Preventing diabetic retinopathy by mitigating subretinal space oxidative stress in vivo

Abstract

Patients with diabetes continue to suffer from impaired visual performance before the appearance of overt damage to the retinal microvasculature and later sight-threatening complications. This diabetic retinopathy (DR) has long been thought to start with endothelial cell oxidative stress. Yet newer data surprisingly finds that the avascular outer retina is the primary site of oxidative stress before microvascular histopathology in experimental DR. Importantly, correcting this early oxidative stress is sufficient to restore vison and mitigate the histopathology in diabetic models. However, translating these promising results into the clinic has been stymied by an absence of methods that can measure and optimize anti-oxidant treatment efficacy in vivo. Here, we review imaging approaches that address this problem. In particular, diabetes-induced oxidative stress impairs dark-light regulation of subretinal space hydration, which regulates the distribution of interphotoreceptor binding protein (IRBP). IRBP is a vision-critical, anti-oxidant, lipid transporter, and pro-survival factor. We show how optical coherence tomography (OCT) can measure subretinal space oxidative stress thus setting the stage for personalizing anti-oxidant treatment and prevention of impactful declines and loss of vision in patients with diabetes.

Introduction

Diabetes impairs vision and causes microvascular disease of the retina that can result in blindness. This diabetic retinopathy (DR) persists as a major health burden despite improved screening and risk factor control. One problem is that current clinical intervention is not started until the appearance of more advanced histopathologic stages of the disease, when modifying the disease trajectory is more difficult. A long-desired goal is to enable effective personalized treatment after the initial diagnosis of diabetes and before the appearance of microvascular histopathology.

Patients with diabetes can appear asymptomatic based on retinal vascular biomarkers but show impaired contrast sensitivity suggesting neuronal dysfunction (Sokol et al., 1985; Trick et al., 1988; Aung et al., 2014; Berkowitz et al., 2015c). Indeed, during the clinically silent phase of DR, photoreceptors and retinal pigment epithelium (RPE), for example, show a host of diabetes-induced metabolic abnormalities, including oxidative stress (MacGregor & Matschinsky, 1985, 1986b, a; Kowluru et al., 1992; Ostroy et al., 1994; Kern & Engerman, 1996; Ostroy, 1998; Lahdenranta et al., 2001; Colantuoni et al., 2002; Phipps et al., 2004; Kim et al., 2005; de Gooyer et al., 2006; Phipps et al., 2006; Garcia-Ramirez et al., 2009; Wong et al., 2011; Du et al., 2013; Dmitriev et al., 2014; Berkowitz et al., 2015c; Du et al., 2015; Kern & Berkowitz, 2015; Liu et al., 2015; Samuels et al., 2015; Berkowitz et al., 2016b; Kern, 2017; Malechka et al., 2017; Roy et al., 2017; Tarchick et al., 2019; Yokomizo et al., 2019). Correction of oxidative stress, caused by poorly-regulated production of free radicals (e.g., superoxide and hydroxyl radicals), is an appealing therapeutic target because many FDA-approved anti-oxidants are available. Indeed, early treatment of diabetes-induced outer retina oxidative stress with exogenous anti-oxidants improves both visual performance and mitigates microangiography in experimental models (Berkowitz et al., 2015e; Kern & Berkowitz, 2015; Berkowitz et al., 2016a).

Neuronal dysfunction, as measured by full-field and multifunctional electroretinography, is apparent before overt retinal microvascular morbidities in many patients with diabetes and in experimental models (Han et al., 2004; Adams & Bearse Jr, 2012; McAnany & Park, 2018; McAnany et al., 2019). However, to our knowledge, clinical electrophysiologic biomarkers are not reliably prognostic for evaluating treatment efficacy, including anti-oxidant therapy. Typically, treating patients with diabetes with anti-oxidants requires educated guesswork regarding dose, timing, and drug combinations since it has not been possible to determine if the selected treatment strategy actually reduces oxidative stress in the retina. Also, many clinical studies implement a one-antioxidant-solves-all approach that may be too simple or started too late to be effective in changing disease outcomes. For these reasons, clinical trials often fail to find clear medical benefits from anti-oxidant therapy.

The above considerations provide the rationale for developing non-invasive methods to optimize anti-oxidant treatment and mitigate the development of later DR in patients. Novel imaging approaches are presented that measure pathogenic diabetes-induced outer retinal oxidative stress, a missing step needed to achieve the goal of personalizing anti-oxidant treatment. First, we describe pioneering evidence for outer retinal oxidative stress in early DR in vivo based on its impairment of rod L-type calcium channel (LTCC) function, which is measured by manganese-enhanced MRI (MEMRI). Then we summarize a non-invasive measure of diabetes-induced retinal-laminae-specific oxidative stress using QUEnch-assiSTed (QUEST) MRI which reveals the photoreceptor-RPE interface as a particular region of interest. Next, we demonstrate how oxidative stress in the photoreceptor-RPE region impacts the hydration status of the subretinal space and thus interphotoreceptor matrix components such as the interphotoreceptor binding protein (also known as interstitial retinoid-binding protein or IRBP), a vision-critical, anti-oxidant, lipid-transport, and pro-survival factor. Finally, we review QUEST OCT, a new translational tool that can immediately help in the early management of anti-oxidant treatment in patients with diabetes with the goal of preventing its progression to DR. Mechanistic questions concerning how oxidative stress in the outer retina leads to microvascular disease (and retinal edema, another common complication of diabetes) are important areas of active investigation but are not a focus of this review.

1). Outer retina LTCC function:

LTCCs are the primary route for calcium entry into retinal neurons and thus having an essential role in vision and DR (see below) (Ko et al., 2007; Krizaj, 2012). Neuronal LTCCs consists of three sub-types (i.e., Cav 1.2, Cav 1.3, and Cav 1.4) with distinct electrophysiologic and pharmaceutical properties (Hasreiter et al., 2014; Simms & Zamponi, 2014). For example, LTCC Cav 1.2 are located in the outer retina (and inner retina) and can be preferentially blocked by d-cis-diltiazem (IC50s of ~45 μM for Cav1.2 L-type calcium channels, ~326 μM for Cav1.3 L-type calcium channels, and ~92 μM for Cav1.4 L-type calcium channels) (Berkowitz et al., 2018a). The Cav1.3 channels have been documented in cells of the inner retina and in RPE cells based on electrophysiological evidence, and their mRNA exists in rod cells and cells of the inner retina (Xu et al., 2002; Habermann et al., 2003; Morgans et al., 2005; Wu et al., 2007; Xiao et al., 2007; Berkowitz et al., 2015f; Shi et al., 2017). However, the functional role of Cav1.3 channels in the retina is unclear, in part because current antibodies for identifying these channels and used for immunological localization are nonspecific and a selective antagonist has not yet been established; also knocking out Cav1.3 channels does not seem associated with a visual phenotype as measured by electrophysiology and water maze testing perhaps because of a compensatory upregulation of Cav1.2 LTCCs (Xu et al., 2002; Habermann et al., 2003; Morgans et al., 2005; Wu et al., 2007; Xiao et al., 2007; Berkowitz et al., 2015f; Shi et al., 2017). The best studied LTCCs in the outer retina are the Cav1.4 subtype which regulate release of neurotransmitters at the first retinal synapse. Mutations in the CACNA1F gene which encode Cav1.4 LTCC α1 subunits are linked with loss of vision in Åland Island eye disease, cone-rod dystrophy, X-linked retinal disorder, night blindness–associated transient tonic down-gaze, and incomplete congenital stationary night blindness (Morgans et al., 2001; Baumann et al., 2004; Morgans et al., 2005). Importantly, oxidative stress can cause Cav1.2 LTCCs to become dysfunctional (Yang et al., 2013; Muralidharan et al., 2017).

LTCCs are also a main route into cells by manganese ion (a paramagnetic MRI contrast agent) which then accumulates in an activity-dependent manner (Drapeau & Nachshen, 1984; Carlson et al., 1994; Piggott et al., 2006; Berkowitz, 2011; Rich et al., 2015). In the MEMRI measurement, a non-toxic amount of paramagnetic manganese is administered systemically to awake and freely moving animals. The content of manganese that accumulates in retinal laminae over several hours is measured later in anesthetized animals based on local increases in 1/T1 (=R1) (Berkowitz et al., 2016a). MEMRI has the spatial resolution and detection sensitivity to non-invasively measure, for example, healthy photoreceptor LTCC function in vivo linked to i) phototransduction, ii) the visual cycle, iii) inhibitory feedback from horizontal cells, and iv) melanopsin activity; in addition, manganese uptake is responsive to optogenetic manipulation (Berkowitz, 2006; Berkowitz et al., 2008; Ivanova et al., 2010; Berkowitz et al., 2015f; Muir et al., 2015; Berkowitz et al., 2016d; Kubota et al., 2019). MEMRI also measures abnormal LTCC function and treatment response in various retinopathies (reviewed (Berkowitz et al., 2016b)). For these reasons, MEMRI has been called the imaging modality of choice for measuring rod LTCC function in vivo and can be performed in humans (Tofts et al., 2010; Ramos de Carvalho et al., 2014; Sudarshana et al., 2019).

1.a.). Outer retina LTCCs and oxidative stress in DR:

In non-diabetic rats and mice, dark-adaptation causes outer retina (i.e., outer nuclear, inner segment, and outer segment layers) cyclic nucleotide channels and LTCCs to open, (i.e., a dark phenotype), and light causes them to close (i.e., a light phenotype) (Berkowitz et al., 2006; Berkowitz et al., 2007; Berkowitz et al., 2009a; Berkowitz et al., 2009b; Berkowitz et al., 2015e; Giordano et al., 2015; Muir et al., 2015; Saliba et al., 2015). In contrast, in diabetic rats and mice, dark-adaptation causes less manganese accumulation indicative of a closed, light-like phenotype (summarized Table 1) (Berkowitz et al., 2007; Berkowitz et al., 2009a; Berkowitz et al., 2015e; Giordano et al., 2015; Muir et al., 2015; Saliba et al., 2015).

Table 1.

In vivo evidence for outer retinal oxidative stress in diabetes (chronological order)

	Method	Species	Year
1) α-lipoic acid corrects impaired outer retina LTCC function	MEMRI	Rat	2007
2) Cu/Zn superoxide dismutase overexpressor mice do not show impaired outer retina LTCC function	MEMRI	Mouse	2009
3) Systemic retinaldehyde treatment corrects impaired outer retina LTCC function	MEMRI	Mouse	2012
4) Peroxisome-targeted catalase corrects impaired outer retina LTCC function	MEMRI	Mouse	2015
5) Photobiomodulation corrrects impaired outer retina LTCC function	MEMRI	Mouse	2015
6) α-lipoic acide corrrects impaired light-dependant subretinal space expansion	MEMRI	Mouse	2015
7) Systemic retinaldehyde treatment corrects oxidative stress and light-dependant subretinal space expansion	ADC MRI, OKT	Mouse	2015
8) Anti-oxidant reduction in subretinal space production of excessive free radicals	QUEST MRI	Mouse	2015

Importantly, the subnormal outer retina manganese uptake in dark-adapted diabetic animals was brought back to non-diabetic levels by anti-oxidant treatments including free radical scavengers (e.g., α-lipoic acid), organelle-targeted genetic therapy (e.g., catalase in peroxisomes), and photobiomodulation (anti-oxidant mechanism unclear) (Table 1) (Berkowitz et al., 2007; Berkowitz et al., 2009a; Berkowitz et al., 2015e; Giordano et al., 2015; Muir et al., 2015; Saliba et al., 2015). Inner retina LTCCs also show the above pattern but it is unclear if this is due to a downstream (indirect) consequence of outer retinal lesions or to a direct effect by diabetes (Berkowitz et al., 2007; Berkowitz et al., 2009a; Berkowitz et al., 2015e; Giordano et al., 2015; Saliba et al., 2015). Nonetheless, the above results were the first to show diabetes-induced oxidative stress in the outer retina in vivo.

The underlying mechanisms by which oxidative stress impairs LTCC function are unclear. Oxidative stress can alter Cav1.2 LTCCs by direct redox modification, although induced acidosis may also be a contributor (Tsai et al., 1997; Yang et al., 2013; Muralidharan et al., 2017; Zhang et al., 2018). We have recently demonstrated that the outer nuclear layer (ONL) (which is dominated by rod nuclei in mice (Carter-Dawson et al., 1978)) contains Cav1.2 LTCCs as measured by MEMRI following agonist Bay K 8644 in C57BL/6J mice and in Cav1.2 L-type calcium channel BAY K 8644-insensitive mutant mice; the ONL may also contain Cav1.3 LTCCs but this has been difficult to clearly establish in mammals (Xu & Lipscombe, 2001; Berkowitz et al., 2018a). It does not appear that diabetes-induced oxidative stress has a similar effect on photoreceptor Cav1.4 LTCCs. If diabetes impaired Cav1.4 LTCC function, we would anticipate a supernormal manganese uptake in dark-adapted photoreceptors (Berkowitz et al., 2007; Berkowitz et al., 2009a; Berkowitz et al., 2015f). This is predicted because dark-adapted Cav1.4 knockout mice would lack horizontal inhibitory signaling back to the photoreceptors that would, in turn, result in a greater-than-normal photoreceptor manganese uptake (Berkowitz et al., 2015f). Instead, a less-than-normal uptake in rod cells is measured in diabetic rodents suggesting that the Cav1.4 LTCCs are operating normally (Berkowitz et al., 2007; Berkowitz et al., 2009a; Berkowitz et al., 2015f). Next we consider the impact of diabetes on photoreceptor outer segment function.

1.b.). Phototransduction lesions and diabetes-induced oxidative stress:

Manganese influx into rod photoreceptors can also occur via cyclic nucleotide-gated channels (Piggott et al., 2006; Rich et al., 2015). Outer segments of rod photoreceptor cells have cyclic nucleotide-gated channels regulated by phototransduction. Intriguingly, levels of α-transducin1 (Gnat1), a key part of the phototransduction pathway, are reduced in experimental diabetes (Kowluru et al., 1992; Kim et al., 2005). As first reported at ARVO in 2015, non-diabetic dark-adapted Gnat1^−/− mice demonstrate diabetes-like outer retina-specific oxidative stress, and transretinal manganese uptake was lower-than-normal mirroring that in dark-adapted diabetic mice; rod calcium levels were also greater-than-normal (Gonzalez-Fernandez et al., 1985; Berkowitz et al., 2015b; Berkowitz et al., 2015g; Liu et al., 2019). Non-diabetic wildtype mice maintained in the dark for 2 months (i.e., in the absence of phototransduction) also show retinal oxidative stress (Liu et al., 2019). In addition, non-diabetic Gnat1^−/− mice demonstrated microvascular histopathology similar to that found in diabetic mice and that damage was not exacerbated by diabetes (Liu et al., 2019). We note that the outer retina oxidative stress in Gnat1^−/− mice likely preceded the induction of diabetes and could have preconditioned the retina (Gargioli et al., 2018). Thus, interpretation of a potential benefit of inhibiting phototransduction in murine diabetic retinopathy is ambiguous since the oxidative stress and vascular degeneration develop independently of diabetes in Gnat1^−/− mice. In diabetic mice, systemically administered 11-cis-retinaldehyde, a key component of the visual pigment rhodopsin, inhibited retina oxidative stress and (as measured by a lucigenin assay) and improved diabetes-induced outer retinal LTCC dysfunction (Berkowitz et al., 2012; Berkowitz et al., 2015e). Electrophysiology transretinal recordings in excised tissue also found a diabetes-induced reduction in dark-adapted rod photo-responses that could be corrected with systemic retinaldehyde treatment, supporting the MEMRI results (Berkowitz et al., 2015e).

1.c.). Visual cycle lesions and diabetes-induced oxidative stress:

Diabetes also reduces rhodopsin regeneration via the visual cycle (Kowluru et al., 1992; Ostroy et al., 1994; Kim et al., 2005; Kern & Berkowitz, 2015; Liu et al., 2015; Malechka et al., 2017). We find that in non-diabetic rd12 mice, which have a loss of function in RPE65 (the key isomerase of the visual cycle located in the RPE), LTCC function is impaired only in the outer retina and is corrected with systemic 11-cis-retinaldehyde (Berkowitz et al., 2009b; Berkowitz et al., 2012; Berkowitz et al., 2015e). In diabetic C57BL/6J mice, systemic 11-cis-retinaldehyde fixed the retinal oxidative stress, corrected the lower-than-normal uptake of manganese in the outer retina and the dark-evoked shrinkage of the subretinal space (see below), and improved visual performance (Berkowitz et al., 2009b; Berkowitz et al., 2012; Berkowitz et al., 2015e)

Together, the above considerations raise the possibility that diabetes-induced oxidative stress, and its consequences on photoreceptor cyclic nucleotide channels, LTCCs, visual performance, and histopathology, have an etiology involving phototransduction and the visual cycle.

2). Outer retina oxidative stress and diabetes:

2.a.). Diabetes-induced outer retina oxidative stress ex vivo:

Consistent with the above evidence, retinal cryosections stained with dichlorofluorescein (DCF, which fluoresces when it interacts with reactive oxygen species) show that the photoreceptor outer segments and RPE in diabetic models are the primary sites of excessive production of free radicals (i.e., oxidative stress, Table 2), including (Du et al., 2013; Liu et al., 2019). Together, these data provide additional independent support for the hypothesis that the originating site for diabetes-induced oxidative stress is in the outer retina in experimental models.

Table 2.

Ex vivo evidence for outer retina oxidative stress in diabetes (chronological order)

	Method	Species	Year
1) Melantonin corrects diabetes-impaired light-induced oxidative response in photoreceptors	DHR123 fluorescence	Syrian hamster	2002
2) Histochemical evidence demonstrating outer retina oxidative stress	DCF, DHE	Mouse	2013, 2015, 2019

To test this hypothesis directly in vivo, we first note that continuously produced free radicals, such as superoxide and nitric oxide, generate a localized region of paramagnetic relaxation and thus a contrast mechanism that is detectable with R1 MRI (Berkowitz et al., 2015a; Berkowitz, 2018). A simple R1 measurement cannot be used as an unambiguous biomarker of excessive free radical production because R1 is affected by many factors such as temperature and oxygen levels. Instead, we measure R1 before and after giving anti-oxidants; reduction of R1 is a necessary and sufficient index of oxidative stress in vivo (Berkowitz et al., 2016c; Berkowitz, 2018). This QUEst-assiSTed (QUEST) MRI index has been extensively tested in animal models and confirmed against several ex vivo assays (Berkowitz, 2018).

2.b.). Diabetes-induced outer retina oxidative stress in vivo:

QUEST MRI studies in vivo show a diabetes-duration–dependent oxidative stress that is specifically localized to the subretinal space (i.e., photoreceptor outer segment - RPE region) before overt histopathology is evident (Berkowitz et al., 2015a). This diabetes photoreceptor – RPE oxidative stress phenotype does not seem to arise from damage to the RPE alone. In non-diabetic mice exposed to an RPE-specific toxin before degeneration (sodium iodate), or in mice with an RPE-specific genetic manipulation that produces oxidative stress, a more spatially extensive pattern of oxidative stress across photoreceptors is observed (Berkowitz et al., 2015a; Berkowitz et al., 2016c). In addition, the above QUEST MRI and DCF data suggest that the impact of diabetes on LTCCs (as measured by MEMRI) are downstream of oxidative stress in the subretinal space.

3). Functions of the subretinal space:

The subretinal space refers to an extracellular region filled with an interphotoreceptor matrix that is bounded posteriorly by apical RPE and anteriorly by the external limiting membrane (ELM) (Mieziewska, 1996; Gonzalez-Fernandez, 2003). Hydration of the subretinal space is regulated by the RPE and not by Müller cells (Reichenbach et al., 2007). In non-diabetic subjects, the subretinal space becomes significantly dehydrated causing shrinkage of the ELM-RPE region in the dark compared to that in the light as measured by microelectrodes, diffusion MRI, and OCT in a variety of species including humans (Huang & Karwoski, 1992; Govardovskii et al., 1994; Li et al., 1994a; Li et al., 1994b; Wolfensberger et al., 1999; Adijanto et al., 2009; Bissig & Berkowitz, 2012; Berkowitz et al., 2015c; Lu et al., 2017; Berkowitz et al., 2018b).

This dark-light volume difference changes the concentration and distribution of components within the interphotoreceptor matrix that are essential for vision (Govardovskii et al., 1994; Wolfensberger et al., 1999; Adijanto et al., 2009; Berkowitz et al., 2015c; Berkowitz et al., 2016b; Berkowitz et al., 2018b). Our working model of the physiology underlying the subretinal space photo-response consists of a signaling pathway involving phototransduction / ion channel function / mitochondrial reserves / pH / and RPE water co-transporters (discussed below; summarized in Figure 1) (Govardovskii et al., 1994; Wolfensberger et al., 1999; Adijanto et al., 2009; Berkowitz et al., 2015c; Berkowitz et al., 2016b; Berkowitz et al., 2018b). In more detail, rods use more energy in the dark than in the light, and there is an associated increase in production of waste water and CO2. In turn, this leads to acidification of the subretinal space and upregulation of pH-sensitive water-removal co-transporters on the apical portion of the RPE (Huang & Karwoski, 1992; Li et al., 1994b; Wolfensberger et al., 1999; Adijanto et al., 2009; Berkowitz et al., 2016a; Li et al., 2016; Majdi et al., 2016; Zhang et al., 2017; Berkowitz et al., 2018b; Berkowitz et al., 2019). Experimental studies support the above signaling pathway outlined in Figure 1. In wild-type mice, a larger ELM-RPE region in the light can be converted to a smaller dark phenotype in vivo by: (i) blocking phototransduction, (ii) increasing mitochondrial respiration with a protonophore, or (iii) inducing acidification with a carbonic anhydrase inhibitor ((Wolfensberger et al., 1999; Berkowitz et al., 2016a; Berkowitz et al., 2018b). Notably, oxidative stress can cause cellular acidification, an early event in the diabetic rat retina (Mulkey et al., 2004; Dmitriev et al., 2014; Majdi et al., 2016). Based on the evidence for the above signaling pathways, we feel that “dark-evoked shrinkage of the subretinal space” is a more accurate description of the underlying physiology than the notion of a “light-evoked expansion of the subretinal space” (Figure 1).

Figure 1:

Working model that outlines how dark causes shrinkage of the ELM-RPE region in non-diabetic subjects. In contrast, diabetes-induced oxidative stress is hypothesized to induce an acidosis (dotted red line) that would be expected to convert a thicker “light” ELM-RPE phenotype into a thinner “dark-like” phenotype (see text for details); this expectation is supported experimentally.

The subretinal space volume change between dark and light was first measured in vivo in rats and mice by diffusion MRI, which evaluated water self-diffusion / mobility as a reflection of the number of barriers encountered in the direction of a diffusion-weighting magnetic field gradient. In 2012, we demonstrated significant subretinal space-only dark-stimulated decreases in water self-diffusion (i.e., reduced mobility and more barriers encountered than in the light) limited to the direction parallel with the major rod cell axis (i.e., axially) (Bissig & Berkowitz, 2012). No dark-based shrinkage was observed in the absence of rod phototransduction (Gnat^−/− mice) or in wildtype mice made acidotic; a subsequent OCT study supported the key role of phototransduction in the dark-evoked contraction of the subretinal space (Wolfensberger et al., 1999; Bissig & Berkowitz, 2012; Berkowitz et al., 2015c; Berkowitz et al., 2016a; Zhang et al., 2017).

3.a.). Oxidative stress impairment of subretinal space hydration in diabetes as measured by diffusion MRI:

Diabetes caused outer retina water mobility in the dark and light to be similar as measured by diffusion MRI; anti-oxidant treatment restored a dark-light volume change function (Berkowitz et al., 2015c; Berkowitz et al., 2019). In addition, systemic retinaldehyde treatment, which has anti-oxidant properties, also corrected the diabetes-induced, dark-dependent, subretinal space-specific lesion (and improved the outer retina LTCC dysfunction) (Berkowitz et al., 2012; Berkowitz et al., 2015e). In summary, these results are consistent with diabetes-induced oxidative stress / acidification of the subretinal space being sufficient for converting a thicker “light” ELM-RPE phenotype to a thinner “dark” phenotype.

Diabetes-induced dysfunction in subretinal space hydration can be linked to other metabolic lesions in this retinal region. For example, a major component in the subretinal space is IRBP (human gene designation RPB3), which is thought to have a half-life in the space of about 11 hours (Uehara et al., 1990; Gonzalez-Fernandez, 2003). IRBP is a retinoid chaperone molecule that contributes to photoreceptor function as an integral part of the visual cycle and also plays an important role as a pro-survival factor with anti-oxidant properties (Uehara et al., 1991; Gonzalez-Fernandez, 2003; Du et al., 2013; Sun et al., 2016; Chen et al., 2017). Notably, the distribution of the IRBP within the subretinal space normally changes dramatically between dark and light conditions in-line with the hydration changes discussed above (Uehara et al., 1990; Huang & Karwoski, 1992; Yamamoto & Steinberg, 1992; Govardovskii et al., 1994; Li et al., 1994a; Wolfensberger et al., 1999; Bissig & Berkowitz, 2012; Berkowitz et al., 2015d; Berkowitz et al., 2016a; Li et al., 2016). Several reports find that ocular IRBP levels are reduced in patients with diabetes and in diabetic mice (Garcia-Ramirez et al., 2009; Malechka et al., 2017; Yokomizo et al., 2019), and patients with diabetes and elevated RBP3 are protected against DR (Yokomizo et al., 2019). We speculate that a diabetes-related decrease in the anti-oxidant IRBP promotes oxidative stress and dark-light dysfunction in subretinal space volume. In turn, the distribution of IRBP in the subretinal space becomes abnormal, contributing to impaired visual performance.

We speculate that the observed subretinal space oxidative stress early in the course of DR has contributions from a decrease in IRBP and possibly 11-cis-retinaldehyde anti-oxidant defenses discussed above, together with the de novo appearance of pro-oxidant activated microglia and macrophages in the subretinal space (Roque et al., 1996; Zeng et al., 2000; Ng & Streilein, 2001; Harada et al., 2002; Naskar et al., 2002; Marella & Chabry, 2004; Zhang et al., 2005; Davies et al., 2006; Combadiere et al., 2007; Xu et al., 2008; Zeng et al., 2008; Santos et al., 2010; Omri et al., 2011; Kezic et al., 2013). Interestingly, transplanted bone marrow from non-diabetic mice to diabetic mice prevents the development of experimental DR (Kanter et al., 2012; Aredo et al., 2015; Bhatwadekar et al., 2017; Rivera et al., 2017; Sharma et al., 2018; Mei et al., 2019; Mellal et al., 2019; Yauger et al., 2019). More work is needed to determine how immunologic cells (which are not normally present in the subretinal space of young mice) modify the dark - light hydration dynamics of the subretinal space. Also, questions remain regarding how diabetes-activated microglia and IRBP interact in the subretinal space.

4). Oxidative stress impairment of subretinal space hydration as measured by QUEST OCT:

The above QUEST diffusion MRI findings underscore a pathogenic role of diabetes-induced oxidative stress in the subretinal space. So far, diffusion MRI has not been tested in humans. On the other hand, the dark – light change in hydration of the subretinal space can also be measured by OCT (Berkowitz et al., 2018b). This raised the possibility that a QUEST OCT study could measure oxidative stress in the subretinal space by comparing the size of the subretinal space (i.e., ELM-RPE thickness) in groups given either saline or anti-oxidants (Berkowitz et al., 2018a; Berkowitz et al., 2019). We tested this hypothesis with systemic d-cis-diltiazem given to dark-adapted mice. d-cis-diltiazem produces a transient and non-damaging oxidative stress response in the outer retina as measured by QUEST OCT and QUEST MRI in vivo, and a lucigenin assay ex vivo (Berkowitz et al., 2018a; Berkowitz et al., 2019). These studies set the stage for future applications of QUEST OCT in a range of oxidative stress–based retinopathies including DR.

5). Cautionary observation about C57BL/6J mice from Jackson Laboratory and DR studies:

Much of the literature on experimental DR models involve studies using C57BL/6J mice from Jackson labs. Yet, these mice have a null mutation in the mitochondria gene NAD nucleotide transhydrogenase (Nnt) (Ronchi et al., 2013). This defect increases vulnerability to oxidative stress and is not found in the general patient population (Meimaridou et al., 2012). We were curious as to how studies of DR in these mice might have been confounded by the null mutation.

To examine this problem we started by comparing non-diabetic C57BL/6J mice from Jackson Laboratory with non-diabetic brown 129S6/ev mice (Taconic Laboratory), which do not carry the Nnt mutation. We found that the RPE-specific toxin sodium iodate produces less rod mitochondrial oxidative stress and rod death in 129S6/ev mice than in C57BL/6J mice (Berkowitz et al., 2017; Berkowitz et al., 2018b). Further, 129S6/ev mice retina have an 11% greater mitochondrial reserve than C57BL/6J mice from The Jackson Laboratory as measured by an ex vivo Seahorse assay (Berkowitz et al., 2018b; Berkowitz et al., 2020). When compared in vivo with those of dark-adapted C57BL/6J mice, rod mitochondria of dark-adapted 129S6/ev mice have a greater capacity to respond to a protonophore-evoked shrinkage of the external limiting membrane - retinal pigment epithelium (ELM-RPE) region as measured by OCT, consistent with the signaling pathway in Figure 1. Briefly, because dark-adapted C57BL/6J mice have a relatively low mitochondrial reserve capacity (Kooragayala et al., 2015; Berkowitz et al., 2018b)), a protonophore is predicted to produce only a small change in respiration, and consequently, the pH change of the ELM-RPE (subretinal space) and the pH-dependent co-transporter water removal are reduced in magnitude (Wolfensberger et al., 1999; Adijanto et al., 2009). Indeed, after a 10 mg/kg intraperioneal injection of the protonophore 2, 4 dinitrophenol (DNP), the ELM-RPE region of 129S6/ev mice is thinner than that of C57BL/6J mice because of the former’s relatively greater mitochondrial reserve capacity (which can respond by increasing respiration), producing more waste water and CO2, lowering pH of the subretinal space, and increasing co-transporter water removal. A similar but smaller effect was seen after 5 mg/kg DNP (not shown). Because these experiments were done in dark-adapted mice, the influence of differences in rhodopsin regeneration between 129S6/ev and C57BL/6J mice due to RPE65 Leu450/Met450 variants is minimized (Lyubarsky et al., 2005).

In summary, the null mutation for Nnt appears to impart a greater difference between dark and light in the thickness of the ELM-RPE region in C57BL/6J mice from Jackson labs compared to mice without the null mutation (e.g., 129S6/ev). In turn, this likely makes detection of oxidative stress more challenging in species without the Nnt mutations like humans with methods like QUEST MRI that do not have the spatial sensitivity of OCT (Berkowitz et al., 2018a; Berkowitz et al., 2018b; Berkowitz et al., 2019). On the other hand, mitochondrial function decreases with age and it is possible that C57BL/6J mice from Jackson Laboratories are an (inadvertent) model of aging (Zhao et al., 2014). More work is needed to compare age-related mitochondrial dysregulation and that produced by the Nnt null mutation.

Conclusions:

The subretinal space is uniquely positioned to facilitate communication between photoreceptors, RPE and Müller cells, and between these cells and the vascular and immune systems (Uehara et al., 1990; Mieziewska, 1996; Jeon et al., 1998; Ng & Streilein, 2001; Gonzalez-Fernandez, 2003; Semenova & Converse, 2003; Johnson et al., 2007; Garcia-Ramirez et al., 2009; Sung & Chuang, 2010; Ishikawa et al., 2015; Chen et al., 2017). As reviewed herein, the major site of oxidative stress in experimental diabetes appears to be in the subretinal space as measured in vivo (QUEST MRI) and ex vivo (DCF stained histology) with downstream consequences on the rod calcium channels (and reduced LTCC function in the inner retina) (MEMRI) (Berkowitz et al., 2016b). This suggest that diabetes-induced oxidative stress in the subretinal space deserves more attention than it has to-date as a major contributor to the pathogenesis of DR. QUEST OCT is a promising and sensitive method for innovatively bridging experimental and clinical realms and prevent patients with diabetes from losing their vision.