferation in myeloma cells (Quwaider et al., 2022). In Drosophila retina, mutation of IRE1 increases the levels of RIDD-targeting mRNAs in photoreceptors, resulting in XBP1-independent defects in rhodopsin-1 protein delivery and rhabdomere morphogenesis (Coelho et al., 2013). The implications of RIDD in mouse and human photoreceptors and other retinal neurons have not been thoroughly studied.

3.2. The PERK pathway

The PERK branch of the UPR is an enzymatic signaling cascade aimed at maintaining homeostatic conditions through regulation of the cellular response to stress (Liu et al., 2000). PERK, one of four eukaryotic translation initiation factor 2 alpha (eIF2α) kinases in mammals (Taniuchi et al., 2016), is the canonical regulator of protein synthesis in response to stresses of the ER, specifically in the form of the UPR (Bertolotti et al., 2000; Liu et al., 2000; Walter and Ron, 2011). At the onset of ER stress, PERK phosphorylates eIF2α to diminish protein synthesis rates (Bertolotti et al., 2000; Ma et al., 2002; Taniuchi et al., 2016). Not only does PERK phosphorylate its namesake translation factor, eIF2α, but, as will be covered in detail, its activity also results in the downstream activation of altered translational and transcriptional programs (Fig. 3) (Jiang and Wek, 2005).

Fig. 3.

The PERK signaling pathway. 1) PERK is held inactive by the binding of its luminal domain by Grp78/BiP. BiP dissociates from PERK’s luminal domain upon misfolded protein accumulation, leaving PERK unbound. 2) PERK dimerizes and becomes activated by autophosphorylation. 3) PERK phosphorylates eIF2α at S51, resulting in the eIF2 complex’s inhibition of eIF2B’s guanine nucleotide exchange function, halting general protein synthesis. 4) Inhibition of general protein synthesis allows key stress related mRNAs (such as ATF4) to be selectively translated. ATF4 is a transcription factor that travels to the nucleus to promote transcription of pro-apoptotic genes such as CHOP, GADD34 and TRB3.

Canonically, PERK is held inactive by the interaction of its ER luminal domain with GRP78 (Bertolotti et al., 2000; Kopp et al., 2019; Walter and Ron, 2011). Upon GRP78’s dissociation from PERK due to the preferential binding of GRP78 to hydrophobic residues of misfolded proteins, PERK dimerizes to form an active homodimer (Kopp et al., 2019; Wang et al., 2018). Following dimerization, PERK becomes activated by phosphorylation at Thr 980 b y autophosphorylation, meaning PERK enzymes phosphorylate each other (Ma et al., 2002; Sood et al., 2000). PERK is also known as eukaryotic translation initiation factor 2-alpha kinase 3 (EIF2AK3), named after its first described role in the phosphorylation of the α-subunit of the eIF2α complex. An active PERK (p-Thr 980) phosphorylates eIF2α at Ser 51 (Gorbatyuk et al., 2020; Sood et al., 2000), a post-translational modification that alters eIF2’s interaction with eIF2B (Kashiwagi et al., 2017; Sudhakar et al., 1999, 2000). The eIF2 complex consists of α, β, and γ subunits, each with distinct roles. The role of eIF2 in translation initiation involves the complex bringing methionine-tRNA (Met-tRNA), the initiator tRNA, to a start codon on a messenger ribonucleic acid (mRNA) in a process dependent on guanosine triphosphate (GTP) being bound to the γ-subunit (Walton and Gill, 1975). eIF2, Met-tRNA, and GTP- called the ternary complex (TC)-converge on the 40 S ribosomal subunit with other translation initiation factors, or eIFs, to form the 43 S pre-initiation complex (PIC) [reviewed extensively in (Hinnebusch et al., 2016; Sonenberg and Hinnebusch, 2009)]. The PIC then scans the mRNA until a start codon is located. When this initiator codon is recognized and bound by the anti-codon on the Met-tRNA, GTP is hydrolyzed by eIF2 and the now inactive ternary complex, is released from the PIC and therefore, must be reactivated before it can participate in another round of protein synthesis. For this to occur, GDP must be exchanged by eIF2B, the nucleotide exchange factor for eIF2. EIF2B binds eIF2β in order to exchange GDP for GTP on eIF2γ (Alone and Dever, 2006; Kimball et al., 1998). Once the exchange occurs, eIF2 can once again participate in translation. However, stress induced eIF2α phosphorylation prevents canonical translation initiation resulting from the failed guanine nucleotide exchange between eIF2 and eIF2B. This is probably due to eIF2B having a higher affinity for p-eIF2α (S51) than it does for eIF2β; therefore, a normally transient complex between eIF2B and eIF2β instead becomes a stable inhibitory complex of eIF2B and p-eIF2α (Gross et al., 1987; Sudhakar et al., 1999). This is thought to result in eIF2B proteins, which are rate-limiting components of TC formation, becoming trapped and preventing traditional translation. When general translation ceases, synthesis of most proteins is repressed; however, translation of some stress-associated mRNAs is promoted in these circumstances, for example, the one encoding activating transcription factor 4 (ATF4) (Vattem and Wek, 2004a). Together with the phosphorylation of eIF2α and the subsequent halt in translation, this alternative translational program is known as the integrative stress response, or ISR. The ISR will be discussed in detail in a later section. ATF4 is a core protein involved in the second part of PERK signaling, one of stress-activated transcription. ATF4 promotes the transcription of various genes including those encoding Growth arrest and DNA-damage inducible 34 (GADD34), C/EBP homologous protein (CHOP), and tribbles homolog 3 (TRB3), each of which has been studied in retinal degeneration and will be discussed in detail in later sections. Of note, CHOP is a pro-apoptotic protein so its production paves the way for ER-stress triggered cell death. GADD34’s primary role in ER stress signaling is promoting the dephosphorylation of eIF2α by protein phosphatase 1 (PP1), thus restoring traditional protein synthesis and sequestering non-canonical protein synthesis (Brush et al., 2003; Connor et al., 2001; Zadorozhnii et al., 2019).

PERK.

PERK is a ~125kD enzyme that spans the ER membrane. PERK contains multiple domains such as the ER luminal domain, the transmembrane domain, and the cytoplasmic portion containing its kinase domain. The crystal structure of PERK’s luminal domain has been delineated. The structure of PERK is similar to that of other two-lobed membrane kinases. The luminal domain of PERK is comprised of its smaller N-terminal lobe (N-lobe), whereas a larger C-terminal lobe (C-lobe) constitutes its cytosolic domain. The C-lobe of PERK contains its catalytic site. The cytoplasmic domain is similar to those of other eIF2α kinases (Cui et al., 2011).

PERK, being studied so extensively in ER stress signaling and neurodegeneration, one may be under the impression that PERK is inherently bad; however, like many proteins of the UPR, transient alterations in their activity can be vital for cellular and organismal survival. For instance, knocking out PERK or inhibiting PERK in young mice leads to very poor survivability in inbred strains (Zhang et al., 2002). This is thought to mainly be due to pancreatic cells’ need for PERK to tightly regulate the production of secreted proteins. Therefore, though PERK signaling may be hyperactive and deleterious in certain neurodegenerative diseases (including those of the retina, as will be discussed in a later section) other tissues simply cannot go without this enzyme.

Though the canonical route to PERK activation remains the most studied, there may be more to this enzyme. For example, Wang et al. recently demonstrated that mammalian PERK can interact with misfolded proteins with a region of its luminal domain, meaning that upon induction of ER stress, PERK itself can act as a sort of chaperone as it binds misfolded polypeptides to initiate its dimerization and ensuing activation, directly arguing against canonical UPR activation (Wang et al., 2018). In addition, this group provided evidence that PERK activation may be independent of its interaction with BiP as cells expressing mutant PERK lacking an active luminal domain did not experience higher than normal PERK signaling but were still susceptible to pharmacological induction of the UPR. Whether this is cell-type specific or a common mechanism across a range of cell types is unknown. van Vliet et al. demonstrated that PERK may be necessary for the formation of ER-plasma membrane associations, a process important for regulating Ca²⁺ stores in the ER (van Vliet et al., 2017).

In people, genetic variants of PERK increase the risk for tauopathy neurodegenerative diseases – Progressive Supranuclear Palsy (Ferrari et al., 2014; Höglinger et al., 2011b; Sanchez-Contreras et al., 2018) and some forms of Alzheimer’s Disease (Liu et al., 2013a; Wong et al., 2019). These variants include amino acid changes that are predicted to disrupt H-bonds in PERK’s ER stress-sensing luminal domain (Park et al., 2023) and impair stability and function in response to ER stress (Yuan et al., 2018). iPSC-generated neurons carrying disease-associated hypomorphic PERK variants displayed increased vulnerability to ER stress toxins and increased tau protein misfolding (Yuan et al., 2018). Chemical or genetic inhibition of PERK increased tau aggregation while activation of the PERK/ISR pathway reduced tau aggregation in vitro (Park et al., 2023). These findings support that people carrying these PERK disease alleles have increased risk for tauopathy neurodegeneration due to reduced PERK signaling that in turn increases tau proteotoxicity and ER stress-induced damage. The role of these hypomorphic PERK alleles in human ocular diseases remains to be studied.

eIF2α.

eIF2α, as mentioned in the previous sections, is a part of a trimeric translation factor complex essential for canonical translation. eIF2α is an ~37kD protein with a key amino acid that is central to stress-induced ternary complex inhibition, S51. eIF2α-S51 appears to largely exist to maintain translational homeostasis. Homozygous mutations at key sites (i.e. S51A) can lead to embryonic lethality in mice (Back et al., 2009; Longo et al., 2021). Therefore, it is of significant importance to use cell-type specific knockouts, knockdowns, or overexpression of dominant mutants of eIF2α when studying this protein mechanistically. Due to eIF2α being the central target of the ISR, it will be covered in more depth in the ISR section.

ATF4.

As mentioned above, one of the following downstream elements of PERK signaling under stress conditions is ATF4. ATF4 is composed of 351 amino acids and is organized into domains crucial for its stability and response to ER stress (Ameri and Harris, 2008). Belonging to basic leucine zipper (bZIP) family transcription factor - ATF4 was described in many reviews (Baird and Wek, 2012; Pavitt and Ron, 2012; Sonenberg and Hinnebusch, 2009). It can function in homodimeric or heterodimeric complexes. Homodimeric complexes are less stable and heterodimeric complex can bind to other bZip family members such as CCAAT box/enhancer-binding protein b (C/EBP) which makes it stable. Heterodimers find DNA targets by binding with cAMP response element (CRE) to regulate transcription (Ebert et al., 2022; Podust et al., 2001).

The ATF4 gene carries two upstream open reading frames (uORF) before the main ATF4 coding region. These two uORFs (1 and 2) act as different critical elements during stress with inhibitory and activating properties. uORF1 facilitates ATF4 coding region translation and uORF2 by contrast blocks ATF4 translation (Pavitt and Ron, 2012; Vattem and Wek, 2004a; Yamaguchi and Wang, 2004). Human ATF4 contains 3 uORF regions. Under stress conditions, scanning starts from the second uORF (Harding et al., 2000; Lu et al., 2004; Pakos-Zebrucka et al., 2016; Pavitt and Ron, 2012; Vattem and Wek, 2004b). Interestingly, a similar mechanism has been characterized for yeast transcriptional activator general control nonderepressible 4 (GCN4) (Hinnebusch, 2005). Chan et al. studied human ATF4 and reported that translation of ATF4 is mediated by internal ribosome entry site (IRES) (Chan et al., 2013). ATF4 was shown to activate C/EBP homolog protein (CHOP or GADD153) and promote ER stress-mediated apoptosis (Marciniak et al., 2004). Enhanced expression of CHOP under ATF4 activities was demonstrated in arsenite-treated rat cells (Fawcett et al., 1999). A study led by Kaufman et al. identified that ATF4 interacting with CHOP can lead to cell death (Han et al., 2013). Therefore, ATF4/CHOP regulation plays an important role in pro-apoptotic events of the cellular stress response.

GADD34.

GADD34, first identified as a pro-apoptotic member of the GADD (growth arrest and DNA-damage) family of proteins, is now perhaps most studied for its role as a phosphatase regulatory subunit responsible for assisting protein phosphatase 1 (PP1) in dephosphorylating eIF2α to promote translational recovery after transient ER stress. In this role, GADD34 provides the ISR with a much-needed feedback loop in order to restore translational homeostasis following a bout of stress. Importantly, ATF4 acts on the GADD34 promoter to facilitate its transcription and restore protein synthesis rates. In addition, c-Jun has been shown to upregulate the GADD34 gene during DNA damage and proteotoxic stress (i.e., Alzheimer’s disease) (Xu et al., 2015b). GADD34 upregulation well correlates with apoptotic signaling; however, the mechanism of GADD34’s induction of apoptosis is not well understood. In one interesting study, Farook et al. showed that its pro-apoptotic role could stem from GADD34-mediated inhibition of AKT phosphorylation (Farook et al., 2013). This group reported that GADD34 interacts with TRAF6 to block polyubiquitination of AKT, a process that would set AKT up for subsequent phosphorylation. In addition, Shi et al. reported that GADD34 interacts with and initiates the PP1-mediated dephosphorylation of SMAD7 (Shi et al., 2004).

Interestingly, there is also evidence that GADD34 may have anti-apoptotic functions, which appear to be independent of its role in restoring protein synthesis. For example, Krokowski et al. demonstrated that GADD34 is essential for the survival of cultured corneal cells subjected to hyperosmotic stress (Krokowski et al., 2017). This group highlighted that GADD34 may be important for the integrity of the Golgi as a lack of GADD34 led to a fractionated Golgi body in hyperosmotic corneal cells. Of note, fractionated Golgi further led to less membrane translocation of receptors in these cells, suggesting that GADD34’s functions may be far more complicated than previously thought. In fact, Dedigama-Arachchige et al. revealed that GADD34/PP1 could act on hundreds of proteins to alter the metabolic state of a cell (Dedigama-Arachchige et al., 2018). The reality that GADD34/PP1 could act on so much more than eIF2α is only recently coming to light.

CHOP.

CHOP, also known as GADD153, is a well-characterized transcription factor that regulates several stress-response genes. CHOP is one example of a transcription factor with a very interesting function: it can form heterodimers with ATF4 and alter ATF4’s transcriptional activity (Talukder et al., 2002). In fact, ATF4/CHOP interaction has been proposed to be a prerequisite for transcription of TRB3 (Ohoka et al., 2005).

It is generally accepted that chronic or prolonged PERK signaling can lead to apoptosis and CHOP appears to be a primary player in this process. CHOP activation leads to the transcription of apoptosis-promoting genes (Sano and Reed, 2013) and inhibition of pro-survival proteins. For example, the genes encoding Bcl-2 interacting mediator of cell death (BIM) (Altman et al., 2009; Puthalakath et al., 2007) and death receptor 5 (DR5) (Kim et al., 2008; Kouhara et al., 2007; Yamaguchi and Wang, 2004) are targets of CHOP. In addition, CHOP down-regulates B-cell lymphoma 2 (BCL-2), a vital inhibitor of apoptosis (McCullough et al., 2001). ATF6 can also promote transcription of CHOP (Yoshida et al., 2000), and therefore may support programmed cell death in prolonged ER stress. BIM can activate another pro-apoptotic protein known as BCL2 antagonist/killer (BAK), which can signal for mitochondria-mediated apoptosis (Ord and Ord, 2003). In addition, Li et al. reported that CHOP can activate endoplasmic reticulum oxidore-ductase 1 alpha (ERO1α), which hyperoxidizes the ER and may lead to programmed cell death (Li et al., 2009a). The group also reported that ERO1α promotes IP3R migration to the mitochondria, further promoting apoptosis. Calcium pumping into the mitochondria promotes the release of cytochrome C, which can result in apoptosis signaling through caspases (Li et al., 2009a).

TRIB3.

TRIB3, or Tribbles homolog 3, has been extensively studied due to its crucial role as a signal mediator binding with various proteins such as kinases, phosphatases, and transcription factors. TRIB3 was reported as a novel gene with different names such as TRB3, NIPK, SKIP3, and SINK (Ord and Ord, 2017). Matsuda et al. studied cell death induced by neuronal growth factor depletion and identified a novel gene called NIPK (neuronal cell death inducible putative kinase) with unknown function (Mayumi-Matsuda et al., 1999). This was the beginning of TRIB3 discovery. The original name of TRIB3 came from homology of tribbles gene in Drosophila. The novel tribbles gene was characterized as a crucial regulator of oogenesis (Rorth et al., 2000). In addition, tribbles was shown as an important regulator of mitosis in Drosophila development (Mata et al., 2000). An interesting systemic analysis from Hernández-Quiles et al. showed a conservative similarity of 55% in pseudokinase domain between human TRIB1/2 and 3; However, they only observed a 9% similarity in the C terminal domain of TRIB1 and TRIB3, which is assumed to be due to their unique functions (Hernandez-Quiles et al., 2021). Structurally, TRIB3 has a pseudokinase domain, PEST (proline, glutamic acid, serine, and threonine rich region) and COP1 binding domain (Eyers et al., 2017; Stefanovska et al., 2021). It was reported that TRIB3 in association with E3 ubiquitin ligase protein COP1 leads to the degradation of Acetyl–coenzyme carboxylase (ACC) in adipose tissue (Qi et al., 2006). Its pseudokinase domain has 12 subdomains with a lack of phosphorylation sequence, and since it lacks a true kinase function, is termed a pseudokinase (Hanks and Hunter, 1995; Hegedus et al., 2007; Prudente et al., 2012). Mouse and human amino acid sequence of TRIB3 is highly conservative. Its biological function includes regulation of cell growth, apoptosis, differentiation, and metabolism (Prudente et al., 2012). TRIB3 has been found to interact with AKT, NF-κB, mTORC2, MAPK (Du et al., 2003; Kiss-Toth et al., 2004; Salazar et al., 2015; Wu et al., 2003).

Various stress conditions induce TRIB3 gene expression, for instance, ER stress (Ohoka et al., 2005; Ord and Ord, 2005) and toxic chemicals (Ord et al., 2009). Earlier reports identified TRIB3 as an interacting partner of ATF4 and characterized its transcriptional activity in yeast two-hybrid analysis (Ord and Ord, 2003). Ord et al. showed that TRIB3 can regulate ATF4 expression in a negative feedback mode with induced stress in vitro (Ord and Ord, 2005). Interestingly, the functions of both ATF4 and CHOP during ER stress are regulated in cooperation with TRIB3 and are important in regulating cell death (Ohoka et al., 2005). Other evidence showed that neuronal cells respond to ER stress-mediated apoptosis through TRIB3 (Zou et al., 2009). These studies indicate that TRIB3 serves as a sensor for ATF4/CHOP axis cell death occurring during ER stress. Conversely, several publications determined involvement of TRIB3 in cell survival (Ord et al., 2007, 2015; Schwarzer et al., 2006) and death mechanisms (Humphrey et al., 2010; Ohoka et al., 2005; Salazar et al., 2013; Wu et al., 2003).

One of the interesting functions of TRIB3 is in the cell’s response to toxic chemicals. A report from Ord et al., described cytotoxic effects of arsenite and molecular changes based on TRIB3 expression transcriptionally and translationally (Ord et al., 2016). In that study, the level of TRIB3 mRNA and protein was increased in response to arsenite-induced stress. When TRIB3 was silenced, CHAC1 (glutathione degrading enzyme) expression was elevated, inducing cell death. Importantly, the sensitivity of cells to arsenite-induced stress was decreased by TRIB3-mediated reduction of CHAC1.

3.3. The ATF6 pathway

ATF6 controls a key UPR signal transduction pathway (Walter and Ron, 2011). ATF6 encodes a basic leucine zipper (bZIP)-domain transcription factor that is tethered to the endoplasmic reticulum membrane and found in all cells (Haze et al., 1999). Under resting conditions, ATF6 may form intermolecular disulfide bridges between luminal domains to generate ATF6 dimers and oligomers in the ER (Koba et al., 2020; Nadanaka et al., 2007). In response to increased ER/oxidative stress, ATF6 is fully reduced (Nadanaka et al., 2006, 2007), and the ATF6 monomer then traffics from the ER to the Golgi (Chen et al., 2002; Sato et al., 2011; Shen et al., 2002). In the Golgi, site-1 and site-2 proteases cleave ATF6 in its transmembrane domain to liberate the bZIP portion of ATF6 into the cytosol (Ye et al., 2000). The severed ATF6 transcription factor then migrates to the nucleus where it transcriptionally upregulates target genes (Fig. 4) (Shoulders et al., 2013; Yoshida et al., 2000, 2001b). ATF6’s transcriptional targets have been identified through genetic ablation, microarray/RNA-seq, and chromatin immunoprecipitation experiments and include ER protein folding chaperones; protein folding enzymes such as oxidoreductases and protein disulfide isomerases; and proteasomal degradation cofactors (Bommiasamy et al., 2009; Kroeger et al., 2018, 2021; Lee et al., 2020, 2022; Okada et al., 2002; Wu et al., 2007; Yamamoto et al., 2004, 2007). These chaperones and enzymes improve the fidelity of protein folding, ensure redox balance within the cell, and promote the degradation of damaged proteins (Adachi et al., 2008). The overall result of ATF6 activation in the cell is the reduction of levels of damaged proteins and the reduction of oxidative and ER stress levels (Nadanaka et al., 2004). Hence, ATF6’s transcriptional program helps cell survival during physiologic, pathologic, and environmental conditions that cause oxidative or ER stress.

Fig. 4.

The ATF6 signaling pathway. Full length ATF6 can be present as a monomer, dimer, or oligomer via disulfide bond. Under ER stress, reduced ATF6 monomers traffics from the ER to the Golgi compartment. S1P and S2P proteases cleave ATF6 in the Golgi apparatus to release the cytosolic bZIP transcriptional activator ATF6 domain. Liberated ATF6 moves to nucleus to transcribe target genes. Class 1 ATF6 mutants Y567N, D564G, G512Lfs*39, and L479Vfs*11 show impaired ER-to-Golgi trafficking (blue box). Class 2 ATF6 mutants R376*, V371Sfs*3, N366Hfs*12, and R324C have fully intact ATF6 cytosolic domain and show constitutive transcriptional activator function (green box). The ATF6 mutant I304_R573del has defect in the in-frame bZip, transmembrane, and luminal domains of ATF6. Class 3 ATF6 mutants N267*, E119Gfs*8, P118Lfs*31, M67V, D28Gfs*36, and D28_T82del do not have a functional bZIP domain (purple box) and fail to up-regulate ATF6 target genes. Activation of the ATF6 pathway of the UPR.

To probe the role of ATF6 signaling in diseases linked to oxidative and ER stress, ATF6−/− mice were generated in 2007 independently by two research teams (Wu et al., 2007; Yamamoto et al., 2007). Mouse embryonic fibroblasts (MEFs) from ATF6−/− mice showed abnormal responses to oxidative and ER stress including: 1) defective induction of ER protein folding chaperones, ER protein folding enzymes; and proteasomal degradation factors. 2) increased cell death in response to chemical agents that induce ER and oxidative stress in cell culture. 3) impaired clearance of misfolded proteins in cell culture (Wu et al., 2007; Yamamoto et al., 2007). Despite these defects in cell culture, ATF6−/− knockout mice are viable, develop to adulthood and produce offspring at normal Mendelian ratios (Wu et al., 2007; Yamamoto et al., 2007). These findings reveal that ATF6 is not essential for mouse development, viability, and survival under normal laboratory environments.

However, ATF6−/− mice show heightened sensitivity to many different physiologic and pathologic stresses. ATF6−/− mice are prone to pancreatic b-cell failure when fed a high-fat diet (Usui et al., 2012). ATF6−/− mice are prone to liver steatosis when they are given intraperitoneal injections of the oxidative and ER stress-inducing toxin, tunicamycin (Yamamoto et al., 2010). Dopaminergic neurons from ATF6−/− mice are prone to die in response to the oxidative stress-inducing neurotoxin, MPTP (Egawa et al., 2011). These studies suggest that ATF6 is important for protecting tissues and cells in animal metabolic and brain disease paradigms experimentally induced by oxidative and ER stress.

In people, ATF6’s link to metabolic disease is less clear: over 20 variants of ATF6 were reported to be associated with type 2 diabetes including Met67Val, Pro145A1a, or Ser157Pro coding changes in ATF6 in Pima Indian and Dutch cohorts (Meex et al., 2007; Thameem et al., 2006). However, these ATF6 variants showed no significant association in another study of Pima Indian, Caucasian, and Chinese patients with type 2 diabetes (Chu et al., 2007). The Met67Val ATF6 variant was linked to increased plasma cholesterol levels in a study of Dutch families with familial combined hyperlipidemia, the variant was found to increase ATF6 transcriptional activity (Meex et al., 2009). However, another study found no differences in ATF6 transcriptional activity when residue 67 was converted from methionine to valine (Lee et al., 2020). While ATF6’s role in the pathogenesis and progression of human metabolic diseases requires more rigorous evaluation, more recently, a separate group of ATF6 variants were identified that directly cause photoreceptor disease in patients. ATF6’s essential role in human vision will be the focus of Section 6.

4. ER stress signaling in the RPE and AMD

The RPE plays a key role in supporting the function and survival of photoreceptor cells in the retina. The RPE is composed of a single layer of cuboidal epithelial cells situated between the choroidal vasculature and the outer segments of the photoreceptors. The RPE cells form the outer BRB through tight junctions at their basolateral side. This monolayer of RPE cells also adheres to a well-organized basement membrane known as Bruch’s membrane, which separates the RPE from the fenestrated endothelium of the choroidal capillaries (Fields et al., 2020; Strauss, 1995). On its apical side, the RPE faces the photoreceptor OS and directly envelops them with its microvilli. The ideal anatomical location allows the RPE to perform essential functions to support the neuroretina, in particular photoreceptor cells. Dysfunction of the RPE can lead to photoreceptor degeneration and contribute to the development of retinal disorders, such as AMD (Bird, 2021; Bonilha et al., 2006).

AMD is the most frequent cause of vision impairment resulting in progressive loss of the central vision in the elderly (Mitchell et al., 2018). In developed countries, approximately 10% of individuals over 65 years and 25% of those over 75 years have been diagnosed with AMD. As life expectancy rises, the prevalence of AMD is increasing, with an estimated prevalence of nearly 300 million people worldwide being affected by 2040 (National Eye Institute, 2019; Wei et al., 2019). The increasing prevalence of AMD highlights its significance as a global health concern. Clinically, AMD can be classified into two stages: early and late AMD. The early stages of AMD are characterized by the presence of small extracellular yellowish deposits or drusen as well as depigmentation and impaired functioning of the RPE layer (Fleckenstein et al., 2021; Lim et al., 2012). Late stages of the disease can be subcategorized as non-neovascular (dry) or neovascular (wet) AMD. Both forms of advanced AMD are characterized by the loss of photoreceptors and the development of geographic atrophy (GA). Neovascular AMD (nAMD) is defined by the presence of pathological angiogenesis in the macula, known as macular neovascularization (MNV) (Chen et al., 2020; Spaide et al., 2020). Common consequences of MNV include exudate formation, hemorrhages, edema in the macula, and fibrotic scar formation often resulting in severe visual impairment (Fleckenstein et al., 2021; Hadziahmetovic and Malek, 2020). Damage or stress to the RPE is believed to promote the production of pro-angiogenic factors and may contribute to MNV (Ambati and Fowler, 2012).

4.1. Role of the ER in maintaining the RPE function

A well-maintained ER machinery is required for the RPE to fulfill its vital supportive roles through a diverse range of intracellular events such as visual cycle that regenerates the visual pigments, phagocytosis that removes the damaged and aged photoreceptor OS, and secretion of neurotrophic and growth factors to nourish photoreceptors and choroidal vessels, just to name a few. The RPE cells are enriched with tubular smooth ER, which occupies a large volume of the cytoplasm in particular in the basal portion of the cells (Porter and Yamada, 1960). The smooth ER forms close tridimensional lattice and can also present but less frequently as the myeloid body (Porter and Yamada, 1960). The smooth ER is the primary site in the RPE where the classic visual cycle occurs. It harbors several key enzymes, such as RPE65, which catalyzes the isomerization of all-trans retinyl esters to 11-cis-retinol, and lecithin: retinol acyltransferase (LRAT), which generates fatty acid retinyl esters providing substrates for RPE65. These enzymes are critical for the regeneration of visual pigments, which are required for photoreceptor cells to respond to light stimuli (Cheng et al., 2020b; Fisher and Ferrington, 2018; Viegas and Neuhauss, 2021). Mutations of RPE65 or LRAT genes lead to congenital or early-onset retinal degeneration, such as Leber congenital amaurosis (LCA) (Cai et al., 2009; Sears and Palczewski, 2016). In diabetes, the visual cycle function is compromised contributing to vision impairment. Inhibiting ER stress successfully restored the expression levels of RPE65 and several other key visual cycle enzymes in the RPE, suggesting that maintaining ER homeostasis is critical for visual cycle function (Kang et al., 2018).

An important and well-studied role of the RPE is detoxifying and phagocytosis of shed photoreceptor OS. The phagocytosis process in the RPE involves phagosome formation and maturation followed by breakdown and resolution of the ingested photoreceptor OS and the membrane-associated components are then recycled back to photoreceptors for their use (Kwon and Freeman, 2020). The sequential stages of phagosome maturation in the RPE were also observed in macrophages exposed to photoreceptor outer segment (POS), suggesting similarities in the phagocytosis process in these cell types (Silène T. Wavre-Shapton et al., 2014). In phagocytes of the innate immune system, the ER plays a critical role in phagosome formation, maturation, and possibly phagolysosome repairing by forming the ER-phagosome contact sites (Ghavami and Fairn, 2022). In macrophages, upon particle entry the ER is recruited and fused with the plasma membrane providing building material for phagosome formation to avoid the use of a highly specialized plasma membrane(Gagnon et al., 2002). In neutrophils, ER transmembrane proteins, such as stromal interaction molecule (STIM) proteins, are responsible for the remodeling of the ER membranes near adjacent phagosomes (Orci et al., 2009). These molecules sense ER Ca²⁺ depletion and induce a conformational change of the ER to increase the availability of membrane contact sites and subsequently form tight phagosomal-ER junctions (Nunes et al., 2012; Stendahl et al., 1994). STIM1 also interacts with store-operated Ca²⁺ entry (SOCE) channels on the phagosome membrane which results in highly localized Ca²⁺ signals necessary to sustain phagocytosis (Lewis, 2007; Nunes et al., 2012; Orci et al., 2009).

The RPE secrete a large array of proteins and growth factors that nourish photoreceptor cells and choroidal vessels. For example, polarized mature RPE cells secret pigment epithelium growth factor (PEDF) from their apical surface providing neurotrophic support to photoreceptors and secret vascular endothelial growth factor (VEGF) from their basal side, which is essential for the maintenance of choriocapillaris (Saint-Geniez et al., 2009; Sonoda et al., 2009). As epithelium, the RPE provides metabolic support to photoreceptors by facilitating the transport of nutrients to these cells. In addition, the RPE is considered an integral part of the metabolic ecosystem within the eye. In this intricate system, glucose derived from the choroid is transported through the RPE to photoreceptors (Swarup et al., 2019). Subsequently, photoreceptors convert glucose into lactate, which serves as a fuel source for the RPE and neighboring retinal cells. Lactate also inhibits glycolysis in the RPE, thereby preserving glucose for utilization by the photoreceptors (Kanow et al., 2017). This intricate metabolic interplay ensures the availability of energy resources and contributes to the overall homeostasis of the retina.

4.2. ER stress signaling in RPE survival

Given its multiple significant roles in supporting the neuroretina, in particular phagocytosis and detoxification of photoreceptor OS, the RPE is equipped with sophisticated systems to reduce chronic physiological and pathological stresses, such as aging, light phototoxicity, cigarette smoke, and so on. These factors with time increase oxidative stress, which triggers ER dysfunction and induces ER stress; sustained ER stress, in turn, exacerbates redox imbalance resulting in oxidative damage. Thus, oxidative stress and ER stress often go hand-in-hand in RPE dysfunction and pathogenesis (Chen et al., 2014a; Huang et al., 2015b; Sreekumar et al., 2016). A recent single-cell transcriptomic study shows that genes associated with cellular response to ER stress and oxidative stress, such as HERPUD1, HMOX1, MDM2, and XBP1, are highly enriched in the macular RPE (Xu et al., 2021). These genes are believed to provide great ability to the macula to respond to stresses and injuries. Knockout or inhibition of XBP1 activation leads to increased RPE apoptosis, tight junction damage, and RPE/photoreceptor induced by oxidative stress or ER stress (Chen et al., 2014a; Ma et al., 2016; Zhong et al., 2012b). Reduced expression of PERK has been observed in the RPE of human donors with early and intermediate AMD (Porter et al., 2019; Saptarshi et al., 2022). Downregulation of PERK increases ER stress, impairs autophagic flux, and reduces antioxidant response in ARPE-19 cells challenged with brefeldin A. This suggests that the PERK pathway, along with the IRE/XBP1 pathway, plays a critical role in the RPE response to ER stress. However, it should be noted that in chronic stress conditions overactivation of the PERK-downstream effectors, ATF4 and CHOP, can lead to ER stress-associated apoptosis and cell death (Chen et al., 2014a; Li et al., 2014b; Ma et al., 2016; Zhong et al., 2012b). Yet, complete silencing of the CHOP gene in the RPE increases cell apoptosis and this effect is associated with a reduction of nuclear factor erythroid 2-related factor 2 (Nrf2), a master regulator of the cell’s anti-oxidant and detoxification function, suggesting that an optimal level of CHOP is crucial for Nrf2 activation and cell survival in stressed RPE (Huang et al., 2015b). In addition to CHOP, other ER stress signaling molecules also regulate or interact with Nrf2. Deletion of XBP1 decreases Nrf2 protein levels and its downstream target gene expression while downregulation of PERK reduces Nrf2 phosphorylation in RPE cells (Chen et al., 2018; Saptarshi et al., 2022). Overexpression of Nrf2 reduced ER stress-induced RPE cell death and sufficiently rescued the RPE in mouse models of retinitis pigmentosa (RP) (Huang et al., 2015b; Wu et al., 2021). These results suggest that the oxidative and ER stress signaling may converge on common pathways, such as Nrf2, to regulate RPE cell survival.

Disturbance in the phagocytotic process and/or delayed clearance of the POS in the RPE can cause increased oxidative stress and subsequent ER stress, whereas sustained ER stress can also lead to dysfunctional phagocytosis. For example, excessive accumulation of visual cycle metabolites such as all-trans retinal (atRAL) due to mutations of retinol dehydrogenase 12 (RDH12) leads to increased oxidative and ER stress resulting in apoptotic RPE cell death (Li et al., 2015; Zhu et al., 2016) and disrupted phagocytosis (Gal et al., 2000; Zhang et al., 2020). In a zebrafish rdh12 mutant model, the POS were not able to penetrate beyond the apical surface of the RPE, indicating an abnormal RPE phagocytosis (Sarkar et al., 2021). The impaired communication between RPE and POS may trigger the activation of chronic ER stress response in photoreceptors causing POS degeneration and eventual cell death. In addition, transmission electron microscopy revealed enlarged, undigested phagosomes in the RPE cells, which suggests an impaired POS recycling and the latter has been linked to the LC3-associated phagocytosis (LAP) pathway involving autophagy proteins (Kim et al., 2013). Regulation of autophagy by the ER stress signaling in RPE pathophysiology will be discussed in Section 9. A recent study shows that tauroursodeoxycholic acid (TUDCA), a bile-derived neuroprotectant, can protect against oxidative stress-induced RPE phagocytic dysfunction through enhancing Mer tyrosine kinase receptor (MerTK) (Murase et al., 2015). While TUDCA has demonstrated an anti-ER stress effect in RPE cells (Chen et al., 2014a), in this study the authors reported that TUDCA did not reduce the increase of ER stress marker and thus concluded that TUDCA may promote the phosphorylation of MerTK through an independent pathway of ER stress. Further studies are needed to determine if inhibition of ER stress can promote phagocytosis via regulation of tyrosine kinase.

4.3. ER stress signaling in RPE barrier function

The integrity of the BRB is essential to maintaining the microenvironmental homeostasis of the neural retinal tissue. Damage to the RPE cell tight junctions and adherens junctions resulting in disruption of the barrier integrity of the outer BRB contributes to the pathogenesis of AMD. Studies have shown that prolonged exposure to cigarette smoke, a most significant environmental risk factor for AMD development, or continuous light can trigger the activation of ER stress signaling, e.g. the PERK pathway, in RPE cells and lead to disruption of the outer BRB (Huang et al., 2015a; Song et al., 2020). Overexpression of ER chaperone ERp29 or treatment with cyanidin-3-glucoside (C3G), a natural water-soluble plant pigment demonstrating anti-oxidant properties, can reduce ER stress and PERK activation, restore tight junction formation, and preserve the barrier function in the RPE (Huang et al., 2015a; Song et al., 2020). While these studies provide indirect evidence supporting the role of PERK activation in RPE tight junction damage, experiments to elucidate the exact role of PERK in RPE barrier formation using loss-of-function and gain-of-function approaches are needed in future research. In addition, both ERp29 and C3G have been found to activate Nrf2 which increases the RPE’s antioxidant defense (Chen and Cubillos-Ruiz, 2021; Peng et al., 2022b). Whether their pro-Nrf2 effects are dependent or independent of the PERK pathway remains unclear.

The implication of the IRE1/XBP1 pathway in regulation of the RPE barrier function has been studied using conditional XBP1 knockout animals and RPE cell culture treated with pharmacological inhibitors of XBP1 splicing (Ma et al., 2016). The authors demonstrated that inducing ER stress by thapsigargin, which depletes the ER calcium storage causing calcium dyshomeostasis, is sufficient to damage RPE junctions. Moreover, deletion of the XBP1 gene in vivo or inhibition of XBP1 activation in cultured ARPE-19 cells and primary primate RPE cells both result in impaired RPE tight junction formation. Mechanistically, inhibition of XBP1 increases intracellular calcium levels, possibly through dysregulation of the calcium channel protein ryanodine receptors on the ER membrane, resulting in activation of the Rho/Rho kinase signaling pathways (Ma et al., 2016). Activation of the Rho/Rho kinase then causes aberrant cytoskeletal rearrangement resulting in the dissociation of the RPE tight junctions. In mammary epithelial cells, activation of transient receptor potential vanilloid 4 (TRPV4), a major calcium channel protein on the plasma membrane, increases XBP1 splicing and expression of tight junction proteins while knockdown of XBP1 blocks TRPV4-induced increase in tight junction component (Islam et al., 2020). These findings further support the role of XBP1 in the regulation of calcium homeostasis and tight junction formation in epithelial cells including the RPE.

4.4. ER stress signaling in epithelial to mesenchymal transition (EMT)

Recent studies have shed light on an association between ER stress and EMT, a biological process responsible for fibrosis development in various cell types, including RPE cells (Ouyang et al., 2022; Zhou et al., 2020). EMT is present in the later stages of neovascular AMD, whereby RPE cells, which are normally a monolayer of polarized epithelial cells, undergo phenotypic changes and lose their epithelial characteristics and barrier function due to growth factor and cytokine stimulation (Tenbrock et al., 2022). Findings from Ouyang et al. demonstrated that inducing mild ER stress by low doses of tunicamycin and thapsigargin increases GRP78 expression, enhances tight junction protein level, and reduces fibrotic genes (EMT markers) including fibronectin and a-smooth muscle actin. Moreover, pretreating RPE cells with low doses of ER stress inducers significantly blocks TGF-β-induced upregulation of EMT marker proteins and suppresses migration of RPE cells. These findings provide strong support that activation of the UPR can protect RPE cells from EMT, although the underlying mechanisms are yet to be investigated (Ouyang et al., 2022). In contrast to the protective effect of low-dose ER stress inducers, other studies demonstrate that inducing ER stress with high doses of tunicamycin and thapsigargin disrupts epithelial tight junctions and increases EMT (Ma et al., 2016; Zhou et al., 2020). Inhibition of ER stress by chemical chaperones suppresses EMT suggesting high level or sustained ER stress promotes EMT contributing to human eye diseases such as cataract and AMD.

4.5. ER stress signaling in choroidal neovascularization (CNV)

Choroidal neovascularization (CNV) is the most common type of MNV and a hallmark feature of wet AMD. The mechanisms of CNV are complex involving dysregulation of the VEGF signaling that stimulates choroidal endothelial cell proliferation, migration, and forming aberrant new vessels and non-VEGF dependent pathways associated with a diverse array of processes including increased inflammation, complement activation, enhanced oxidative stress, and ER stress. Activation of the PERK pathway by ER stress contributes to RPE dysfunction and CNV formation in AMD. Evidence from earlier studies revealed a critical role of ATF4 in the upregulation of pro-angiogenic factors such as VEGF in RPE cells under conditions of stress or hypoxia (Oskolkova et al., 2008; Roybal et al., 2004). Inhibition of PERK by GSK2606414 reduces VEGF expression in RPE cells (Jiang et al., 2017). In addition, recent work highlights the important role of microRNAs (miRNAs) in angiogenesis (Suárez and Sessa, 2009). Activation of the PERK-ATF4 pathway in RPE cells inhibits the transcription of the miR-106b-25 cluster resulting in a downregulation of miR-106 b. Decreased miR-106 b level then increases the production of VEGF contributing to pathological angiogenesis in retina and choroid in animal models of laser-induced CNV and OIR (Menard et al., 2020). In addition to the PERK-ATF4 pathway, activation of the IRE1/XBP1 pathway and the ATF6 pathway also participates in CNV pathogenesis (Arjunan et al., 2021). Knockdown of IRE1a or ATF6 partially inhibits VEGF-induced in vitro angiogenesis and potentiates the anti-angiogenic effect of the anti-VEGF treatment in an animal model of CNV, suggesting that the IRE1a and ATF6 pathways may regulate angiogenesis through VEGF-independent mechanisms (Liu et al., 2013b). In addition to RPE cells, activation of the UPR in macrophages promotes their polarization toward a pro-angiogenic M2 phenotype (Li et al., 2021b). Treatment of choroidal endothelial cells with conditioned medium from M2 macrophages promotes cell proliferation and tube formation and this pro-angiogenic effect was blocked by inhibition of XBP1 splicing in macrophages. These results suggest that XBP1 activation promotes M2 macrophage differentiation and potentially contributes to CNV formation.

5. ER stress signaling in inherited retinal degeneration

Inherited retinal degeneration (IRD) refers to the heterogeneous group of retinal disorders affecting both rod and cone photoreceptors, leading to severe vision loss and blindness in some cases. One example of an IRD is retinitis pigmentosa (RP), which primarily affects rod photoreceptor viability, leading to the subsequent loss of cone photoreceptors, while cone-rod dystrophy (CORD), another example, is characterized by impaired color vision, blind spots in the center of the visual field, and peripheral vision loss. Overall, mutations in more than 300 genes that eventually lead to visual impairment have been identified either in rods or cones. The presence of similar clinical phenotypes resulting from mutations in different genes can pose challenges to the diagnosis of IRDs (Gorbatyuk et al., 2020; Hu et al., 2021a). Moreover, no available treatment to prevent irreversible vision loss exists for the majority of IRDs. However, ongoing research in gene therapy holds great promise in addressing this critical demand, thus supporting the hope of improving vision loss in individuals with IRD. It is important to note that future advancements in retinal gene therapy rely strongly on the identification of feasible therapeutic targets. Moreover, the success of gene therapy is determined based on two critical factors: the ability to effectively target specific cell types in the retina, and the ability to achieve adequate levels of therapeutic transgene expression. Therefore, advancements in our understanding of the molecular mechanisms involved in IRD would not only significantly improve the diagnostic assessment of patients through genetic testing, but also facilitate the development of gene and/or cell therapy.

RP is the most common form of IRD associated with vision loss, which ranges from mild to severe. Human RP is classified based on its syndromic or non-syndromic nature, as well as the mode of inheritance (Daiger et al., 2013). Syndromic RP includes conditions such as Usher syndrome and Bardet–Biedl syndrome. In contrast, non-syndromic RP encompasses all modes of inheritance, including autosomal dominant, autosomal recessive, X-linked, and the remaining forms that are yet to be determined. Non-syndromic types account for approximately 70–80% of all cases of RP (Dias et al., 2018; Verbakel et al., 2018).

A typical manifestation of RP is night vision loss accompanied by a progressive decline in the visual field. Fundus abnormalities in RP patients commonly include bone spicule pigmentation. Diagnosis can further be supported by an electroretinogram, which typically reveals the characteristic loss of photoreceptor function, which predominantly affects rod photoreceptors in the early stages of the disease, as opposed to cones (Verbakel et al., 2018). The molecular mechanism of photoreceptor degeneration has been linked to changes in retinal metabolism, receptor expression, and neuronal network remodeling, which eventually lead to photoreceptor cell death. The latter can be observed through the distinct IRD phases ranging from photoreceptor stress to photoreceptor degeneration, and the occurrence of this phenomenon is expected to influence the effectiveness of therapeutic approaches aimed at restoring or preserving vision (Pfeiffer et al., 2020). The 2007 study by Lin et al. was one of the first to introduce UPR signaling into vision research as a molecular mechanism of IRD, reporting that the degenerating retinas of P23H RHO transgenic rats exhibited the activation of UPR signaling (Lin et al., 2007). This activation was evidenced by an increase in CHOP and GRP78 mRNA levels. Furthermore, P23H RHO knock-in mice crossed with ERAI mice (carrying a GFP reporter of UPR activity) or Ub-GFP mice (carrying a GFP reporter of proteasome activity) showed selective GFP induction in rods expressing P23H rhodopsin providing evidence of proteostatic activation in diseased rods (Alavi et al., 2015; Chiang et al., 2015; Lobanova et al., 2013). Since then, mounting evidence of UPR’s contribution to the mechanism of IRD has accumulated. However, the question of whether the sustained activation of UPR is beneficial or harmful for degenerating photoreceptors is still under investigation due to existing discrepancies in the impact of individual UPR mediators on photoreceptor homeostasis. These discrepancies include the use of different animal models of IRD and the analysis of different time points of disease progression.

The role of PERK has been investigated in rat and mouse models of IRD. A study conducted by Cheetham’s lab with P23H RHO transgenic rats demonstrated that PERK inhibition with GSK2606414 A, which led to an inhibition of eIF2α phosphorylation, was correlated with reduced ERG retinal function (Athanasiou et al., 2017). A similar effect of PERK inhibition has been reported in P23H RHO knock-in mice, where it led to a significant increase in photoreceptor cell death (Comitato et al., 2020). Moreover, the authors reported that long-term stimulation of PERK has a protective effect by phosphorylating the nuclear factor erythroid 2-related factor 2 (NRF2) transcription factor, which is associated with antioxidant responses. In Chiang et al.’s study, it was revealed that in cells expressing T17M, P23H, Y178C, C185R, D190G, K296E, and S334ter rhodopsin proteins, selective activation of PERK prevents mutant rhodopsin from accumulating in cells (Chiang et al., 2012). In our study with rd16 mice manifesting severe retinal degeneration, we also manipulated the level of PERK mediators in the retinas using GSK2606414 and a genetic approach to knockout PERK in the photoreceptors by employing the PERK floxed and iCre mice. We found that while these manipulations modulated the p-eIF2α level, they did not lead to a complete recovery in translation or affect further retinal function loss (Starr and Gorbatyuk, 2019a). Specifically, neither an increase nor a decrease in the scotopic A- and B-wave ERG amplitudes were detected in the PERK-deficient photoreceptors. Therefore, we concluded that in rd16 mice with rapid retinal degeneration, the PERK deficit does not play a dramatic role in IRD progression, as it may in other models (Athanasiou et al., 2017). Another study conducted with rd1 mice suggested that the time-dependent upregulation of p-PERK coincided with preceded photoreceptor apoptosis (Yang et al., 2007). At the peak of apoptosis, p-PERK was primarily located in the photoreceptor’s inner segments, the outer nuclear layer, or both. The authors thus proposed that ER stress modulators may be strong candidates as therapeutic agents in the treatment of retinal degenerative diseases.

Recently, it has been suggested that the PERK mediator could be responsible for thapsigargin (TG)-induced CXCL10 and CCL2 mRNA expression in photoreceptors (Zhu et al., 2017). In this particular study, the knockdown of PERK-attenuated CXCL10 and CCL2 mRNA elevation was associated with significant decreases in p–NF–κB RelA and STAT3. In contrast to PERK, knockdown of XBP1 robustly enhanced TG-induced CXCL10 and CCL2 expression and the associated p—NF–κB RelA and STAT3 levels. The authors further investigated the role of PERK in diseased photoreceptors by treating them with advanced glycation end products and high glucose and found that PERK is a positive regulator of CXCL10 and CCL2 expression, while XBP1 negatively regulates these chemokines (Zhu et al., 2017).

The role of ATF4 in the progression of IRD has also been elucidated. Elevated ATF4 expression has been found in retinal protein fractions isolated from different models of IRD (Bhootada et al., 2016; Starr et al., 2018). Thus, in T17M RHO transgenic retinas manifesting severe retinal degeneration, the overexpressed ATF4 turned out to have a pro-death role (Bhootada et al., 2016). Consequently, in these mice, we observed ATF4 overexpression concomitantly with an increase in CHOP and caspase-3/7 activity. In this particular study, we proposed that ATF4 possibly contributed to the mechanism of photoreceptor cell loss, since ATF4 knockdown in the T17M RHO retinas retarded retinal degeneration and promoted photoreceptor survival, as measured by scotopic and photopic ERGs and photoreceptor cell counting. The conclusion of the study was confirmed by experiments conducted in C57BL6 retinas overexpressing ATF4 by means of AAV delivery (Rana et al., 2014). Therefore, we concluded that future ADRP therapy that regulates ATF4 expression could be developed to treat retinal degenerative disorders associated with activated UPR.

In contrast, a protective role for ER stress-induced ATF4 has recently been proposed (Huang et al., 2021). In ER stress-manifesting cone-derived 661 W cells and mouse retinas of CH3 and C57BL6 mice, researchers found that treatment with 2, 3, 5, 6-tetramethylpyrazine (TMP) not only ameliorates retinal photoreceptor function loss and alleviates ER stress, but also enhances ATF4 expression. Further examination allowed the investigators to determine that the proportion of insoluble prion protein (PRP) versus soluble PRP was reduced both in vitro and in vivo. The intrinsic mechanism of the TMP therapeutic effect was proposed to be associated with the ATF4-mediated inhibition of PRP aggregation (Huang et al., 2021). In addition, the protective role of ATF4 in the induction of XBP1 expression has been recently highlighted, specifically in a study in which ATF4-mediated control of IRE/XBP1 pathway has been proposed as a novel mechanism (Tsuru et al., 2016).

The role of TRIB3, a downstream ATF4 mediator, in retinal degeneration has been studied to a lesser extent. In a retinal detachment (RD) model, it has been found that the number of TRIB3-positive photoreceptor cells was significantly induced after RD and peaked at 3 days post-RD. The knockdown of TRIB3 protects photoreceptors against ER stress-induced apoptosis. The authors concluded that TRIB3 may be a crucial molecule in photoreceptor apoptosis induced by ER stress (Yan et al., 2016). In a study conducted with rd16 mice, we investigated the role of TRIB3-mediated regulation of AKT/mTOR (Saltykova et al., 2021). We previously showed that the AKT/mTOR axis is inhibited in rd16 mice (Starr et al., 2018). Knowing that TRIB3 is a pseudokinase that inhibits AKT and mTOR, we genetically ablated TRIB3 in rd16 retinas, which resulted in preservation of photoreceptor function in degenerating retinas, associated with restoration of the p-AKT/p-mTOR activity and photoreceptor homeostasis in TRB3−/−, rd16 retinas. Based on these findings, we propose that TRIB3 may retard retinal degeneration and be a promising therapeutic target for treating retinal degenerative disorders (Saltykova et al., 2021).

CHOP is known to be a pro-apoptotic protein. Therefore, this fact served as a reason to test its role in retinal degeneration, which is known to manifest photoreceptor apoptotic cell death. In degenerating retinas, we and other investigators found that the ablation of CHOP in mice with IRD surprisingly resulted in no rescue of degenerating photoreceptors (Adekeye et al., 2014; Chiang et al., 2016; Nashine et al., 2013). The T17M RHO and P23H RHO mice did not obtain any benefits for the protection of vision loss from CHOP ablation in their retinas. Moreover, in T17M RHO CHOP−/− photoreceptors, we found a 22–24% decline in the thickness of the outer nuclear layer, which was associated with a 70% reduction in the a-wave ERG amplitude (Nashine et al., 2013). However, another study has shown that CHOP may regulate pathological responses, such as inflammation, that are upregulated during later stages of disease progression (Adekeye et al., 2014). The investigators found a regional protective effect from CHOP ablation in severely degenerated central retina in older P23H RHO mice. Indeed, our study with CHOP knockout mice manifesting Tn-induced ER stress demonstrated significantly lower IL-1b expression as compared to the C57BL6 retina overall, suggesting that CHOP controls this cytokine expression (Rana et al., 2014).

GADD34 is another pro-apoptotic protein known to be a subunit of a protein phosphatase 1-GADD34 complex, providing a feedback loop to dephosphorylate p-eif2α upon activation of PERK signaling. We recently studied its role in two animal models of IRD: rd16 and P23H RHO mice (Saltykova et al., 2022; Starr and Gorbatyuk, 2019a). These mice differ in terms of the rate of retinal degeneration (rapid vs. relatively slow), the affected proteins (CEP290 vs. RHO), and possibly even the molecular mechanisms of retinal pathogenesis. Despite these facts, UPR activation is a common signaling in their retinas. We found that the ablation of GADD34 exacerbated retinal degeneration in both models by increasing the number of apoptotic photoreceptor cells. Moreover, in P23H RHO retinas, GADD34 ablation caused a decline in the scotopic a-wave ERG amplitudes. In this particular study, we also found that, similar to CHOP, GADD34 controls Il-6 cytokine expression, and its ablation enhances the Tnfα mRNA expression in P23H RHO retinas, thus contributing to retinal pathogenesis (Saltykova et al., 2022). These studies have also revealed that future experiments should be conducted to better understand the roles of pro-apoptotic CHOP and GADD34 proteins in degenerating retinas.

6. ER stress signaling in cone photoreceptor disease and achromatopsia

6.1. ATF6 is essential for cone function in people

Achromatopsia is a heritable cone dysfunction disease characterized by loss of color vision, severely impaired visual acuity, photosensitivity, and nystagmus (Zobor et al., 2015). Mutations in cone phototransduction genes account for the majority of achromatopsia cases (Chang et al., 2009; Grau et al., 2011; Kohl et al., 1998, 2000, 2002, 2012), but a fraction of patients with clinical symptoms of achromatopsia lack mutations in cone phototransduction genes, suggesting additional disease genes. In 2015, genetic sequencing of these achromatopsia patients with intact cone phototransduction genes identified the most recent achromatopsia disease gene, ATF6 (Ansar et al., 2015; Kohl et al., 2015; Xu et al., 2015a). ATF6 achromatopsia disease variants carry single-nucleotide changes, small deletions, or duplications that introduce missense mutations, premature stop codons, or damaged splicing sites in ATF6 (Fig. 4). More recently, using advanced sequencing technologies, large multi-exon deletions in the ATF6 gene locus were identified in achromatopsia patients (Lee et al., 2020). By contrast to other achromatopsia disease genes, ATF6 is not part of the cone phototransduction system and its expression is not restricted to cones. Instead, ATF6 is a key regulator of ER stress signaling and protein homeostasis and is found in all mammalian cells.

Biochemical and computational studies with recombinant proteins, patient fibroblasts and stem cells demonstrated that ATF6 disease alleles were uniformly associated with loss of transcriptional function (Chiang et al., 2017; Kohl et al., 2015; Lee et al., 2020; Skorczyk-Werner et al., 2017). Interestingly, at least 3 different pathomechanisms underlying loss of transcriptional function were identified (Chiang et al., 2017). Class 1 ATF6 achromatopsia mutations prevent ATF6 from exiting the ER to Golgi apparatus, and, therefore, prevent generation of the cytosolic transcription factor fragment (Chiang et al., 2017; Skorczyk-Werner et al., 2017). Class 2 ATF6 mutations introduce premature stop codons that likely trigger nonsense-mediated decay, but for some variants, could also lead to generation of the cytosolic fragment (Chiang et al., 2017). In contrast, class 3 mutations directly damage the bZIP domain, deleting ATF6’s transcriptional activity (Chiang et al., 2017; Lee et al., 2020). Patient fibroblasts carrying these ATF6 achromatopsia variants all showed increased vulnerability to ER stress-induced damage and cell death in response to ER toxins, similar to findings with ATF6−/− MEFs (Chiang et al., 2017; Kroeger et al., 2018; Wu et al., 2007; Yamamoto et al., 2007). However, the different pathomechanisms that impair ATF6 transcriptional activity raise the possibility that there could be phenotype differences linked to different ATF6 achromatopsia alleles. Indeed, patients carrying the Class 1 ATF6 D564G variant reported rod dysfunction in addition to cone dysfunction, while patients with other ATF6 variants were limited to cone disease (Skorczyk-Werner et al., 2017).

6.2. Cones fail to develop outer segments in ATF6 mutant retinal organoids

Retinal organoids generated from achromatopsia patients’ iPSCs carrying ATF6 variants have shed insight into cellular and molecular pathomechanisms underlying cone dysfunction (Kroeger et al., 2021; Lee et al., 2022). The pace of differentiation and morphology/size of retinal organoids showed no differences in ATF6 defective retinal organoids compared to controls. But, microscopic inspection of ATF6 mutant organoids revealed a “smooth” surface throughout differentiation whereas control retinal organoids developed bulbous projections consistent with formation and extension of cone inner/outer segment (Kroeger et al., 2021). Confocal immunohistochemical examination with cone markers, peanut agglutin and red/green cone opsin, confirmed severe defects in cone IS/OS extension in ATF6 mutant retinal organoids (Kroeger et al., 2021). Sequencing of ATF6 mutant retinal organoids also identified a significant loss of cone phototransduction apparatus genes and pathways. By contrast, cell death/apoptosis was not a significant process in ATF6 mutant organoids. Failure in cone IS/OS extension was observed using multiple different ATF6 disease variant iPSCs (Kroeger et al., 2021). Furthermore, isogenic gene-edited ATF6−/− hESCs also generated retinal organoids with the same morphologic defect (Kroeger et al., 2021). Together, these findings identify a robust sub-cellular phenotype in developing cones on retinal organoids that account for photopic vision loss: defective cone IS/OS formation and extension. These retinal organoid findings are also congruent with adaptive optics imaging of ATF6 patients’ fovea showing an absence of cone IS/OS (Kroeger et al., 2021; Mastey et al., 2019).

ATF6 mutant retinal organoids also revealed additional subcellular and genetic defects besides cone IS/OS mal-development that may also contribute to cone dysfunction in people. Ultrastructural and transcriptomic analysis revealed extensive mitochondria damage (malformed cristae, enlarged/dilated mitochondria) coupled with significant induction of oxidative phosphorylation genes (Lee et al., 2022). These mitochondria defects likely impact rods and cones because they were widely found in superficial ultrastructural sections of ATF6 mutant retinal organoids (Lee et al., 2022). Also, Müller glia showed activated gene signatures in transcriptomic data from ATF6 mutant retinal organoids (Lee et al., 2022). Since ATF6 is expressed in all retinal cells (Lee et al., 2020), these mitochondria defects and Müller glia activation could be a direct consequence of ATF6 dysfunction. These defects may also reflect secondary pathology due to primary dysfunction of the developing cones in the ATF6 mutant retinal organoids.

6.3. ATF6 is dispensable for cone development but protects retina from aging and proteotoxicity in mice

While clinical and retinal organoid findings show that ATF6 is essential for cone development in people, ATF6 is dispensable for cone development in mice (Kohl et al., 2015; Lee et al., 2021). ATF6−/− mice carry exon deletions that prevent generation of ATF6 protein (Wu et al., 2007), similar to patient ATF6 disease multi-exon deletion alleles (Lee et al., 2020). However, by contrast to the congenital photopic cone dysfunction seen in people, at all mouse ages <18 months, no defects in retinal lamina thickness, cone photoreceptor numbers by PNA staining, and cone opsin protein expression were found in ATF6−/− mice (Kohl et al., 2015). Consistent with this, no defects in ERG responses (photopic and scotopic) were observed in ATF6−/− mice. However, in aged mice (>18 months), photopic and scotopic ERG responses declined in ATF6−/− mice along with retinal thinning by histology (Kohl et al., 2015). Interestingly, when crossed with P23H rhodopsin knock-in mice, earlier retinal dysfunction accompanied by reduced rhodopsin protein turnover was seen (Lee et al., 2021). Our study reveals that the loss of ATF6 leads to a significant accumulation of P23H rhodopsin. This accumulation is accompanied by hyperactivation of the IRE1 pathway, as a compensatory response to the absence of ATF6. However, as P23H rhodopsin mice age, we find that loss of ATF6 accelerates retinal degeneration (Lee et al., 2021). Thus, our findings provide direct in vivo evidence that ATF6 is important in rod photoreceptors to maintain rhodopsin protein quality in a mouse model of retinitis pigmentosa, and its absence worsens retinal degeneration. These findings indicate that ATF6 does protect murine retina from aging and genetic ER/oxidative stressors like the misfolded P23H rhodopsin protein. These data support that ATF6 supports retinal homeostasis – especially protection from ER/oxidative stress and misfolded proteins - throughout life. In people, ATF6 supports another essential function – promoting cone subcellular structural formation during retinal development.

6.4. ER stress and cone disease arising from CNGA3, CNGB3, and cone opsin mutations

In addition to ATF6’s direct role in the pathogenesis and progression of cone disease in people, ER stress and UPR activation are also implicated in many other human cone diseases linked to misfolded ER client proteins (secreted and transmembrane proteins). For instance, hundreds of mutations in the CNGA3 and CNGB3 transmembrane channel proteins lead to cone dysfunction and achromatopsia (Michalakis et al., 2022). Missense mutations in the S-opsin transmembrane protein cause S-cone dysfunction and tritanopia (Weitz et al., 1992). These and many other mutations linked to cone disease damage protein structure leading to retention of the misfolded receptor or channel protein in the endoplasmic reticulum. ER protein misfolding elicits ER stress and triggers UPR activation (Walter and Ron, 2011). However, the consequences of ER stress and UPR activation in these cones carrying misfolded proteins remain to be determined.

7. ER stress signaling in diabetic retinopathy and angiogenesis

As one of the most common complications of diabetes, diabetic retinopathy (DR) poses a significant risk for the development of visual impairment or even severe blindness in patients. Approximately one-third of diabetic patients experience some degree of DR, and around 10% of affected individuals progress to more advanced forms, including diabetic macular edema (DME) and proliferative DR (PDR) (Yau et al., 2012). Recent investigations suggest that DR is a neurovascular disorder characterized by neural retina deterioration and accompanying microvascular abnormalities (Simo et al., 2018). The molecular mechanisms underlying DR are complex, and the specific pathophysiological processes driving the progression from diabetes mellitus to DR are not yet fully understood. Various factors have been linked to the development of DR, including abnormal glucose and lipid metabolism, oxidative stress, inflammatory cytokine exudation, and autophagy. Moreover, all these factors are known to regulate the ER stress response.

7.1. Activation of the PERK signaling in DR

Studies on diabetic mice and rats have revealed that persistent increase or fluctuations in hyperglycemia result in the upregulation of the PERK/eIF2α/ATF4/CHOP signaling in retinal cells ((Chen et al., 2012; Zhong et al., 2012a)Kong et al., 2018; Li et al., 2009c; Ma et al., 2017; Ma et al., 2014; Zhang et al., 2014). Meanwhile, the damage to retinal cells is primarily associated with dysregulation of the PERK/CHOP and IRE1 pathways, which may exert opposite effects (Elmasry et al., 2018; Kong et al., 2018; Li et al., 2011, 2014a; McLaughlin et al., 2019; Yang et al., 2019). It has been proposed that dyslipidemia, a biomarker of diabetes, and lipid oxidation activate UPR signaling in diabetic retinas (Fu et al., 2014). Indeed, we and other groups reported increased UPR activation in the retina in db/db mice, a type 2 diabetes model (Ma et al., 2017; Tang et al., 2011), and in human retinas from donors with type 2 diabetes (Du et al., 2013). In pancreatic β cells, excessive accumulation of free fatty acids (FFAs) observed in the serum of type 2 diabetic patients triggers apoptosis through activation of the PERK/eIF2α/ATF4/CHOP pathway while suppressing both the IRE1 and ATF6 pathways (Cnop et al., 2007). These findings on apoptosis and β-cell dysfunction were confirmed by another study on diabetic liver cells conducted by Cao and colleagues (Cao et al., 2012). In the diabetic retina, lipid peroxidation can alter the phospholipid composition and fluidity of the ER membrane, leading to ER stress, accumulation of fatty acid metabolites, increased mitochondrial beta-oxidation, and subsequent stimulation of ROS production. This could in turn activate the UPR and trigger apoptosis. Understanding the role of ER stress signaling in lipid peroxidation induced retinal cell death holds great promise in identifying new pathways in the pathogenesis of DR.

In the early stage of DR, circulating inflammatory cytokines, such as VEGF, TNFα, and IL-1β, damage the retinal microvessels in diabetic eyes. For example, TNF-α, IL-1β and/or IL-6 have been shown to induce ER stress in hepatocytes (Zhang et al., 2006) and the T-cell-derived cytokine interferon-γ (IFN-γ) activates PERK and PERK-associated apoptosis in oligodendrocytes (Lin et al., 2005). The mechanism underlying this activation is likely associated with the release of calcium from the ER and the generation of ER stress. All these events can occur in cells when metabolic factors, such as cholesterol, nonesterified fatty acids, glucose, and homocysteine, induce the ER stress response and the inflammatory response simultaneously (Zhang and Kaufman, 2008). Regarding the latter, the inflammatory response could be directly triggered by UPR signaling. Indeed, in the last decade, our labs have shown that activated UPR causes cytokine and chemokine overproduction in the retina (Chen et al., 2012; Huang et al., 2015c; Li et al., 2009b; Rana et al., 2014; Wang et al., 2013; Zhong et al., 2012c). In one study, we found that IL-1β and IL-6 expression was responsive to the treatment of C57BL6 retinas with tunicamycin, a UPR inducer, or subretinal injection with AAV2/5 overexpressing ATF4, in C57BL6 mice (Rana et al., 2014). In our other studies, we demonstrated that treatment of C57BL6 mice with periocular injections of tunicamycin-induced retinal TNFα and VEGF expression (Li et al., 2009b) and moreover, intravitreal injection of adenovirus overexpressing ATF4 significantly increase retinal levels of monocyte chemoattractant protein 1 (MCP1) and inflammatory cell infiltration (Huang et al., 2015c). We further demonstrated that ATF4 and its downstream network are directly responsible for the regulation of pro-inflammatory cytokine gene expression. This series of experiments confirmed that ATF4 is the transcriptional factor controlling the expression of both cytokines and chemokines.

7.2. The PERK signaling in regulation of vascular function and angiogenesis

We previously reported high glucose treatment induces the activation of UPR with enhanced GRP78 expression, PERK and eIF2a phosphorylation, and ATF4 expression in cultured human retinal microvascular endothelial cells (RMECs). In human pericytes, fluctuations in glucose concentration, but not constantly high glucose, increases ATF4 and CHOP expression with a concomitant elevation of MCP-1 production (Zhong et al., 2012c). Treatment of the cells chemical chaperone TUDCA suppresses MCP-1 secretion, suggesting a role of ER stress-mediated ATF4/CHOP activation in MCP-1 production. In agreement with these findings, it was reported that advanced glycation end products (AGE) and modified low-density lipoprotein activate the PERK pathway in human retinal pericytes, which was ameliorated by chemical chaperone UDCA (Chung et al., 2017). Our in vivo studies support a role of the activated PERK signaling in retinal angiogenesis, a hallmark pathological feature of PDR. We found that retinal levels of p-PERK, p-eIF2α, ATF4, and CHOP are significantly increased in mice with oxygen-induced retinopathy (OIR) (Chen et al., 2012; Li et al., 2009c; Wang et al., 2013). Heterozygous knockout of ATF4 or inhibition of ATF4 function decreased VEGF expression, reduced retinal inflammation, and ameliorated BRB breakdown and vascular leakage in diabetic mice (Chen et al., 2012; Zhong et al., 2012a). We further showed that ATF4 deficiency reduced the rate of retinal neovascularization and angiogenic gene expression (Flt1, Vegf1, Hif1, and Tgb1) in OIR mice (Wang et al., 2013). Mechanistically, we found that overexpression of ATF4 enhanced, while inhibition of ATF4 attenuated, the basal and LPS-stimulated phosphorylation of NF-κB, P38, and JNK and pharmacological inhibition of NF-κB, P38, or JNK significantly reduced ATF4-stimulated MCP-1 secretion in endothelial cells (Huang et al., 2015c). Recently, we expanded the search of ATF4 downstream targets regulating the pericyte and endothelial cell loss in DR and found that TRIB3 activation is responsible for the increase in acellular capillary area and the decrease in pericytes in the retina in a mouse model of type 1 diabetes (Pitale et al., 2021). The retinal neovascular area was also dramatically reduced in TRIB3^−/− mice with OIR. We proposed that the mechanism underlying TRIB3-mediated vascular cell control is most likely linked to the control of cytokine and pro-inflammatory gene expression. Taken together, the results from our studies over the past decade suggest a critical role of PERK signaling in the regulation of vascular endothelial and pericyte damage and pathological angiogenesis in DR.

7.3. The PERK signaling in retinal ganglion cell homeostasis in DR

Mounting evidence suggests that the degeneration of retinal neurons can occur before clinical manifestation as typical microvascular alterations in DR and retinal cell death is often first detected RGCs (Simo et al., 2022). Oshotari and colleagues reported that RGC apoptosis in diabetic retinas exposed to high levels of glucose in vitro correlated with the activation of the PERK pathway and increased levels of CHOP (Oshitari et al., 2011). In STZ-induced diabetic rats, activation of PERK UPR mediators (e.g., CHOP), JNK, and caspase 12 protein were found increased and associated with apoptosis (Yang et al., 2013). In a study using a transformed mouse RGC cell line, Zhang and colleagues reported that RGC apoptosis occurs due to activation of ATF4 and CHOP mediators of PERK signaling under conditions mimicking diabetes (Zhang et al., 2018). Finally, our lab demonstrated that human and mouse diabetic RGCs overexpress TRIB3 protein and that the retinal ablation of ATF4-regulated TRIB3 proteins increases RGC survival in the diabetic retina (Pitale et al., 2021). Intriguingly, a recent study shows that increased ER stress in retinal neurons, especially RGCs, in OIR suppresses reparative angiogenesis through reducing neuron-derived angiogenic guidance cue neutrin-1 (Binet et al., 2013). Thus, targeting neuronal ER stress may also improve retinal vascular function in ischemic retinopathy including DR.

7.4. The PERK signaling in glia activation in DR

There are three types of glial cells, including astrocytes, microglia, and Müller cells, in the retina. These cells, in particular Müller cells, are highly responsive to metabolic changes, injuries, or stressors that may cause neuronal and vascular dysfunction [reviewed in (Coughlin et al., 2017; Kelly et al., 2018)]. Increased activation of Müller glial cells or reactive gliosis has been reported in diabetic retinas (Coughlin et al., 2017; Simo et al., 2022). In our previous studies, we investigated the role of the PERK signaling in Müller glial activation in conditions mimicking diabetes (Zhong et al., 2012a). We have shown increased expression of p-eIF2α, ATF4, and CHOP in retinal Müller cells in STZ-diabetic mice. In cultured Müller cells, high glucose is sufficient to induce ER stress and PERK pathway activation and inhibition of ER stress inhibits high glucose-induced ATF4 and CHOP upregulation. Further, we demonstrated that ATF4 interacts with hypoxia inducible factor 1α (HIF-1α) and JNK pathways, playing a critical role in the regulation of inflammatory factor expression and secretion from Müller cells. We also showed that ATF4 is required for VEGF secretion from Müller cells in diabetic conditions (Zhong et al., 2012a). These findings collectively indicate an important role of the PERK pathway in Müller glial activation in DR. Relative to Müller cells, the role of the PERK signaling in microglia and astrocytes is less well explored in the diabetic retina. In one study, Wang and associates showed increased ATF4 expression in astrocytes of optic nerves but not in the retina of STZ-diabetic rats (Wang et al., 2020). In another study, Lind et al. examined the UPR activation in cultured astrocytes and rat brain during experimental diabetes (Lind et al., 2013). They found no alterations in p-eIF2α, ATF4, and CHOP in astrocytes with extended high glucose treatment for up to 4 weeks or in STZ-treated rats with up to 7 months of diabetes. The authors concluded that in sharp contrast to retinal Muller cells and diabetic retina, the astrocytes and STZ-diabetic brain are relatively resistant to diabetes-induced ER stress. Future research should elucidate if the PERK and other UPR pathways regulate the function of astrocytes and microglia in DR.

7.5. Activation of the IRE1/XBP1 signaling in DR

Like the PERK signaling but to a less extent, studies have shown that the IRE1/XBP1 pathway is activated in diabetic retinal tissue and actively involved in the regulation of oxidative stress, inflammation, and energy metabolism in retinal cells (Li et al., 2011; McLaughlin et al., 2018, 2019; Yang et al., 2019). Activation of the IRE1/XBP1 pathway, demonstrated by increased levels of p-IRE1a and spliced XBP1, has been observed in the retina of Akita mice and STZ-induced diabetic mice, both of which are common animal models of type 1 diabetes, and db/db mice, a model of type 2 diabetes (Li et al., 2011; Ma et al., 2014, 2017). In cultured human retinal microvascular endothelial cells and Müller cells, the IRE1/XBP1 pathway can be activated by hypoxia and hyperglycemia, two major insults pertinent to the pathogenesis of DR (Li et al., 2011; Yang et al., 2019). Exposure of retinal endothelial cells to hypoxia or inflammatory cytokine TNF-α increases the expression of adhesion molecules ICAM-1 and VCAM-1, resulting in enhanced endothelial inflammation, tight junction damage, and vascular leakage. Using transcriptomic analysis, a recent study identified that XBP1 is downregulated in retinal pericytes isolated from 3-month-diabetic animals (Rangasamy et al., 2020). The exact role of XBP1 in pericyte dysfunction and pericyte loss during diabetes remains to be investigated.

7.6. The IRE/XBP1 signaling in retinal inflammation

Activation of the IRE1/XBP1 pathway has been implicated in regulation of retinal inflammation. In an earlier study, we demonstrate that preconditioning with mild ER stress induces a transient activation of the IRE1/XBP1 pathway resulting in increased expression of XBP1s in retinal endothelial cells. ER stress preconditioning protects the cells from TNF-a-induced endothelial inflammation and retinal vascular leakage and this protective effect is mediated by XBP1 (Li et al., 2011). We showed that overexpression of spliced XBP1 negatively regulates the activation of IRE1α and inhibits NF-kB-mediated ICAM-1 and VCAM-1 expression, suggesting a protective role of spliced XBP1 against endothelial inflammation (Li et al., 2011). Interestingly, a recent study using next-generation sequencing and bioinformatic analysis of XBP1-binding motifs identified XBP1 as a repressor of IRE1 mRNA expression during the UPR in HeLa cells (Gebert et al., 2021). The negative regulation of the IRE1 activity by XBP1s needs further in-depth study in retinal cells.

In a recent study, we determine the role of XBP1 in regulation of Müller cell activation and cytokine production. We found that conditional knockout of the XBP1 gene in Müller cells leads to a substantial increase in VEGF and TNF-a production resulting in enhanced retinal inflammation and vascular leakage in diabetic animals (Yang et al., 2019). Consist with these findings, as discussed earlier, inhibition of XBP1 in 661w cells leads to increased inflammatory cytokine production while deletion of PERK reduces ER stress-induced inflammation, suggesting a differential role of the XBP1 and PERK pathways in photoreceptor-derived inflammatory factor production (Zhu et al., 2017).

7.7. The IRE/XBP1 signaling in retinal neurodegeneration in DR

Neuronal dysfunction and degeneration are considered integral components in DR pathogenesis, ultimately resulting in vision impairment and blindness. To determine the role of XBP1 in retinal neuronal survival and function, we generated retina-specific conditional XBP1 knockout mice. We found that XBP1 deficiency does not affect retinal development but causes a significant reduction in retinal function and degeneration of retinal neurons with aging (McLaughlin et al., 2018). These findings are consistent with previous observations in Drosophila that mutations of IRE1, but not mutations of XBP1, cause defective rhodopsin delivery in photoreceptor development and degeneration of rhabdomeres (Coelho et al., 2013). The latter is the light-sensing organelle in drosophila functionally equivalent to the photoreceptor OS in vertebrates. Furthermore, IRE1 is required for rough ER differentiation and expansion in developing drosophila photoreceptors (Xu et al., 2016). In contrast, mice with selective deletion of the IRE1 gene in photoreceptors show normal photoreceptor development but lead to reduced outer nuclear layer thickness after 6 months of age (Massoudi et al., 2023). The results from these studies strongly suggest that the IRE1/XBP1 pathway is dispensable for photoreceptor/retinal neuronal development but is important for maintaining ER homeostasis, cellular survival, and function in retinal neurons.

In a later study, we determined the effect of XBP1 deficiency on diabetes-induced retinal dysfunction and neurodegeneration. We found that loss of XBP1 increases the susceptibility of retinal neurons to diabetic metabolic stress resulting in accelerated photoreceptor degeneration, RGC loss, and increased glial activation (McLaughlin et al., 2019). The number of photoreceptor ribbon synapses was also significantly reduced in XBP1 knockout mice during diabetes, suggesting a role of XBP1 in regulation of photoreceptor synaptic integrity (McLaughlin et al., 2019)(McLaughlin et al., 2023). In line with these observations, Massoudi et al. showed that loss of IRE1 in photoreceptors leads to accelerated photoreceptor degeneration in a RhoP23H mutation-induced RP model (Massoudi et al., 2023). These findings support an important role of the IRE1/XBP1 pathway in maintaining retinal neuronal function and structural integrity during normal and stress conditions. Further discussion can be found in our recent review article (McLaughlin et al., 2022). More recently, UPR dysregulation was identified during 1-deoxyspingolipid induced toxicity to retinal organoids, and this finding may shed insight into macular telangiectasia (Rosarda et al., 2023).

8. The ER-mitochondria crosstalk

The ER is involved in the regulation of multiple vital cellular processes, namely protein synthesis and folding, lipid biosynthesis, and calcium metabolism (Schwarz and Blower, 2016). To achieve homeostasis within such a functionally multifaceted organelle, the ER acts in part with the mitochondria through a protein-and-membrane complex called the mitochondria-associated ER membrane (MAM) to orchestrate many of the signaling pathways involved in ER function (Fig. 5) (Hayashi et al., 2009). Before the advent of advanced biochemical techniques, the MAM was first described in 1958 b y Copeland and Dalton as a unique association between the ER and mitochondria after examining pseudobranch glands of teleosts (Copeland and Dalton, 1959). It was not until 1990 that it was highly purified from rat liver by Vance as “Fraction X,” which was further described as a mitochondrial membrane fraction that was suspected to take part in the transfer of lipids between the ER and mitochondria (Vance, 1990). The most recent liquid chromatography (LC)-mass spectrometry (MS)-based proteomic analyses in rat retinas have now revealed the MAM to be a scaffold consisting of as many as >2660 proteins that serve as the physical interface for biochemical crosstalk between the ER and mitochondria (Wang et al., 2022b). A smaller and simpler structure analogous to the mammalian cell MAM has also been discovered in yeast as well, termed the ER-mitochondria encounter structure (ERMES) (Kundu and Pasrija, 2020). Similar to the MAM in mammalian cells, the ERMES has also been deemed crucial for lipid exchange between the ER and mitochondria (Kawano et al., 2018). Evidence of such ER-mitochondrial associations across eukaryotic cells of varying complexities suggests that these associations are basic and vital to normal cell function as well as to the maintenance of both ER and mitochondrial homeostasis.

Fig. 5.

The mitochondria-associated ER membrane (MAM). A). Schematic diagram of MAM structure and function. MAM is formed through several pairs of tethering proteins localized to the outer mitochondrial membrane and ER membrane, including Mfn1/2-Mfn2, FPTPIP51-VABP, VDAC-GRP75-IP3R, PTPIP51-MOSPD2, and others. The VDAC-GRP75-IP3R is responsible for Ca2+ trafficking from the ER to the mitochondria. Sigma-1R can bind to IP3R and regulate calcium transfer to the mitochondria. B). Electron microscopic image shows a close contact of mitochondria with the ER in the RPE of a wild-type mouse. Double-headed arrow denotes the distance between the ER and the mitochondria.

Structurally, the MAM tethers the ER to the mitochondria so that they may crosstalk at an approximate distance of 10–25 nm, though this range may vary based on cell condition and spatial occupancy by ER ribosomes (Csordas et al., 2006; Wang et al., 2021a). While some tethering proteins solely act to physically link the ER to the mitochondria (e. g., the MOSPD2-PTPIP51 complex (Di Mattia et al., 2018)), other tethering proteins (e.g., the IP3Rs-Grp75-VDACs complex (Tubbs et al., 2014), VAPB-PTPIP51 complex (Qiao et al., 2017; Stoica et al., 2014), and REEP1 (Lim et al., 2015)) may play a more dynamic role in maintaining and/or adjusting the morphology of the ER and mitochondrial membrane surfaces (Wang et al., 2021a). According to electron microscopy studies, MAM tethering proteins tend to connect the ER to the mitochondria at (1) a single site that covers roughly 10% of the outer mitochondrial membrane (OMM) surface, (2) at multiple sites covering 50% of the OMM, or (3) at nearly all sites along the OMM covering nearly 100% of the mitochondrial surface area (Fujimoto and Hayashi, 2011).

Largely a function of structure, the MAM coordinates numerous processes such as calcium regulation, mitochondrial remodeling, inflammation, reactive oxygen species (ROS) production, apoptosis, and lipid transfer based on the composition of its proteome located within and between the ER and OMM (Fig. 5) (Lee and Min, 2018). For example, PERK is an important ER stress sensor protein that is heavily embedded within the MAM. The loss of PERK in MAM has been shown to not only weaken the ER-mitochondrial tethering by the MAM but also disrupt ER morphology, impair calcium signaling, and dysregulate the ER-mitochondrial crosstalk necessary for ROS-induced cell death (Verfaillie et al., 2012). The MAM interface also serves as an important site for inflammasome formation as it houses NOD-like receptor protein 3 (NLRP3) in its activated state (Missiroli et al., 2018; Zhou et al., 2011). It is even involved in the antiviral response via Gp78, an E3 ubiquitin ligase involved in the ERAD pathway that also localizes to the MAM (Jacobs et al., 2014). Hence, the MAM’s diverse proteomic composition allows it to serve as a key crossroad site for a variety of ER-mitochondrial signaling.

Given the MAM’s intricate involvement in key signaling pathways contributing to ER and cellular homeostasis, disturbances in the MAM proteome can lead to impairment of ER-mitochondrial communication, further resulting in calcium dysregulation, aberrant metabolism, mitochondrial damage, and oxidative stress. Such consequences have been linked to the pathogeneses of diseases including but not limited to insulin resistance (Cheng et al., 2020a; Tubbs et al., 2014), atherosclerosis (Wang et al., 2021b), cancer (Yang et al., 2023a), and neurodegenerative diseases such as Alzheimer’s disease, Parkinson’s disease, and amyotrophic lateral sclerosis (Hedskog et al., 2013; Liu and Yang, 2022). Although more studies are needed to elucidate the exact scale of the MAM’s impact on regulating systemic health, growing evidence has made it clear that there is indeed an association between MAM dysfunction and disease on a macro level. As such, the MAM proteome may offer additional avenues for the study of potential therapeutic targets.

8.1. MAM and calcium regulation

The MAM plays an indispensable role in regulating calcium signaling between the ER and mitochondria. The ER is the primary site of intracellular calcium storage. In the healthy state, a low constitutive level of calcium is transmitted from the ER to mitochondria to regulate essential cellular processes such as apoptosis, ATP production, and metabolism (Marchi et al., 2018). Numerous MAM proteins involved in the direct supply of calcium from the ER to mitochondria have been discovered, the most well-studied of which has been the inositol 1,4,5-trisphosphate receptor (IP3R)-75 kDa glucose-regulated protein (GRP75)-voltage-dependent anion channel (VDAC) MAM tethering complex (Bononi et al., 2012). IP3Rs located on the ER membrane are physically attached to GRP75 within the MAM and facilitate the release of ER calcium into the cytoplasm. Once released, the cytoplasmic calcium is shuttled into the mitochondrial intermembrane space (MIMS) via VDAC porins. Impaired calcium signaling is critically involved in the process of cell death, as calcium overload within the mitochondrial matrix disrupts normal metabolism and triggers apoptosis (Giorgi et al., 2012; Zhivotovsky and Orrenius, 2011). Interestingly, the distance and surface area at which the MAM tethers the ER to the mitochondria are not trivial in the context of calcium signaling and cell death; in one study, large ER-mitochondrial contacts as created by the MAM (i.e. contacts covering 20% of the mitochondrial surface) have been associated with higher concentrations of IP3R-mediated calcium release and greater saturation of calcium binding to aequorin chimeras in the MIMS (Rizzuto et al., 1998). Another study found that mitochondria were exposed to much higher levels of calcium when in close contact with the ER than with the rest of the cytoplasm, lending significance to the fact that the physical ER-mitochondria interface plays an integral role in calcium homeostasis (Rizzuto et al., 1993).

Several studies have sought to investigate other protein targets associated with the IP3R-GRP75-VDAC complex in the MAM that, when impaired, may also influence normal ER calcium trafficking. One such protein is mammalian target of rapamycin complex 2 (mTORC2), a kinase involved in pathways of cell proliferation, growth, and metabolism that also phosphorylates kinases linked to the pathogenesis of cancer and diabetes (Oh and Jacinto, 2011). One study found that mTORC2 localizes to the MAM in a growth factor-dependent manner and modulates calcium flux through Akt-dependent IP3R phosphorylation (Betz et al., 2013). The same study also showed that an mTORC2 knockout significantly compromises MAM integrity by reducing the number of ER-mitochondrial contact sites by about 40%. Cyclophilin D (CYPD) is another protein closely linked to the IP3R-GRP75-VDAC complex that, when downregulated in mice, similarly disrupts MAM integrity and blunts IP3R-mediated calcium transfer, which further induces hepatic insulin resistance downstream (Rieusset et al., 2016). The IP3R complex in the MAM additionally interacts with phosphatase and tensin homolog deleted on chromosome 10 (PTEN), a tumor suppressor protein that also localizes to the signaling domains in the MAM involved in ER-mitochondria calcium transfers (Bononi et al., 2013). When silenced, PTEN was found to impair calcium release from the ER and diminish cellular sensitivity to apoptosis. Through proteins associated with the IP3R-GRP75-VDAC trimeric complex alone, the MAM offers multiple insights into the mechanism of calcium regulation between the ER and mitochondria.

8.2. MAM and mitochondrial remodeling

Mitochondrial remodeling refers to the dynamic events that shape and alter mitochondrial morphology, including but not limited to mitochondrial fusion and fission, mitophagy, and mitochondrial biogenesis (Gottlieb and Bernstein, 2016). Proper mitochondrial remodeling is crucial to normal cell function as studies have shown that inhibition of fission, fusion, and mitophagy events leads to the accumulation of dysfunctional mitochondria and macro-level consequences. In mouse hearts, for example, cardiomyocyte-specific knockout of Dynamin-related protein (Drp)-1, which is an inducer of mitochondrial fission, resulted in a generalized loss of mitochondria and full-blown heart failure after six to seven weeks (Song et al., 2015). Other prominent proteins in mitochondrial remodeling include mitofusin (Mfn), a mitochondrial fusion factor attached to the OMM, Opa1, a dynamin-related GTPase in the MIMS that controls fusion and fission events, Mff, an OMM protein that mediates fission, and mitochondrial elongation factor 1 (Mief1), also another fusion factor (Gottlieb and Bernstein, 2016).

Though the mechanisms are not always clear, the MAM has been shown to house a few of the above proteins and serve as a key entity in regulating mitochondrial remodeling dynamics. In mouse embryonic fibroblasts, the loss of MFN2 was associated with a MAM deficiency (Hu et al., 2021b). Moreover, during mitochondrial fission induced by energy stress, the cellular energy sensor AMP-activated protein kinase (AMPK) translocated from the cytosol to the MAM and mitochondria, highlighting the MAM’s role in mediating energy stress-induced fission. In retinal precursor cells, an Mfn2 knockdown exacerbated MAM dysregulation and mitochondrial dysfunction under both normal and oxidative stress conditions, again highlighting the role of Mfn2 in mitochondrial maintenance (Yang et al., 2023b). During apoptosis, DRP1 is recruited to the MAM by C1q/TNF-related protein 1 (CTRP1), which itself partly localizes to the MAM based on electron microscopy studies (Sonn et al., 2021). Podocytes exhibit an abundance of MAMs when undergoing excessive mitochondrial fission mediated by A-kinase anchoring protein 1 (AKAP1)-DRP1 signaling in diabetic nephropathy (Li et al., 2023). In another study, levels of DRP1 in the MAM were also found to increase in response to hypoxia, which can trigger mitophagy (Wu et al., 2016).

Several other MAM proteins involved in mitochondrial fission, mitophagy, and biogenesis have also been discovered. siRNA silencing of FUN14 domain containing1 (FUNDC1), a novel MAM protein that functions as a mitophagy receptor, resulted in mitochondrial elongation and prevention of mitophagy in HeLa cells under hypoxic conditions (Wu et al., 2016). Phosphofurin acidic cluster sorting protein 2 (PACS2) is an important regulator of MAM formation that has been studied in human kidney 2 (HK-2) cells (Li et al., 2022). In addition to restoring MAM integrity, PACS2 was also suggested to play a role in decreasing high glucose-induced mitochondrial fission by inhibiting the recruitment of DRP1. The MAM may also participate in mitochondrial biogenesis via the transport of ER cholesterol to mitochondria; one study reported the presence of dihydroceramide desaturase Δ4-dihydroceramide desaturase 1 (DEGS1) not only in the ER but also in the MAM (Planas-Serra et al., 2023). By inducing a DEGS1 deficiency in patient fibroblasts, the authors observed aberrant mitochondrial morphology and consequent defects in oxidative phosphorylation, disruption of lipid droplet biogenesis, and abnormal lipid and phospholipid metabolism. Accumulating evidence thus suggests that MAM-associated proteins, both those that are resident in the MAM and others that may regulate the MAM remotely, are involved in the maintenance and life cycle of mitochondria. Further studies are needed to explore the exact extent to which other remodeling proteins localize to and interact with the MAM during remodeling events.

8.3. MAM and inflammation

Inflammation is an overarching term that refers to a host’s protective immune response mounted against pathogenic microbes on both a systemic and intracellular level. On an intracellular level, the activation of inflammasomes as the gateway to systemic inflammation has been of great interest, especially because it reveals further information regarding the mechanistic origins of inflammation. Additionally, inflammasomes have been linked to numerous ailments—neurodegenerative disorders, autoimmune diseases, cancers, and metabolic disorders—and may therefore offer possibilities for therapeutic targets if understood thoroughly (Heneka et al., 2018; Sharma and Kanneganti, 2021; Wilson and Cassel, 2010). The inflammasome itself is a protein complex consisting of a sensor protein, an adaptor called apoptosis-associated speck-like protein (ASC), and a zymogen procaspase-1 protein. The most well-studied subfamilies of inflammasome, differentiated by their sensor proteins, include NLRP3, NLRP1, NLRC4, and AIM2. Given that mitochondria are a major source of ROS, which can trigger the activation of the NLRP3 inflammasome (Latz et al., 2013), it is no surprise that MAMs have been found to participate in the inflammatory process via inflammasome activation (Zhou et al., 2011). To date, the only inflammasome known to interact with the MAM is NLRP3 (Missiroli et al., 2018), which localizes to the MAM with its adaptor ASC in the presence of NLRP3 inflammasome activators (Zhou et al., 2011).

There are several stress pathways through which the MAM may trigger inflammasome activation, such as hypoxia, mitochondrial fission, autophagy, and calcium overload (Thoudam et al., 2016). Hypoxia is known to cause mitochondrial dysfunction by depriving the mitochondria of oxygen needed for ATP generation, ultimately leading to superoxide formation, oxidative stress, and eventual mitochondrial damage (Jassim et al., 2021). The damaged mitochondrial components can then be recognized as damage-associated molecular patterns (DAMPs) by inflammasome components, instigating inflammasome complex formation (Thoudam et al., 2016). Interestingly, in some cases hypoxia and mitochondrial turnover via mitophagy have actually provided a context for the prevention of inflammation; one study found that the mitophagy receptor FUNDC1 works to inhibit NLRP3 inflammasome activation by promoting mitophagy in the setting of hypoxia (Zheng et al., 2021). Mitochondrial fission has been associated with inflammation as well in lipopolysaccharide (LPS)-stimulated microglial cells (Park et al., 2013); LPS was found to induce the translocation of Drp1 from the cytosol to the MAM, where it would not only promote fission by cleaving mitochondria but also coincide with the expression of pro-inflammatory cytokines. Overload of mitochondrial matrix calcium caused by abnormal ER-mitochondrial calcium transfer by the MAM is another cause of mitochondrial ROS generation, leading to mitochondrial collapse and apoptosis. Here again, the mitochondrial contents released during apoptosis are detected as inflammasome-inducing DAMPs, triggering inflammation (Thoudam et al., 2016). Given the association between inflammasome activation and various diseases, it is evident that many studies have sought to identify targets in MAM-associated inflammation to discover avenues for therapy. Though inflammasome formation can be harmful with overactivation, future research in therapeutics should always be mindful of the potential negative consequences of quelling what is otherwise a necessary step in the development of the inflammatory response against pathogens.

8.4. MAM in diabetic retinopathy

The study of the MAM’s role in the pathogenesis of DR is a relatively new field, but one that warrants exploration. DR is a common and serious but preventable complication of diabetes that affects over a third of diabetic adults worldwide (Lee et al., 2015). Though DR is often understood to be a primarily microvascular disease, growing evidence supports the notion that it may be caused first by neurodegenerative changes and retinal inflammation prior to the manifestation of vascular changes (Joltikov et al., 2018; Lee et al., 2018; Simo et al., 2018). As disturbances in MAM activity are associated with pathways of inflammation (Missiroli et al., 2018) and neurodegenerative disorders (Raeisossadati and Ferrari, 2022), it is reasonable to investigate whether the MAM may play a contributive role in DR. Thus far, studies have highlighted the importance of MAM formation and proper ER-mitochondrial communication in insulin signaling as a whole (Cheng et al., 2020a; Tubbs et al., 2014; Wang et al., 2022a), but relatively few have delved into their effect in DR.

An initial study of rat retinal MAM in diabetes has examined whether the diabetic condition is associated with changes in the MAM proteome (Wang et al., 2022b). Interestingly, in rats with STZ-induced long-term Type 1 diabetes, 179 out of 2664 MAM proteins (6.72%) discovered in the retina were significantly altered in concentration, many of which were found to be involved in key processes of cell survival, inflammation, calcium regulation, and protein synthesis and trafficking. Of note, the identified altered MAM proteins were linked to several non-MAM major protein regulators of inflammation, diabetes, and/or DR, illustrating the extent of the MAM’s indirect involvement in a diverse array of signaling pathways. These preliminary data hint at the likely possibility that aberrations in the MAM proteome contribute to the development of DR, but further studies are needed to gauge the degree and mechanisms of its involvement. Differences in the MAM proteome may also exist depending on whether a Type 1 or Type 2 model of diabetes is utilized.

While abnormal concentrations of MAM proteins may potentially feed into DR pathology, normal concentrations of functioning MAM proteins may also instigate DR in the presence of triggers such as ER stress. For example, PERK, as a key ER stress sensor, is activated in DR given that ER stress is implicated in DR pathogenesis (Li et al., 2009b). When activated, PERK transmits ROS signals between the ER and mitochondria via its tethering function at the MAM and thereby promotes ROS-mediated apoptosis (Verfaillie et al., 2012). As long as ER stress is involved in the manifestation of DR, the MAM is equally likely to be involved as well due to its importance in maintaining ER homeostasis.

8.5. Sigma-1R, a MAM protein in retinal protection and disease

Sigma-1R is an ER-resident MAM protein that is classified as a non-opioid receptor. It has been well established that Sigma-1R is a necessary modulatory protein in calcium and lipid exchange between the ER and mitochondria, though there is also growing evidence of its involvement in autophagy, ER stress response, and protein folding (Hayashi and Su, 2007). Much is also known regarding its role in neuroprotection, making Sigma-1R a common protein of interest as a therapeutic target for disorders such as Alzheimer’s disease, dementia, and Parkinson’s disease (Lachance et al., 2023). Thus, it is not surprising that studies have also begun to investigate the potential therapeutic effects of this protein in the retina. Recent data from a study have suggested that Sigma-1R also serves a protective role in an optic nerve crush (ONC) model of glaucoma (Li et al., 2021a); transgenic expression of Sigma-1R in RGCs from Sigma-1R knockout mice resulted in higher RGC counts, attenuated apoptosis, and increased RGC activity following ONC. Similar results could also be seen when Sigma-1R was activated via pentazocine. These study findings associate well with previous data revealing the restorative role that Sigma-1R plays towards mitochondrial function when Sigma-1R is overexpressed and/or activated in RGCs (Ellis et al., 2017). Specifically, administration of Sigma-1R agonists and transgenic overexpression of Sigma-1R in rat RGCs that were oxygen- and glucose-deprived led to restoration of the mitochondrial membrane potential, restoration of cytochrome c activity, and decreased caspase activity. There have also been several other studies correlating accelerated RGC dysfunction and death with a Sigma-1R deficiency in mice, further underscoring the relevance of this protein in the prevention of retinal neurodegeneration (Ha et al., 2012; Mavlyutov et al., 2011).

9. ER stress signaling and autophagy

Autophagy is a self-digested pathway and a major catabolic process activated to protect cells against certain pathological conditions. The last few decades have been important for the cell biology field in establishing a link between UPR and autophagy. The link between UPR and autophagy, however, seems to be more complicated than originally thought. These two systems may act independently, or the induction of one system may interfere with the other. Autophagy can be classified into three subtypes: macroautophagy, microautophagy, and chaperone-mediated autophagy. Macroautophagy (here referred to as autophagy) is accompanied by the process of formation of double-membrane vesicles, autophagosomes that sequester cytosolic proteins and organelles and serve as cargo degradation machinery. Later, these autophagosomes fuse with lysosomes, and as a result, intracellular components are degraded as a part of macromolecule recycling (Maiuri et al., 2007). Activation of autophagy occurs under two different stress conditions: starvation when the activation is aimed at promoting survival and accumulation of misfolded proteins leading to ER stress response when autophagy acts to eliminate damaged intracellular components (Levine and Klionsky, 2004; Martinez-Vicente and Cuervo, 2007; Matus et al., 2008). Among the 30 recently identified mammalian autophagy genes (Gallagher et al., 2016), the most common autophagy marker LC3 (Atg8) and the first identified mammalian gene Beclin-1 (Atg6) are regulated by BCL-2 protein members at the ER membrane.

During UPR, autophagy activation may be necessary to serve as a mechanism for eliminating damaged ER organelles or controlling ER expansion. For instance, the cellular process aimed at the timely removal of damaged ER is known as ER-phagy (He et al., 2021; Reggiori and Molinari, 2022; Yang et al., 2021) ER-phagy is a selective form of autophagy that safeguards cells from the damage caused by excessive ER stress and maintains ER homeostasis. ER-phagy is a multistep process that requires specific receptors and the core autophagic machinery to promote the degradation of ER components. These receptors can be found in the ER or cytosol and are subsequently recruited to the ER membrane. To date, eleven ER-phagy receptors have been identified from various species (He et al., 2021). Among them, six exist in mammals (FAM134B, RTN3L, TEX264, ATL3, SEC62, and CCPG1) and two in budding yeasts (Atg40 and Atg39), which are directly anchored on the ER membrane by their reticulon domains or transmembrane domains. The remaining identified receptors (CALCOCO1, C53, and Epr1) are located within the ER by binding to resident ER membrane proteins. Cells employ various mechanisms to tightly control ER-phagy, including the regulation of the expression of key ER-phagy players. Transcription factors play a pivotal role in controlling the expression levels of these proteins. Notable among these transcription factors are Transcription Factor EB/E3, CCAAT/Enhancer-Binding Protein β, Histone Deacetylase Large Complex, and Regulatory Protein Mig 1 and 2. Dysfunctional ER-phagy may be related to several neurodegenerative diseases, such as Hereditary Sensory and Autonomic Neuropathy (HSAN), Niemann-Pick Type C Disease, autosomal dominant hereditary spastic paraplegia, Alzheimer Disease, and Parkinson Disease (He et al., 2021) (PMID: 34571977). Although several studies have suggested a connection between ER-phagy receptors and these diseases, these associations remain speculative at this point. It’s worth noting that, as of the current knowledge, no retina-associated diseases related to ER-phagy have been described.

Many laboratories have shown that ER stress triggers autophagy, and this effect is regulated by UPR stress sensors, such as IRE1α and PERK signaling (Fig. 6) (Ding et al., 2007; Kouroku et al., 2007; Maiuri et al., 2007; Vicencio et al., 2008). Thus, ATG5 and ATG7 have been proposed to connect autophagy with ER stress through PERK signaling (Zheng et al., 2019). The protective effect of ATG5/7 overexpression on chondrocyte survival has been found to rely strongly on PERK signaling. Herewith siRNA-mediated PERK or Nrf2 knockdown reduced expressions of ATG5, ATG7, and LC3-I/LC3-II while expression of p62 was oppositely increased. Meantime, knockdown of ATF4 manifests an antagonistic effect suggesting that the overall PERK signal is the pivot for autophagy, ER homeostasis, and ER-phagy. Interestingly, overexpression of ATG5 and ATG7 in turn not only enhances autophagy but also inhibits ER stress in chondrocytes (Zheng et al., 2019). Together, ATF4 and its transcriptional target CHOP, have been proposed to regulate more than a dozen different ATG genes (B’Chir et al., 2013). Three classes of autophagy genes were distinguished recently according to their dependence on ATF4 and/or CHOP binding to specific promoter cis-elements. It was found that CHOP and ATF4 together are essential for the transcriptional activation of p62, NB1, and ATG7. The CHOP protein alone activates ATG10, Gabarap, and ATG5, while ATF4 by itself regulates the expression of Atg16l1, Map1lc3b, Atg12, Atg3, Beclin1, and Gabarap12 genes (B’Chir et al., 2013). Overall, this study reveals a novel regulatory role of the eIF2a–ATF4 pathway activated as a result of different stresses, amino acid starvation, and ER stress in tuning the autophagy gene transcription program.

Fig. 6.

The UPR and autophagy. A) All three UPR pathways alter autophagy. IRE1 and ATF6 can regulate autophagy by acting on regulators of Beclin1 while the PERK pathway results in the upregulation of transcription of autophagy related genes. These pathways alter the transcriptional landscape to promote autophagy. B) The UPR alters autophagy through TRB3’s interaction with P62 resulting in its subsequent impairment of autophagy.

Another UPR signaling pathway, the IRE1 pathway, is also known to activate autophagy. However, the mode of this activation is indirect. The IRE1 interaction with adapter proteins, such as tumor necrosis factor (TNF), receptor-associated factor 2 (TRAF2), and apoptosis signal-regulating kinase 1 (ASK1), results in the formation of the IRE1/TRAF2/ASK1 complex that activates c-Jun N-terminal kinase (JNK) (Prestes et al., 2021). In turn, JNK phosphorylates transcription factor c-Jun, inducing expression of Beclin1 (Liu et al., 2020). Moreover, this effect could be as protective against cell death, depending on the duration of ER stress (Lindner et al., 2020; Liu et al., 2020).

Approaches to investigate the role of ATF6 signaling in the regulation of autophagy have been taken as well. However, how ATF6 mediates ERS-induced autophagy remains unclear. It has been reported that silencing ATF6 inhibits autophagy and affects LC3 conversion in fibroblasts treated with hydroxycamptothecin, an antineoplastic drug, overall suggesting a role of ATF6 in autophagy activation (Tao et al., 2021). Another potential mode of ATF6-mediated regulation of autophagy is an increase in the expression of death-associated protein kinase1 (DAPK1), which in turn phosphorylates Beclin1 (Gade et al., 2014; Zhou et al., 2016). Additionally, it is known that mAtg9 trafficking, which is critical for autophagosome formation, also occurs via the ATF6-regulated expression of DAPK1 (Zhou et al., 2016). In this study, the authors found that generation of a HepG2 cell line with a stable ATF6 and DAPK1 knockdown results in a decrease in the conversion ratio of LC3 upon the treatment of quinocetone, a potent synthetic antimicrobial agent used for improving feed efficiency and controlling dysentery in food-producing animals (Zhou et al., 2016). The UPR signaling has also been linked to autophagy in the vision research field.

9.1. Inherited retinal degeneration

A study conducted by Yamoah et al. demonstrated that the retina of rd10 mice mimicking retinitis pigmentosa (RP) in humans caused by mutant PDE6b manifests RNA binding protein (RBP) aggregation (Yamoah et al., 2023). Moreover, this accumulation takes place as an early pathogenic event, damaging non-photoreceptor retinal cell types and is independent of Pde6b gene defects. The authors also revealed that robust increases in levels of the protective ER calcium (Ca²⁺) buffering chaperone Sigma-1R, together with other ER-Ca²⁺ buffering proteins in photoreceptors and non-photoreceptor neuronal cells, occur before any noticeable photoreceptor degeneration in these mouse retinas. In addition, the changes in the ER resident proteins were accompanied by altered expression of autophagy proteins p62 and LC3, abnormal ER widening, large autophagic vacuole formation detected by EM, and stress granule formation. The authors concluded that progressive neurodegeneration in the rd10 mouse retina is associated with early disturbances of proteostasis and autophagy, along with abnormal cytoplasmic RBP aggregation (Yamoah et al., 2023).

The human rod photoreceptor-specific rhodopsin protein was the first gene identified to cause causative RP (Dryja et al., 1990a, 1990b; Thiagalingam et al., 2007). More than 150 different mutations in RHO are associated with 25% autosomal dominant (ad) inherited RP (Meng et al., 2022). For this disease, two classes of RHO mutations have been described. Class A mutations (R135G, R135L, R135W, V345L, and P347L) trigger loss of rod function in the entire retina, with an early onset of blindness (Cideciyan et al., 1998). Class B mutations, in turn, can be divided into two subclasses: B1 and B2. The subclass B1 mutations (T17M, P23H, T58R, V87D, G106R, and D190G) lead to a milder phenotype characterized by normal rod activation kinetics and preserved rod outer segment length, although mutation-specific abnormalities in the rod visual cycle are detected. Subclass B2 mutations (G51A, Q64ter, and Q344ter) show no regional retinal predisposition for disease (Meng et al., 2022).

The P23H RHO mutation is the most prevalent RHO mutation in North America, accounting for 10% of cases of adRP. Models expressing the P23H RHO mutations include mice and rats. Moreover, both transgenic and knock-in animals have been generated, although fine mechanisms of P23H RHO aggregation and degradation have been explored in vitro. For example, genetic and proteomic analysis of P23H rhodopsin mouse retinas demonstrated induction of ER-associated protein degradation in vivo (Chiang et al., 2015; Kim et al., 2022), and Intartaglia et al. demonstrated that induction of autophagy causes clearance of the P23H RHO aggregates in vitro (Intartaglia et al., 2022). Previously, this group identified the ezrin protein as an inhibitor of autophagy and lysosomal functions in the retina. To identify potential pharmacological targets, the investigators inhibited ezrin and treated the P23H RHO mice with NSC668394 inhibitor daily and followed the treated mice with electrophysiological recording and molecular biological assessment of the retinal proteins. The authors revealed that the approach inhibiting ezrin promotes lysosomal clearance of diseased-linked P23H RHO agglomerates, which in turn reduced ER stress, provided robust decreases in photoreceptor cell death, and ameliorated both retinal morphology and function. For example, reductions in PERK, XBP1s, ATF6, and CHOP and an increase in BIP were associated with a decline in LAMP1, p62, and LC3II in the treated mouse group. Therefore, a better understanding of how the therapeutic use of autophagy inducers could be applied in the degenerating retina is necessary to move the field forward in clinical studies or to evaluate the therapeutic efficacy of autophagy inducers (Intartaglia et al., 2022).

A group led by David Zacks recently reported that knock-in P23H RHO mice demonstrate elevated levels of autophagy flux, and that the pharmacological stimulation of autophagy in these mice accelerates retinal degeneration (Qiu et al., 2019). Overall, the P23H RHO mouse retinas manifested the activation of UPR markers concomitantly with increases in Beclin1, p-p62, p62, and GFP-LC3 puncta. In this study, the authors found that genetic deletion of Atg5 gene improved photoreceptor structure and function, while mTOR inhibition by CCI-779 resulted in increased autophagy and accelerated retinal degeneration in P23H RHO mice (Intartaglia et al., 2022). Contrary to the study led by Conte (Intartaglia et al., 2022), the study by Zacks lab proposed autophagy-induced cell death as a contributing factor in retinal degeneration caused by P23H RHO. Later, this group also revealed that the ratio of autophagy to proteosome (A:P) in P23H RHO is an overall important marker defining photoreceptor cell homeostasis and that shifting the A:P ratio is key to reducing proteotoxic cell death (Qiu et al., 2019). In the referred study, the P23H mice were treated with a chemical chaperone (4-phenylbutyric acid) to improve rhodopsin folding or with a selective phosphodiesterase-4 inhibitor (rolipram) to increase proteasome activity. Both treated P23H RHO mouse groups exhibited a reduced ER stress response, decreased autophagy flux, increased proteasome activity, and decreased activation of cell death pathways (Qiu et al., 2019).

Using transgenic T17M RHO mice modeling adRP in humans, our group also detected aberrant autophagy in the retina (Bhootada et al., 2016). We found that at P30, the T17M RHO retinas show increased ER stress markers, such as PERK, p-eIF2α, ATF4, and p-ATF6, in addition to a reduction in LC3 conversion and beclin1 and p62 levels. Conversely, ATF4 knockdown significantly reduced retinal degeneration in T17M RHO mice, leading to photoreceptor survival as measured by scotopic and photopic ERGs and photoreceptor nuclei row counts. This delay was accompanied by a dramatic decrease in UPR signaling, restoration of LC3 conversion, and levels of Beclin1 and p62. We therefore proposed that an increase in autophagy genes may also manifest temporal protection in T17M RHO ATF4 ± mice and this strategy may contribute to photoreceptor health at early time points through increasing a clearance of intracellular cargo; however, this approach may not be sufficient to maintain cellular homeostasis at later time points (Bhootada et al., 2016).

Recently, in an Rh1^P23H Drosophila model of adRP, it was proposed that degradation of wild-type rhodopsin is mediated by selective autophagy induced by IRE1/XBP1 signaling and insufficient proteasome activities (Zhao et al., 2023). The authors revealed that the upregulation of PERK signaling during ER stress prevents autophagy and suppresses retinal degeneration in the fly adRP model. Thus, by comparing the Rh1^P37H–GFP, PERK^RNAi, and Rh1^P37H-GFP of flies’ retinas, the authors found that deficiency in the PERK pathway overactivates IRE1, leading to the accumulation of Rh1^P37H due to insufficient proteasome activity and degradation of wild-type Rh1 by autophagy, ultimately causing neuron dysfunction and degeneration. The authors proposed a pathological role of autophagy under these neurodegenerative conditions and indicated that promoting PERK activity could be used to treat ER stress-related neuropathies, including adRP (Zhao et al., 2023).

Not only animal models of adRP manifest concomitant upregulation of UPR markers and autophagy. We examined rd16 mice, which mimics Leber congenital amaurosis and manifest severe and rapid ciliopathy due to expression of the mutant gene encoding CEP290 (Collin and Garanto, 2017). In these animals, we found persistent activation of the integrated stress response (p-eIF2α, ATF4, TRIB3, GADD34, and CHOP) and compromised activities of AKT and mTOR at P15 and P20 (Starr et al., 2018). We also found an increase in beclin1, LC3 conversion (I to II), and p62 accumulation, suggesting an increase in autophagy flux (Saltykova et al., 2021). However, when we visualized the rd16 retina expressing the RGP-eGFP-LC3 transgene driven by the control of the CAG promoter under fluorescence microscopy, we found a decrease in the number of red puncta (RFP and acid insensitive) localized in the IS of rd16 photoreceptors. The latest suggests that the conversion from autophagosome (neutral pH) to autolysosome (acidic pH) in rd16 photoreceptors could be reduced. We then decided to investigate the consequences of reduced mTOR activation and ablated TRIB3 in the rd16 retina. The TRIB3 protein serves as a pseudokinase to prevent phosphorylation of AKT and mTOR (Borsting et al., 2014; Salazar et al., 2015). Therefore, in rd16 TRIB3^−/− retinas, in addition to restored p-AKT and p-mTOR levels, we found reduction in Beclin1, reduced LCA3II/I ratio, and diminished p62 accumulation, which together with an increased number of red puncta, a marker of healthy autolysosomes, are evidence of the restoration of autophagolysosome function (Saltykova et al., 2021). However, at this point, we do not know whether improvement of autophagolysosome flux is an mTOR dependent event. Independently of mTOR, TRIB3 could interact with autophagic receptor p62 and abrogate the binding of LC3 and ubiquitinated substrates to P62, which induces the blockage of autophagic flux and subsequent ubiquitin proteasomal system (UPS) defects, leading to p62 accumulation (Fig. 6) (Hua et al., 2015). Therefore, interrupting TRIB3-p62 interaction could be a potential novel strategy against different retinopathies associated with p62 depositions and defective UPS.

9.2. Age-related macular degeneration (AMD)

AMD is a disease with a special need for the efficient removal of cellular waste in RPE cells due to elevated oxidative stress causing protein misfolding and ER stress. Waste clearing in the RPE includes UPS and the autophagosome-lysosomal system (ALS). Compared to UPS, ALS could degrade damaged organelles in addition to the degradation of proteins. Studies on RPE cells from AMD donors and mice with AMD-like phenotypes suggest that autophagy increases during aging and AMD (Mitter et al., 2014). Various environmental risk factors (cigarette smoke and light exposure) have been linked to AMD pathology. Thus, in a study conducted with mice exposed to cigarette smoke, we found dysregulated ER stress-dependent antioxidant responses associated with incidents of apoptosis and activation of autophagy in the RPE/choroid samples (Chen et al., 2014a). In this study, we employed strategies aimed at suppressing ER stress or inhibiting CHOP activation by either pharmacological chaperones or genetic ablation. These approaches were found to attenuate apoptosis in RPE cells exposed to hydroquinone, causing oxidative stress. In a study by Zhang et al. using RPE-specific deletion of Atg5 and Atg7 in the mouse retina, the authors found an accumulation of p62 in the RPE, suggesting autophagy deficiency and disrupted autophagy flux (Zhang et al., 2017). The Atg5^ΔRPE and Atg7^ΔRPE mice occasionally manifested retinal degeneration (35% and 45% of the entire group, respectively), which gradually increased with age. In addition, RPE atrophy and choroidal neovascularization were occasionally observed in mice of advanced age. Therefore, the authors concluded that autophagy deficiency induced by RPE-specific deletion of Atg5 or Atg7 predisposes, but does not necessarily drive, the development of AMD-like phenotypes or retinal degeneration (Zhang et al., 2017).

Overall, the lack of animal models truly reflecting all aspects of AMD retinal pathology in humans most likely significantly impedes the investigation of crosstalk between UPR and autophagy. Therefore, the majority of the studies have been conducted in cell models; one that is of particular importance in this field is the study by Porter et al., which showed significant dysregulation of EIF2AK3 (PERK) gene expression in the RPE of human donors with early and intermediate AMD (Porter et al., 2019). In ARPE-19 cells exposed to brefeldin A, the authors found that PERK downregulation results in increased ER stress and impaired apoptosis induction, antioxidant responses, and autophagic flux (Saptarshi et al., 2022). They also showed that PERK regulates the rate of autophagy induction in a time-dependent manner and that PERK downregulation is an integrative event leading to reduced RPE cell survival in AMD development (Saptarshi et al., 2022). Finally, the authors proposed PERK as a potential future therapeutic target for AMD (Saptarshi et al., 2022).

Accumulated unfolded proteins activate UPR and are targeted for degradation by UPS or autophagy, the two major proteolytic systems for clearance of misfolded or damaged proteins. The study conducted by Zhan et al. investigated how these two systems communicate and coordinate with each other in RPE cells to eliminate intracellular misfolded and damaged proteins (Zhan et al., 2016). The authors employed ARPE-19 cells treated with proteasome inhibitors MG132 and chloroquine and found that when the level of ubiquitinated protein aggregations is significantly increased after the treatment of MG132, the RPE cells also manifest an increase in the levels of LC3-I, LC3-II, and LAMP1. Moreover, the levels of γ-tubulin and p62 were also increased in MG132-treated cells. Alternatively, the inhibition of lysosomal activity with chloroquine increases the levels of ubiquitin conjugates, LC3-II, and p62. In this study, the authors concluded that UPS and ALS are interrelated and that dysfunction of the ALS results in dysfunction of the UPS and severely compromises the capacity to eliminate misfolded and other forms of damaged proteins.

9.3. Diabetic retinopathy

Compared to the nondiabetic retina, the human diabetic retina manifests activation of the UPR markers (Du et al., 2013; Pitale et al., 2021) and an increase in autophagy proteins ATG5, Beclin1, and LC3-II (Fu et al., 2016). Interestingly, a literature review suggests that autophagy may play a dual role either protecting or damaging retinal cells in DR pathogenesis (Gong et al., 2021). For example, in human pericytes, heavily oxidized glycated low-density lipoprotein (HOG-LDL) at low doses induces mild ER stress triggering protective autophagy while at higher concentrations increases ER stress leading to autophagic and apoptotic death, suggesting the impacts of ER stress and autophagy on cell viability are similar and dose-dependent (Fu et al., 2016). Another study demonstrated the protective role of autophagy in the context of activated UPR. In this study, the authors found an increase in ER stress markers, including p-PERK, p-eIF2a, and CHOP, and a concomitant inhibition of autophagy in ARPE-19 cells cultured under high glucose (25 mM) or hypoxia (1% oxygen) (Miranda et al., 2012). Treatment with fenofibrate protected the cells by upregulating LC3-II and thus increasing autophagy flux associated with inhibition of stress-mediated signaling and improvement of tight junction (Miranda et al., 2012). By contrast, a study by Devi et al. found that high glucose induced pro-inflammatory factor TXNIP expression in retinal Müller cells, which evokes a process of cellular defense, ultimately leading to the activation of ER stress, inflammation, apoptosis, and autophagy, contributing to the development of DR (Devi et al., 2012).

Overall, currently published studies pointed out existing crosstalk between ER stress and autophagy, and in many cases, the inhibition of ER stress is linked to alteration of autophagy flux. It is also worth mentioning that autophagy in the retina manifesting in different retinal degenerative conditions may play distinct roles. This most likely explain the fact that different strategies inducing or inhibiting autophagy flux benefit the degenerating retina. Moreover, not only degenerative conditions, but also experimental conditions, such as time points and doses of pharmacological treatments, may contribute to the results of the study. Nevertheless, manipulation with autophagy, either ER stress-mediated or direct, could be effective strategies to slow down inherited or age- and diabetes-related retinal pathogenesis.

10. ER stress signaling in mitochondrial regulation

Because of the close functional and physical inter-connections between ER and mitochondria (described in detail in Section 8), ER stress impacts morphology and function of mitochondria. A growing body of data supports that the UPR – in addition regulating ER homeostasis – also regulates mitochondria in response to ER stress (Malhotra and Kaufman, 2011; Rainbolt et al., 2014; Vannuvel et al., 2013). The PERK signal transduction cascade is especially important in regulating mitochondria function and morphology in response to ER stress. PERK signaling promotes mitochondria elongation and fusion in cell culture studies through its translational and transcriptional programs (Lebeau et al., 2018; Perea et al., 2023). The PERK-regulated ATF4 transcription factor also induces many mitochondria quality control genes (Han et al., 2013); mitochondrial respiratory chain assembly and activity (Balsa et al., 2019); mitochondria membrane phospholipid composition (Perea et al., 2023). In cell culture, sustained PERK signaling also triggers cell death through multiple pro-apoptotic pathomechanisms including induction of CHOP (Marciniak et al., 2004); attenuation of anti-apoptotic IAPs (Hiramatsu et al., 2014); and disruption of ATP levels (Hiramatsu et al., 2020).

In retina, the link between ER stress and mitochondria morphology and function has not been as extensively examined as in vitro. Interestingly, ATF6 mutant retinal organoids from achromatopsia patients showed excessive ER stress, as to be expected with disruption of a UPR signaling pathway, but also showed extensive mitochondria structural abnormalities by ultrastructural analyses and significant induction of oxidative phosphorylation pathways in developing photoreceptors (Lee et al., 2022). These mitochondria abnormalities in retinal organoids suggest that ER stress may also lead to morphologic and functional consequences on mitochondria in specialized retinal cell types. In turn, disruption of mitochondria could also contribute to pathogenesis and progression of retinal diseases due to ER stress.

11. The integrated stress response (ISR)

As mentioned in the PERK section, the UPR results in the phosphorylation of eIF2α by PERK. However, a multitude of cellular stresses lead to elevation in p-eIF2α levels. In addition to PERK, there are at least three other kinases — heme-regulated kinase (HRI), protein kinase double-stranded RNA-dependent (protein kinase R or PKR) and general control nonderepressible 2 (GCN2) — in vertebrates that phosphorylate eIF2α at Serine 51 in response to a variety of stresses (Costa-Mattioli and Walter, 2020; Hinnebusch, 1984). HRI is mainly activated in blood cells due to heme deprivation. PKR is activated during viral infection and GCN2 is activated in amino acid starvation. Interestingly, Wu et al. recently published their exciting findings on a fifth eIF2α kinase, FAM69C, which acts on eIF2α and promotes stress granule formation in primary microglia (Wu et al., 2023). Additional work will need to be carried out to determine what the primary triggers of FAM69C as well as its role in other cell types and tissues including the retina. Each of these pathways converge on and phosphorylate eIF2α to stop conventional protein synthesis and enabling translation of ATF4. ATF4 promotes transcription of many genes involved in a variety of cellular processes ranging from development to apoptosis. The most studied targets of ATF4 are the ones that ATF4 promotes following a cellular stress resulting in elevated p-eIF2α levels. Together, this translational inhibition, the subsequent activation of an altered translational program and stress-induced gene targeting by ATF4 is the ISR (Donnelly et al., 2013; Harding et al., 2003).

Aside from PERK, very little research has been conducted to assess the role of eIF2α kinases in IRD. This is mainly because many IRDs have misfolded proteins (Gorbatyuk et al., 2010; Kunte et al., 2012; Murray et al., 2015), meaning the ISR in these retinas should be dependent on PERK. However, we cannot definitively exclude the possibility that other eIF2α kinases are active in these diseases. We previously attempted to answer this question using a small molecule inhibitor of PERK, GSK2606414, and while inhibition of PERK led to normalized levels of p-eIF2α (S51), it only marginally increased translation rates, which could possibly indicate activity of other eIF2α kinases (Starr et al., 2018). Given the inhibition of PERK results in diminished eIF2α phosphorylation without affecting translation rates, we then investigated whether an increase in p-eIF2α more prominently arrests protein synthesis. Knocking out Gadd34—one of the two known PP1 regulatory subunits that facilitates the dephosphorylation of eIF2α, in two different models of IRD, rd16 and P23H RHO mice—we learned there was also no significant difference in the levels of protein synthesis associated with p-eIF2a increase (Saltykova et al., 2022; Starr and Gorbatyuk, 2019b). Altogether this series of experiments indicated that either during chronic ER stress, translational regulation could occur through other mechanisms (Starr et al., 2018, 2019) or p-eIF2 inhibition reaches a threshold in degenerating photoreceptors.

Due to the nature of the ISR’s alleged role in various neurodegenerative diseases, groups have worked on targeting this signaling for therapeutics. The ideal pharmacological compound would be one that targets the ISR effectively without destroying the tissues that rely on it so heavily. The small molecule ISRIB was developed for this purpose. ISRIB inhibits the ISR by preventing the interaction with p-eIF2α(S51) and eIF2B (Zyryanova et al., 2021). Though ISRIB has had promising results in studies of various diseases (Chang et al., 2022; Hosoi et al., 2016; Wong et al., 2018) and retinal disorders related to wet AMD (Yasuda et al., 2021) and glaucomatous RGC loss (Larhammar et al., 2017), no reports on a benefit for IRD has been published to date.

12. Emerging pharmacological regulators of the UPR

The elucidation of the molecules and mechanisms regulating UPR lead to development of small molecule screens to identify chemicals that selectively modulate the IRE1, PERK, and ATF6 signal transduction pathways in the absence of ER stress or protein misfolding. These small molecules provide tools to harness the activities of specific UPR signaling molecules and test consequences in different disease settings, and those compounds that show beneficial effects in disease models have been advanced into additional preclinical studies to potentially develop into pharmaceutical agents to treat disease in people. Here, we discuss several of the most robust small molecule regulators of UPR signaling and highlight interesting biologic effects in disease models.

For the IRE1 pathway, a small molecule activator, IXA4, has been identified that promotes generation of the XBP1s transcription factor and selective upregulation of XBP1s target genes but not transcriptional programs regulated by other arms of the UPR (e.g., PERK/ATF4 and ATF6) (Grandjean et al., 2020). This small molecule IRE1 pathway activator was effective in vivo in a mouse model of obesity where intraperitoneal injections of IXA4 improved systemic glucose levels and reduced hepatic steatosis (Madhavan et al., 2022). A small molecule aldehyde compound, 4u8c, covalently modifies and inactivates that RNase domain of IRE1, thereby inhibiting this UPR signaling pathway (Cross et al., 2012) (Fig. 7).

Fig. 7.

Small molecules targeting ATF6, PERK, and IRE1 pathways. Ceapin-A7 inhibits ATF6 signaling by trapping the ATF6 molecule in the ER, thereby preventing the generation of ATF6 transcriptional activator. AA147 selectively activate the ATF6 signaling by inhibiting the activity of protein disulfide isomerases to increase reduced ATF6 monomers in the ER. IXA4 selectively upregulates IRE1/XBP1s target genes. 4u8c, covalently modifies and inactivates that RNase domain of IRE1. The PERK inhibitor, GSK2606414 and GSK2656157 bind to the PERK kinase and thereby inhibit the PERK pathway. The ISRIB binds and stabilizes the active form of eIF2B which is normally rendered inactive by PERK signaling. Salubrinal inhibits p-eIF2α dephosphorylation.

For the PERK pathway, small molecules including salubrinal,guanabenz, sephin1, and raphin1 have been identified that enhance PERK signaling by inhibiting intracellular phosphatases that normally counteract the kinase activity of PERK and related eIF2a kinases (Boyce et al., 2005; Das et al., 2015; Krzyzosiak et al., 2018; Tsaytler et al., 2011). The specificity of these compounds for target phosphatases as well as the molecular mechanisms of phosphatase inhibition are incompletely understood. However, in experimental neurodegeneration disease models, extension of phosphorylated eIF2a levels (and thereby extension of PERK signaling by application of these molecules has prevented neuropathology (Das et al., 2015; Park et al., 2023). The GSK2606414 and related GSK2656157 small molecules bind the PERK kinase and thereby inhibit the PERK pathway (Axten et al., 2012, 2013). These compounds are not PERK-specific as they can also inhibit other kinases (Mahameed et al., 2019; Rojas-Rivera et al., 2017), but administration of these compounds prevented neurodegeneration in mouse prion disease model (Moreno et al., 2013). The ISRIB compound interferes with PERK signaling via binding and stabilizing the active form of eIF2B which is normally rendered inactive by PERK signaling (Zyryanova et al., 2021) (Fig. 7). The autosomal recessive Vanishing White Matter demyelination disease arises in people carrying destabilizing eIF2B missense mutations, and ISRIB has been effective in rescuing eIF2B activity and preventing neuropathology for some of these disease mutations in animal studies (Wong et al., 2018). A second generation ISRIB derivative has advanced into early-stage clinical safety trial (ClinicalTrials.gov Identifier: NCT05740813) and, if successful, may offer a pharmacologic treatment particularly suited to Vanishing White Matter patients carrying disease eIF2B mutations.

For the ATF6 pathway, a small molecule activator of this pathway has been identified, AA147 (Plate et al., 2016). AA147 covalently targets protein disulfide isomerases (Paxman et al., 2018) that in turn, lead to increase in the pool of monomeric reduced full-length ATF6 available to traffic to Golgi for generation of the ATF6 transcription factor fragment. AA147 has been effective in activating ATF6’s transcriptional program in cell culture and mouse disease models (Blackwood et al., 2019; Kroeger et al., 2018; Wang et al., 2022c). Ceapins are selective ATF6 inhibitors that tether ATF6 protein in the ER with ABCD3 protein in peroxisome, and in doing so, prevent the generation of the ATF6 transcription factor fragment at the Golgi (Torres et al., 2019) (Fig. 7).

AA147 Restores Transcriptional Activity of ATF6 Achromatopsia Disease Mutations.

AA147 enhances ATF6 signaling by promoting ATF6 trafficking from ER to Golgi. This raised the possibility that AA147 may be well tailored to restore function of the Class 1 ATF6 disease variants found in achromatopsia patients, where ATF6 is retained in ER. In cells expressing recombinant ATF6 bearing Class 1 luminal domain variant proteins, AA147 addition restored generation of the ATF6 transcription factor fragment (Chiang et al., 2017; Kroeger et al., 2021), and in retinal organoids derived from achromatopsia patients carrying Class 1 mutations, AA147 increased ATF6 transcriptional output and most intriguingly, restored development of cone photoreceptors, specifically growth of the IS/OS structure (Kroeger et al., 2021). These findings demonstrate that chemical restoration of ATF6 function in developing ATF6-mutant retinal organoids restores cone growth, specifically the extension of IS/OS structure and cone gene expression. Based on these results, ATF6 activators like AA147 are particularly well-suited for ATF6 Class 1 disease variants, and more broadly may offer benefits for photoreceptor growth, viability, and structure in other vision loss diseases.

13. Conclusions and future directions

Studies from nearly over two decades have provided compelling evidence for the important roles of ER stress signaling in maintaining the cellular homeostasis of retinal neurons during development, maintenance, and aging. Loss and/or reduced functional UPR pathways mediated by IRE1/XBP1 and ATF6 contribute to retinal degeneration associated with aging and age-related diseases (Kohl et al., 2015; Massoudi et al., 2023; McLaughlin et al., 2018). Paradoxical results have been reported in the context of the impacts and consequences of defective PERK pathway, in part due to the great diversity of the downstream effector of this UPR branch (Athanasiou et al., 2017; Starr and Gorbatyuk, 2019a; Zhu et al., 2017). The close interactions and overlapping between the UPR signaling and other cellular stress response pathways mediated by autophagy, ISR, mitochondria, calcium regulation, etc. further add to the complexity of the biological and functional outcome of UPR activation or inactivation. Moreover, even the same UPR molecule can demonstrate distinct functions in different cell types, animal models, and treatments. Therefore, the interpretation of the findings should be carefully justified based on the experimental systems and conditions. Nevertheless, research to uncover the structural and functional characteristics of the ER and ER stress signaling would provide insightful information to improve our understanding of the fundamental mechanisms underlying the survival and function of retinal neurons and vascular cells. More importantly, it would help the discovery of signaling pathways of cellular response by which retinal cells adapt to physiological and pathological environments and identify molecular targets for the development of new treatments for retinal diseases. Interestingly, many variants of UPR, ISR, and protein quality control genes are found in the human population and linked to human diseases (Table 1). Patients carrying loss-of-function ATF6 and WFS1 gene variants develop vision loss, but it is unknown how disease variants in other protein quality control genes affect the retina. Given the importance of UPR, ISR, and protein quality control mechanisms in photoreceptor disease, diabetic retinopathy, retinal inflammation, and the many other ocular diseases highlighted in this review, future studies should investigate if human genetic variation in ER and protein quality control genes directly causes vision loss diseases (as demonstrated with WFS1 and ATF6) or acts as risk factors and genetic modifiers for the pathogenesis or progression of retinal diseases.

Table 1.

UPR and ER stress gene variants linked to human diseases.

Gene	Molecular Function	Human Disease (Phenotype)	Reference
WFS1	ER membrane ion channel	Wolfram Syndrome (juvenile diabetes, optic nerve atrophy)	Inoue et al. (1998)
EIF2AK3 (PERK)	Unfolded Protein Response kinase	Wolcott-Rallison Syndrome (juvenile diabetes, failure to thrive, early death)	Delépine et al. (2000)
EIF2B	Integrated Stress Response translation initiation factor	Vanishing White Matter disease	Leegwater et al. (2001)
XBP1	Unfolded Protein Response transcription factor	Bipolar Disorder risk factor	Kakiuchi et al. (2003)
SIL1	ER Co-chaperone for BiP/GRP78 chaperone	Marinesco-Sjogren Syndrome	Anttonen et al. (2005)
EIF2AK3 (PERK)	Unfolded Protein Response kinase	Progressive Supranuclear Palsy (tauopathy neurodegeneration) risk factor	Höglinger et al. (2011a)
EIF2AK4 (GCN2)	Integrated Stress Response kinase	Pulmonary hypertension	Eyries et al. (2014)
ATF6	Unfolded Protein Response transcription factor	Achromatopsia	Kohl et al. (2015)
ATF6	Unfolded Protein Response transcription factor	Cone-Rod Dystrophy	Skorczyk-Werner et al. (2017)
EIF2AK1(HRI)	Integrated Stress Response kinase	Leukoencephalopathy	Mao et al. (2020)
EIF2AK2(PKR)	Integrated Stress Response kinase	Leukoencephalopathy	Calame et al. (2021)

As discussed in Sections 4 - 7, activation of the UPR has been implicated in both inherited and age-related retinal degenerative diseases as well as in diabetes-induced neurovascular damage and angiogenesis. In animal models of inherited retinal degeneration, improving protein homeostasis by overexpressing GRP78 or enhancing ATF6 activation using pharmacological activators reduces photoreceptor loss, improves retinal function, and restores cone photoreceptor growth. Treatment of P23H RHO mice with pharmacological chaperones that promote the production and correct localization of rhodopsin protein also reduces retinal inflammation and ameliorates photoreceptor degeneration (Vats et al., 2022). These findings suggest that restoring protein homeostasis is a promising approach for the treatment of photoreceptor degeneration in certain inherited retinal diseases. In parallel with the influence of genetic factors on protein dyshomeostasis, environmental factors also lead to chronic stress that disrupts protein homeostasis in retinal cells. For instance, accumulation of ubiquitinated protein aggregation and tau-like proteins was observed in both the retina tissue and the RPE, which is believed to contribute to RPE degeneration in AMD (Leger et al., 2011). Overexpression of the UPR regulators, ATF6 and XBP1, has shown great beneficial effects on neuroprotection in vivo in preclinical models of Parkinson’s disease and Huntington’s disease (Vidal et al., 2021). Thus, enhancing the function of the adaptive UPR by genetic or pharmacological approach (discussed in Section 12) may provide new hope for age-related degenerative retinal disease.

In addition to protein homeostasis, maintaining a well-balanced lipid content and composition is critical for retinal development, neuronal survival, synapse formation, and function. For example, lack of cholesterol, the major sterol component of the lipid bilayer of the cell member, causes progressive retinal degeneration in Smith-Lemli-Opitz syndrome (SLOS), whilst aberrant accumulation of cholesterol increases oxidative stress and inflammation contributing to retinal degeneration in multiple diseases, such as DR and AMD (Dasari et al., 2010; Elmasry et al., 2018; Fliesler et al., 2004; Wu et al., 2012). Emerging evidence suggests dysregulation of lipid metabolism, in addition to protein dyshomeostasis, is an important cause of ER stress. Several studies have demonstrated oxidative cholesterol metabolites such as 27-hydroxycholesterol (Dasari et al., 2010), highly oxidized glycated LDL (Wu et al., 2012), 12/15-lipoxygenase (Elmasry et al., 2018), and 7-Ketocholesterol (Pariente et al., 2023), can increase ER stress in cultured retinal cells. However, the exact mechanism by which lipid dysregulation activates the UPR signaling and the role of the UPR pathways in lipid-mediated retinal cell damage and dysfunction in disease development remains largely unknown. Related to lipid and protein homeostasis, the ER plays an important role in assisting and regulating the biogenesis, remodeling, and function of intracellular organelles. For example, as discussed in Section 8, the ER closely interact with the mitochondria via MAM, which plays a significant role in calcium homeostasis, mitochondrial remodeling and function, as well as regulation of oxidative stress and inflammatory signaling. Recent proteomic studies identified a number of differentially expressed retinal MAM proteins to be involved in key pathogenic pathways of DR, such as glucose metabolism, retinal degeneration, fibrosis, and angiogenesis. However, the implication and regulation of these MAM proteins in DR pathogenesis remain to be explored. In macrophages and neutrophils, the ER has been shown to form close contact with the plasma membrane and participate in the process of phagocytosis. To our knowledge, whether the ER-phagosome contact sites exist in the RPE and are implicated in POS phagocytosis and clearance have not been studied and warrant investigation. Understanding the molecular pathways of the ER and ER stress signaling that regulate protein and lipid homeostasis in each specific retinal cell type, and in the retina as an integrated neural tissue involving active neuron-glia-vascular interactions, and developing novel therapeutics targeting these pathways will likely lead to new approaches for the prevention and treatment of retinal disease.

Abbreviations

ACC

acetyl-coenzyme carboxylase

AdRP

autosomal dominant retinitis pigmentosa

AGE

advanced glycation end products

AKAP1

A-kinase anchoring protein 1

ALS

autophagosome-lysosomal system

AMD

age-related macular degeneration

AMPK

AMP-activated protein kinase

ASC

apoptosis-associated speck-like protein

ASK1

apoptosis signal-regulating kinase 1

ATF4

activating transcription factor 4

ATF6

activating transcription factor-6

BCL-2

B-cell lymphoma 2

BRB

blood-retinal barrier

bZIP

basic leucine zipper

C/EBP

CCAAT box/enhancer-binding protein b

C3G

Cyanidin-3-glucoside

CEP290

centrosomal protein of 290 kD

CHOP

C/EBP homologous protein

CNS

central nervous system

CNV

choroidal neovascularization

CORD

cone-rod dystrophy

CRE

cAMP response element

CTRP1

C1q/TNF-related protein 1

CYPD

Cyclophilin D

DAMPs

damage-associated molecular patterns

DAPK1

death-associated protein kinase 1

DEGS1

dihydroceramide desaturase Δ4-dihydroceramide desaturase 1

DME

diabetic macular edema

DR

diabetic retinopathy

DR5

death receptor 5

Drp-1

Dynamin related protein 1

EIF2AK3

eukaryotic translation initiation factor 2-alpha kinase 3

eIF2α

-subunit of the eukaryotic translation initiation factor 2 complex

EMT

epithelial to mesenchymal transition

ER

endoplasmic reticulum

ERAD

ER-associated degradation

ERMES

ER-mitochondria encounter structure

FFAs

free fatty acids

FUNDC1

FUN14 domain containing 1

GA

geographic atrophy

GADD34

growth arrest and DNA-damage inducible 34

GCN2

general control nondepressible 2

GRP75

75-kDa glucose-regulated protein

GTP

guanosine tri-phosphate

HK-2

human kidney 2

HRI

heme-regulated kinase

HSP

hereditary spastic paraplegia

IFN-γ

interferon-γ

IP3R

inositol 1,4,5-triphosphate receptor

IRD

inherited retinal degeneration

IRE1

inositol-requiring protein-1

IRES

internal ribosome entry site

IS

inner segments (of photoreceptors)

ISR

integrated stress response

JNK

c-Jun N-terminal kinase

LC

liquid chromatography

LCA

Leber congenital amaurosis

LDL

low-density lipoprotein

LPS

lipopolysaccharide

LRAT

lecithin:retinol acyltransferase

MAM

mitochondria-associated ER membrane

Mfn

mitofusin

Mief1

mitochondrial elongation factor 1

MIMS

mitochondrial intermembrane space

miRNA

micro ribonucleic acid

MNV

macular neovascularization

mRNA

messenger ribonucleic acid

mTORC2

mammalian target of rapamycin complex 2

nAMD

neovascular AMD

NLRP3

NOD-like receptor protein 3

NRF2

nuclear factor erythroid 2-related factor 2

OIR

oxygen-induced retinopathy

OMM

outer mitochondrial membrane

ONC

optic nerve crush

OS

outer segments (of photoreceptors)

PACS2

phosphofurin acidic cluster sorting protein 2

PDR

proliferative DR

PEDF

pigment epithelium growth factor

PERK

protein kinase RNA-like endoplasmic reticulum kinase

PEST

proline, glutamic acid, serine, and threonine rich region

PIC

pre-initiation complex

PKR

protein kinase RNA

POS

photoreceptor outer segment-

PP1

protein phosphatase 1

PRP

prion protein

PTEN

phosphatase and tensin homolog deleted on chromosome 10

RAV

ribosome-associated vesicles

RBP

RNA binding protein

RD

retinal detachment

RGC

retinal ganglion cell

RIDD

regulated IRE-1 dependent decay

RIDDLE

RIDD lacking endomotif

RMECs

retinal microvascular endothelial cells

ROS

reactive oxygen species

RP

retinitis pigmentosa

RPE

retinal pigment epithelium

SLOS

Smith-Lemli-Opitz Syndrome

STZ

streptozotocin

TC

ternary complex

TMP

tetramethylpyrazine

TNF

tumor necrosis factor

TRAF2

receptor-associated factor 2

TRIB3

tribbles homolog 3 or tribbles pseudokinase 3

TRPV4

transient receptor potential vanilloid 4

uoRF

upstream open reading frames

UPR

unfolded protein response

VDAC

voltage dependent anion channel

VEGF

vascular endothelial growth factor

XBP1

X-box binding protein 1

XBP1s

spliced X-box binding protein 1

XBP1u

unspliced X-box binding protein 1

Data availability

Data will be made available on request.