Sulforaphane recovers cone function in an Nrf2-dependent manner in middle-aged mice undergoing RPE oxidative stress

• Introduction
• Methods
• Results
• Discussion

Abstract

Purpose: Sulforaphane (SFN) is an isothiocyanate derived from cruciferous vegetables that has therapeutic efficacy in numerous animal models of human disease, including mouse models of retinal degeneration. However, despite dozens of clinical trials, the compound remains to be tested as a clinical treatment for ocular disease. Numerous cellular activities of SFN have been identified, including the activation of Nrf2, a transcription factor that induces a battery of target gene products to neutralize oxidative and xenobiotic stresses. As Nrf2 expression and function reportedly decrease with aging, we tested whether the loss of the transcription factor limits the therapeutic efficacy of SFN against retinal degeneration.

Methods: Six- to 8-month-old wild-type and Nrf2 knockout mice were treated with SFN beginning 1 month after ribozyme-mediated knockdown of superoxide dismutase 2 (SOD2) mRNA in the RPE. The impacts of MnSOD (the protein product of SOD2) knockdown and the efficacy of SFN were evaluated using a combination of electroretinography (ERG), spectral domain optical coherence tomography (SD-OCT), and postmortem histology.

Results: SFN restored the ERG photopic b-wave suppressed by MnSOD loss in wild-type mice, but not in the Nrf2 knockout mice. In contrast, ERG scotopic a- and b-wave loss was not restored for either genotype. SFN significantly improved retinal thickness in the Nrf2 knockout mice with MnSOD knockdown, but this was not observed in the wild-type mice. In both genotypes, SFN treatment reduced morphological markers of RPE atrophy and degeneration, although these improvements did not correlate proportionally with functional recovery.

Conclusions: These findings highlight the capacity of SFN to preserve cone function, as well as the potential challenges of using the compound as a standalone treatment for age-related retinal degeneration under conditions associated with reduced Nrf2 function.

1-isothiocyanato-(4R)-(methylsulfinyl) butane (SFN) is a phytochemical present in cruciferous vegetables, including broccoli, radishes, cabbage, and Brussels sprouts, with the highest concentrations found in broccoli sprouts [1]. This isothiocyanate is produced by plants in response to predation via the vesicular release of the hydrolytic enzyme myrosinase, which enters damaged cells and converts the precursor compound glucoraphanin into glucose and SFN [2]. This predation response is also elicited by chewing or heating these vegetables. The relatively high bioavailability of SFN is further promoted via myrosinase produced by microbiota resident in the human intestine [3].

Since the discovery of SFN in 1992 [4], the phytochemical has shown therapeutic efficacy in rodent models of human diseases ranging from arthritis and cancer to chronic obstructive pulmonary disease and autism (reviewed in [5,6]). These promising findings spurred dozens of clinical trials evaluating the compound, but to date, SFN has not received approval from the U.S. Food and Drug Administration (FDA). Confounding factors that may be limiting the clinical applications of SFN include dosing amounts and frequency, as well as formulations [6]. Compounding these challenges is the growing list of cellular proteins and pathways targeted by SFN. These include HDAC inhibition [7-12], induction of mitochondrial fusion [13], inhibition of Phase I metabolic enzymes [14], and induction of apoptosis [15-17]).

SFN has been extensively studied in the context of activating Nrf2, a master anti-stress transcription factor that regulates the expression of Phase II detoxification genes, antioxidant genes, and genes encoding enzymes of anabolic and bioenergetic pathways [18]. SFN stabilizes Nrf2 by thionoacylating multiple cysteines within KEAP1 [19-23], the substrate adaptor of a multisubunit E3 ligase that marks Nrf2 for proteasomal degradation [19,21,24]. SFN modification of KEAP1 liberates the KEAP1-Nrf2 complex from the E3 ligase, thus preventing Nrf2 from being marked for degradation with polyubiquitin chains. The stabilized transcription factor enters the nucleus and induces target gene expression. The protein products of Nrf2 target genes cooperatively maintain and restore redox and proteome homeostasis, making Nrf2 activation by SFN an attractive therapeutic option for diseases linked to unchecked oxidative stress, including those impacting the retina (e.g., dry age-related macular degeneration [dAMD]) [25].

Preserving RPE cell function and integrity during aging is a long-standing treatment goal, as atrophy of this cell layer precipitates or accompanies photoreceptor loss during AMD development [26]. Whether SFN has clinical utility for AMD remains to be established in part because Nrf2 expression and Nrf2-mediated transcription can be compromised with aging. Dysmorphic macular RPE in postmortem AMD eyes have reduced expression of Nrf2 and Nrf2 target gene products [27,28], and the RPE cells of aged mice have impaired Nrf2 signaling [29]. Furthermore, cigarette smoking is the leading environmental risk factor for AMD and impairs Nrf2 signaling in an age-dependent manner [30].

In the present study, we directly tested whether Nrf2 is required for the therapeutic efficacy of SFN in a mouse model of retinal degeneration induced by RPE mitochondrial oxidative stress. This model uses a ribozyme to knockdown manganese superoxide dismutase (MnSOD) expression in the RPE of C57BL/6 mice, inducing pathologies consistent with those observed in AMD despite the lack of a macula in mice [31-33]. MnSOD localizes within mitochondria and dismutates superoxide produced by the electron transport chain into hydrogen peroxide (H₂O₂) and oxygen. The H₂O₂ is subsequently converted into oxygen and water. Suppressing MnSOD expression in the RPE layer leads to oxidative stress that alters RPE and choroidal morphology, increases autofluorescent bis-retinoids in extracellular deposits, and causes disorganization and loss of the RPE and underlying photoreceptors. These deficits reduce vision and ERG amplitudes [31-34]. Complementary work in which the SOD2 locus was genetically ablated from the RPE of either C57BL/6 or BALB/c mice using Cre-lox manipulation revealed a time-dependent progression of changes initiated by RPE oxidative stress. Subsequent pathological features observed included increased lipofuscin-associated autofluorescence, scotopic and photopic ERG deficits, reduced outer nuclear layer (ONL) thickness, increased RPE cell area in the central retina, disorganized mitochondrial cristae, decreased mitochondrial DNA content within the RPE, and compensatory upregulation of glycolytic enzymes, indicative of RPE cells switching from respiration to glycolysis [34,35]. Transition of the RPE to glycolysis deprives photoreceptors of sufficient glucose and underscores the metabolic interdependence and cooperation between the two cell types [36-40]. Similarly, SOD2 gene ablation from muscle decreases oxidative phosphorylation and upregulates the expression of several glycolytic enzymes [41].

In the present study, we used the SOD2 ribozyme knockdown approach in RPE cells to evaluate the efficacy of interventional SFN administration in mitigating disease pathologies in wild-type and Nrf2 knockout (KO) mice. Global Nrf2 KO mice have been valuable for elucidating the contributions of oxidative stress to the pathophysiology of a range of cancers as well as liver, lung, and kidney diseases (e.g., [42-45]). In the context of the RPE and neuroretina, Nrf2 knockout mice model aspects of wet and dry AMD in an age-dependent manner, with phenotypes manifesting primarily after 1 year of age [46]. To amplify these AMD-like phenotypes, investigators have coupled aged global Nrf2 knockout mice (i.e., 12–18 months old) with dietary manipulations (e.g., high-fat or high-glycemic diets) [47,48] or with co-knockout of peroxisome proliferator-activated receptor gamma coactivator-1α (PGC-1α) [49]. In the work reported here, Nrf2 KO mice were fed standard chow, and the studies were completed before mice reached 1 year of age to limit ocular deficiencies caused by absence of the transcription factor.

We report that administering SFN three times weekly to RPE-MnSOD knockdown mice beginning 1 month after SOD2 mRNA-targeting ribozyme injections restored cone function and improved morphological markers of RPE atrophy and degeneration. Notably, this efficacy was dependent on Nrf2 expression. These findings highlight the therapeutic utility of SFN to counter retinal degeneration, as well as potential limitations of the compound in scenarios of compromised Nrf2 expression and function.

To test whether the efficacy of SFN to mitigate retinal degeneration requires Nrf2, we performed the study outlined in Figure 1. Six- to 8-month-old, strain-matched wild-type (WT) and global Nrf2 KO mice in a C57BL/6J background negative for the Rd8 mutation were subjected to baseline ERG and OCT measurements. The animals were then injected subretinally with AAV1-RzSOD2, an adeno-associated virus (AAV) expressing a ribozyme targeting superoxide dismutase 2 (SOD2) mRNA to induce oxidative stress in the RPE [31-33]. The contralateral eye of each animal was not injected. One month after the subretinal injections, ERG and OCT measurements were recorded again. The mice were subsequently administered 50 mg/kg of SFN (or PBS vehicle) by intraperitoneal injection three times per week (Monday, Wednesday, and Friday) for 3 months. Final ERG and OCT measurements were recorded before the mice were euthanized. Mice were euthanized by intraperitoneal injection of xylazine and ketamine (2.5 and 17.5 mg/kg body weight respectively), followed by the cervical dislocation. Postmortem histology of the posterior eye was performed using H&E staining to evaluate the retinal morphology and specific antibodies to detect markers of inflammation. The mice were 10–12 months of age at study completion. In keeping with convention, we use the term “SOD2” when referring to the gene and mRNA species and “MnSOD” when referring to the enzyme.

A comparison of the ERG profiles from the 6- to 8-month-old mice before MnSOD knockdown revealed that relative to the untreated WT mice, the Nrf2 KO mice had modestly decreased scotopic a- and b-waves (Figure 2A,B, respectively), a phenotype previously observed in mice lacking this transcription factor [46], whereas photopic b-waves were comparable between genotypes (Figure 2C). Although there were no significant ERG differences detected between the left (OS) and right (OD) eyes of the WT mice (Figure 2D–F), the Nrf2 KO animals curiously displayed modestly reduced scotopic a-waves in the OS compared to OD eyes (Figure 2D). AAV1-RzSOD2-mediated knockdown (KD) of MnSOD enzyme expression induced a significant reduction in all ERG waves for the WT and Nrf2 KO mice 1 month after ribozyme subretinal injection (Figure 2G–I). The extent of the ERG decreases was comparable between genotypes (Figure 2G–I), indicating that the absence of Nrf2 did not amplify the functional consequences of MnSOD suppression in the RPE. Three months of SFN treatment conferred no significant improvement on the scotopic a- and b-waves in the SOD2 kDa eyes for either the WT or Nrf2 KO mice (Figure 2J–K) but significantly improved the photopic b-waves in the WT-SOD2 kDa eyes (Figure 2L). This photopic b-wave recovery required Nrf2 expression, as a comparable improvement in cone function did not occur in the Nrf2KO-RzSOD2-SFN mice (Figure 2L).

In parallel with the longitudinal ERG measurements, SD-OCT was performed to track morphological integrity in vivo. Comparing untreated 6- to 8-month-old WT and Nrf2 KO mice before any treatments, we did not observe significant differences in retinal thickness between genotypes (Figure 3A) or between left and right eyes within a genotype (Figure 3B). However, we observed modest RPE disruptions in a subset of Nrf2 KO mice (Figure 3C versus 3D, arrows). One month after ribozyme-mediated SOD2 mRNA KD, we observed the expected retinal thinning in the WT and Nrf2 KO mice (Figure 3E), and this thinning was exacerbated by 12% in the Nrf2 KO animals. Three months of SFN treatment conferred no significant protection on MnSOD KD retinas in the WT animals (Figure 3F). In contrast, SFN modestly recovered the retinal thickness in the Nrf2 KO-MnSOD KD retinas (Figure 3F). Similarly, SFN improved the retinal thickness in the uninjected eyes of the Nrf2 KO animals, with statistics bordering on significance (Figure 3F, p=0.051).

Complementary postmortem histological analyses of retinas from 10- to 12-month-old mice were performed on four sets of mice for each genotype [1]: untreated [2], SFN treatment in the absence of MnSOD KD [3], vehicle treatment following MnSOD KD, and [4] SFN treatment following MnSOD KD. Analysis of the untreated eyes revealed comparable retinal morphology (Figure 4A, top and bottom panels). The WT and Nrf2 KO mice treated with SFN in the absence of MnSOD KD had mild morphological defects (Figure 4B). Specifically, in the WT mice, the RPE layer consistently showed patchy areas of atrophy (Figure 4B, top panels, red arrows), whereas the RPE layers in some Nrf2 KO mice contained vacuoles (Figure 4B, bottom panel, red arrows). The WT mice subjected to MnSOD KD and treated with vehicle displayed gross disorganization of the neural retina, retinal thinning, severe atrophy of the RPE layer, and the presence of subretinal cellular infiltrates likely representing infiltrating inflammatory cells (Figure 4C, top panel). Likewise, the Nrf2 KO mice subjected to MnSOD KD and treated with vehicle exhibited highly disorganized neural retina morphology, severe retinal thinning and RPE atrophy, and extensive subretinal infiltrates accompanied by fibrosis (Figure 4C, bottom panel, red arrows). Treatment of the WT mice with SFN following MnSOD KD improved the morphology of the retina and the RPE compared to vehicle treatment, although the morphology was not comparable to baseline levels (Figure 4D, top panel). Similarly, SFN partially preserved the retinal and RPE morphology in the Nrf2 KO mice administered SFN after MnSOD KD, although RPE atrophy, infiltrates, and fibrosis were still detectable (Figure 4D, bottom panel).

Transverse retinal sections were labeled for the inflammatory markers Iba1 and F4/80. Iba1 staining was minimal in untreated animals but significantly increased in both genotypes after MnSOD KD, with increased numbers of ramified microglia in the inner retina and subretinal space (Figure 5A). This was markedly reduced in the MnSOD KD mice following SFN treatment. Staining for the macrophage marker F4/80 revealed a similar profile with maximum staining confirming macrophage infiltration in the WT and Nrf2 KO-MnSOD KD mice, which was reduced by SFN treatment (Figure 5B).

This study tested the capacity of SFN to restore retinal integrity and function in mice following oxidative stress–induced damage to the RPE layer induced by the ribozyme-mediated knockdown of MnSOD in the RPE. This model increases mitochondrial superoxide levels, which yield dAMD-like pathologies [31-35]. SFN treatment was initiated 1 month after delivery of an SOD2-targeting ribozyme to evaluate the interventional utility of this phytochemical during active disease and to complement previous studies testing the prophylactic efficacy of the compound (e.g., [51-53]). SFN was administered three times a week for 3 months such that the study was terminated 4 months after ribozyme injection, a time point at which AMD-like structural and functional pathologies from MnSOD knockdown are readily detectable [31]. This study further compared the efficacy of SFN in WT mice versus Nrf2 KO mice, as SFN activates Nrf2, but age-dependent reductions of Nrf2 in the RPE [28,29] and other cell types (e.g., [30]) may limit the clinical feasibility of exploiting Nrf2 activation in elderly patients.

SFN restored the ERG photopic b-wave deficit induced by MnSOD loss in a Nrf2-dependent manner (Figure 2L), consistent with unchecked oxidative stress within the RPE causing functional deficiencies in photoreceptor cones (e.g., [34,35]) and Nrf2 target gene products neutralizing this stress sufficiently to recover cone function [54,55]. This functional rescue is promising, as cone loss associated with oxidative stress is characteristic of AMD [56,57] and retinitis pigmentosa (RP) [58]. In contrast, SFN did not recover rod function (Figure 2J,K). Cones are likely more vulnerable to oxidative damage than rods due to the higher mitochondrial content [58] and higher sensitivity to the mitochondrial toxin, FeSO₄ (e.g., [59]). The elevated mitochondrial content of cones may be accompanied by a more robust antioxidant capacity to neutralize endogenous free radicals generated by mitochondrial respiration, conferring on cones higher resistance to oxidative stress originating in neighboring RPE cells and rods. The differential sensitivity between rods and cones to exogenous stresses has been observed in diabetic mice; rod a- and b-wave ERGs decreased by 50%, whereas cone a- and b-wave ERGs were not deleteriously impacted [60]. In addition, Nrf2 overexpression selectively rescued cones in several mouse models of retinal degeneration (rd) [55], as did antioxidant supplementation [61], and greater functional rescue of cones by SFN has been reported for the rd10 model of RP [62].

Rod photoreceptors may incur irreparable damage from chronic oxidative stress, consistent with scotopic ERG deficits in untreated Nrf2 KO mice (Figure 2A,B), whereas cone photopic ERGs were comparable between genotypes (Figure 2C). Zhao et al. reported a selective reduction in rod function in mice genetically ablated for Nrf2 [46]. The higher susceptibility of rods to Nrf2 loss could also stem from differences in rods and cones with respect to fatty acid compositions [63], metabolic and bioenergetic requirements [39], and autophagic capacity [58,64-66]. Nrf2 activity is required for efficient autophagy and lysosomal function in the RPE [46], and SFN protects RPE cells from blue light–induced inflammation and apoptosis by inducing autophagy [67]. These findings, along with reports that SFN can either induce or inhibit autophagy depending on the cell type and experimental conditions (e.g., [68,69]), highlight an important future direction to determine the extent by which SFN differentially impacts the autophagic machinery in rods versus cones. The balance between autophagy and proteasomal degradation in photoreceptors governs the capacity of these cells to retain function in response to stress (e.g., [70]). As SFN can increase lysosomal and mitochondrial biogenesis (e.g., [68,71-73], we favor that the benefit to cones stems from SFN concomitantly increasing the removal of damaged mitochondria by mitophagy while inducing the biogenesis of replacement mitochondria.

SFN mitigates retinal degeneration in rodent models of ischemia/reperfusion [51,74,75], light damage [52,53], genetic models [62,76], and diabetes [60,77,78]. The compound improves ERGs [52,53,62,74,76,78], reduced retinal thinning [74,75], and suppressed proinflammatory cytokines [51,77]. These positive effects have been consistently reported and include studies that initiated SFN treatment before the onset of pathologies (e.g., [51-53]. In this context, SFN preserves rod ERG function when administered before light-induced retinal damage [52,53], implying that protecting rods depends on mitigating the initial (photo)oxidative insult before damage.

Despite failing to recover rod function in Nrf2 knockout mice with MnSOD knockdown, SFN modestly increased retinal thickness in these mice (Figure 3F) and improved the morphological markers of RPE atrophy and retinal degeneration in both genotypes (Figure 4D). However, these improvements did not correlate proportionally with functional recovery in either genotype. Incomplete morphological and functional rescue has been observed with other therapeutic interventions (e.g., [79,80]), likely reflecting irreparable damage of photoreceptors during the first month of MnSOD knockdown before SFN treatment. Further protection may be achievable if SFN treatment is initiated simultaneously with ribozyme delivery.

This study has potential caveats and limitations. First, control mice injected with an AAV-inactive ribozyme were not included due to the limited number of Nrf2 KO animals available because of breeding issues and reduced litter sizes. Second, MnSOD expression can be induced as a direct Nrf2 target gene product, as most convincingly demonstrated by the Nrf2 antibody chromatin immunoprecipitation (CHIP) approach in cells treated with astragaloside [81]. Constitutive MnSOD expression, however, is independent of Nrf2 (e.g., [82]). Numerous other transcription factors regulate MnSOD expression, including NF-kB, specificity protein 1 (SP1), AP-2, and C/EBP (reviewed in [83], raising the possibility that the efficacy of SFN in wild-type mice results from Nrf2-mediated transcription of SOD2 mRNA that counters the efficiency of the ribozyme. However, the evidence does not favor this scenario because the activity of the ribozyme was not detectable during the time period that SFN was administered, as indicated by a plateauing of the ERG deficits between the first and fourth months after ribozyme delivery in vehicle-treated mice, irrespective of genotype (Appendix 1). Moreover, labeling of the RPE in paraffin sections with an MnSOD antibody did not show a detectable change in enzyme expression in response to SFN compared to RPE cells from vehicle-treated eyes (Appendix 1). As SFN treatment was not initiated until 1 month after ribozyme administration when the ERG deficits plateaued, the improved photopic ERG readings in the WT mice are more consistent with the compound restoring cone function independent of replenishing MnSOD expression in the RPE.

An additional consideration is that the SOD2 mRNA knockdown is targeted to the RPE layer, but SFN was delivered systemically. Thus, the impacts of SFN are not restricted to the RPE; however, the individual relative contributions of each cell type (e.g., RPE, photoreceptors, or M ü ller cells) to the overall efficacy of SFN are unknown. These relative contributions depend on [1] the expression levels of Nrf2 and other factors activated or inhibited by SFN (e.g., histone deacetylases) [2, 7-12], the susceptibility and responses of each cell type to oxidative stress and mitochondrial dysfunction, and [3] the metabolic flexibility and bioenergetic needs of each cell type. Notably, the MnSOD levels in the inner segments of photoreceptors were not detectably changed by SFN in the WT mice but trended higher in the Nrf2 KO mouse inner segments (Appendix 1). However, functional rescue was observed only in the WT animals, indicating that MnSOD induction in photoreceptors is not likely a major contributor to SFN efficacy in this study.

Extrapolating the in vivo impacts of SFN on the mouse retina to human AMD is complicated by the lack of a macula in rodents. Thus, whether SFN exerts similar protective effects on cones in the macula or fovea cannot be inferred from these studies. Further, because the photoreceptor subpopulations in this specialized region of the primate retina may have distinct physiologic and bioenergetic requirements [39], the impacts of SFN on each of these cell types may differ from the response of photoreceptors in the peripheral retina of humans or in the mouse retina. The relative levels of Nrf2 transcriptional activity in photoreceptor subpopulations within and surrounding the macula during health and disease is currently unknown, but would be valuable for determining the therapeutic potential of manipulating Nrf2 to ameliorate retinal degeneration [54,55]. An additional caveat is that this study utilizes a genetic knockout model of Nrf2, whereas only a partial Nrf2 deficiency has been reported in human RPE [28]. Therefore, SFN may confer some benefit on cone function for humans retaining partial Nrf2 transcriptional activity in the RPE.

Moving forward, it will be important to determine which cell types mediate SFN efficacy, as well as the mechanisms conferring neuroprotection. We speculate that these findings reflect the differential responses of rods and cones to reduced ATP production by mitochondria in the RPE layer. Genetic ablation of Nrf2 and knockdown of MnSOD compromise mitochondrial integrity and ATP generation [34,41,84]. The suppression of mitochondrial respiration in RPE cells deleteriously impacts photoreceptor survival and function [38], as RPE cells utilize oxidative phosphorylation to spare glucose for photoreceptors (e.g., [34,37,85-87]). Suppressed ATP production by RPE mitochondria induces a cellular switch to glycolysis, thus reducing glucose availability for photoreceptors, as documented in the RPE [34] and muscle [41] following SOD2 gene ablation. Limiting glucose oxidation by photoreceptors further reduces lactate production and secretion for use by the RPE and Müller glia resulting in disrupted bioenergetic homeostasis across the posterior eye (reviewed in [37,85]). Glucose starvation of photoreceptors may adversely affect rods disproportionately compared to cones, as rods have a higher reliance on glycolytic metabolism [88-90] but contain lower glycogen stores than cones [90]. This may account for the preferential restoration of cone function by SFN (Figure 2). Reduced mitochondrial respiration in RPE cells also decreases oxygen consumption, resulting in a local hyperoxia that can be toxic to photoreceptors [91-93].

ATP deficits and mitochondrial dysfunction in RPE cells impact the mammalian target of rapamycin (mTOR) kinase complexes (mTORC1 and mTORC2), which integrate nutrient availability with bioenergetics, protein synthesis, and proliferation (reviewed in [94]). RPE mTOR activity is mediated by ATP levels as well as by mitochondrial (dys)function (e.g., [95,96]), and chronic mTOR signaling in the RPE causes dedifferentiation, leading to photoreceptor and choroid degeneration [96-98]. However, SFN induces mitochondrial biogenesis [71,99,100], a cellular response accompanied by enhanced mTOR activity that may underlie the neuroprotection conferred by the compound [101]. Notably, enhanced mTOR activity protects cones in mouse models of RP [102,103]. Thus, the efficacy of SFN given every other day in the present study could involve periodic elevations of mTOR activity in cones. Additional factors contributing to the efficacy of SFN may involve Nrf2-mediated induction of the rate-limiting enzymes for numerous bioenergetic pathways (e.g., the pentose phosphate pathway, the polyol pathway, the hexosamine pathway, the β-oxidation of fatty acids, or other intermediary pathways of metabolism) [18,84,104].

In conclusion, the findings of this study underscore that SFN has efficacy in maintaining retinal function and morphological integrity but likely requires optimization of the timing of administration and dosing frequency [2], preserves cone function in a Nrf2-dependent manner, and [3] may optimally be used as an adjuvant treatment with other therapies that enhance photoreceptor function. These findings highlight the therapeutic value of SFN and the potential limitations of the compound in ocular cells with compromised Nrf2 activity.