Deuterated docosahexaenoic acid protects against oxidative stress and geographic atrophy‐like retinal degeneration in a mouse model with iron overload

Abstract Oxidative stress plays a central role in age‐related macular degeneration (AMD). Iron, a potent generator of hydroxyl radicals through the Fenton reaction, has been implicated in AMD. One easily oxidized molecule is docosahexaenoic acid (DHA), the most abundant polyunsaturated fatty acid in photoreceptor membranes. Oxidation of DHA produces toxic oxidation products including carboxyethylpyrrole (CEP) adducts, which are increased in the retinas of AMD patients. In this study, we hypothesized that deuterium substitution on the bis‐allylic sites of DHA in photoreceptor membranes could prevent iron‐induced retinal degeneration by inhibiting oxidative stress and lipid peroxidation. Mice were fed with either DHA deuterated at the oxidation‐prone positions (D‐DHA) or control natural DHA and then given an intravitreal injection of iron or control saline. Orally administered D‐DHA caused a dose‐dependent increase in D‐DHA levels in the neural retina and retinal pigment epithelium (RPE) as measured by mass spectrometry. At 1 week after iron injection, D‐DHA provided nearly complete protection against iron‐induced retinal autofluorescence and retinal degeneration, as determined by in vivo imaging, electroretinography, and histology. Iron injection resulted in carboxyethylpyrrole conjugate immunoreactivity in photoreceptors and RPE in mice fed with natural DHA but not D‐DHA. Quantitative PCR results were consistent with iron‐induced oxidative stress, inflammation, and retinal cell death in mice fed with natural DHA but not D‐DHA. Taken together, our findings suggest that DHA oxidation is central to the pathogenesis of iron‐induced retinal degeneration. They also provide preclinical evidence that dosing with D‐DHA could be a viable therapeutic strategy for retinal diseases involving oxidative stress.


| INTRODUC TI ON
Oxidative stress plays a major role in the pathogenesis of neurodegenerative and retinal diseases (Shichiri, 2014). The retina is subject to oxidative damage because of free radicals generated by abundant mitochondria, especially when photosensitizers in the mitochondria are exposed to blue light (King et al., 2004).
Bisretinoids produced as byproducts of the visual cycle are also blue light photosensitizers (Sparrow et al., 2000). Polyunsaturated fatty acids (PUFAs), which are abundant in plasma and mitochondrial membranes, are especially vulnerable to oxidative stress, since reactive oxygen species (ROS) initiate a lipid peroxidation (LPO) chain reaction.
The abstraction of bis-allylic hydrogens is the rate-limiting step of ROS-driven PUFA oxidation. Substitution of deuterium atoms for hydrogen atoms at bis-allylic sites can slow down the LPO chain reaction due to an isotope effect (Shchepinov, 2020) (Figure 1). PUFAs cannot be synthesized de novo from carbon sources, for example, acetate. Typically, linoleic acid and alpha-linolenic acid, respectively, serve in the diet as the major precursors for biosynthesis of all the n-6 and n-3 PUFAs . This ensures that D-PUFAs are incorporated into mitochondrial and cellular membranes after oral dosing, replacing a fraction of the PUFAs naturally occurring in membranes, and conferring resistance to oxidative stress and LPO. D-PUFAs have been studied in multiple conditions involving oxidative stress and LPO (Andreyev et al., 2015;Berbée et al., 2017;Hill et al., 2011Hill et al., , 2012. A deuterated version of linoleic acid (11,11-D 2 -Lin, RT001) inhibited LPO and rescued cell death in both animal models and clinical trials in several neurodegenerative diseases, including Friedreich's ataxia (FRDA) (Zesiewicz et al., 2018), infantile neuroaxonal dystrophy (INAD) (Kinghorn et al., 2015), and progressive supranuclear palsy (PSP) (Angelova et al., 2021). D-PUFAs also reduced LPO and hold therapeutic potential in preclinical studies for Alzheimer's (Raefsky et al., 2018), Parkinson's (Beal et al., 2020;Shchepinov et al., 2011), and Huntington's diseases (Hatami et al., 2018).
Oxidative stress has been implicated in several retinal diseases, including age-related macular degeneration (AMD) (Beatty et al., 2000), light-induced damage (Cheng et al., 2019;Tanito et al., 2005), iron-related retinal degeneration (Katz et al., 1993;Shu et al., 2020), Leber's hereditary optic neuropathy (Kirches, 2011), and retinitis pigmentosa (Campochiaro et al., 2015;Komeima et al., 2006;Tuson et al., 2009). Docosahexaenoic acid (cervonic acid; DHA, C22:6, n-3) is the most abundant PUFA in the retina, representing up to 40% of all total fatty acids in human rod photoreceptor outer segments (Fliesler & Anderson, 1983). DHA is crucial for the integrity of photoreceptors and visual function (Benolken et al., 1973). While ingestion of DHA-rich fatty fish is associated with lower AMD risk, n-3 PUFA supplementation has shown no appreciable benefits in patients with AMD (Souied et al., 2013) or retinitis pigmentosa (Hoffman et al., 2014). Moreover, high doses of DHA may increase risk in conditions involving oxidative stress, due to its high sensitivity to oxidation (Tanito & Anderson, 2009). The addition of DHA to the human RPE cell line ARPE-19 (Dunn et al., 1996) increased oxidative stress and LPO under high-intensity light exposure (Liu et al., 2014). Levels of carboxyethylpyrrole (CEP), a protein adduct specifically derived from the oxidation of DHA, are elevated in retinal tissues (Beatty et al., 2000) and plasma (Ardeljan et al., 2011;Ni et al., 2009) from patients with AMD. Furthermore, immunization of mice with CEP adducts led to an AMD-like retinal degeneration (Hollyfield et al., 2008). These pieces of evidence suggest that nonenzymatic oxidation of DHA in the retina may play a crucial role in the pathogeneses of retinal disorders involving oxidative stress.
Iron injection resulted in carboxyethylpyrrole conjugate immunoreactivity in photoreceptors and RPE in mice fed with natural DHA but not D-DHA. Quantitative PCR results were consistent with iron-induced oxidative stress, inflammation, and retinal cell death in mice fed with natural DHA but not D-DHA. Taken together, our findings suggest that DHA oxidation is central to the pathogenesis of iron-induced retinal degeneration. They also provide preclinical evidence that dosing with D-DHA could be a viable therapeutic strategy for retinal diseases involving oxidative stress.

K E Y W O R D S
age-related macular degeneration, deuterium, docosahexaenoic acid, iron, isotope effect, lipid peroxidation, oxidative stress, polyunsaturated fatty acid acid ethyl ester for 11 weeks followed by a washout in a pharmacokinetic study to establish a dosing regimen for efficient retinal uptake.
To determine the protective effect of D-DHA against LPO, we fed mice a diet containing D-DHA for 1-4 weeks before giving an IVT injection of iron or control saline. In mice fed with D-DHA for 4 weeks, >50% substitution of DHA with D-DHA in the neural retina was observed. This regimen provided nearly complete protection against iron-induced retinal damage by inhibiting oxidative stress and DHA oxidation.

| Dietary D-DHA efficiently incorporated into the neural retina and RPE cells
To determine the pharmacokinetics of ocular D-DHA uptake, incorporation, and elimination from the neural retina and RPE/choroid/ sclera, twelve-week-old C57BL/6J mice were fed a 0.5% D-DHA containing diet (0.5 g D-DHA/100g food, Table S1) for 77 days followed by an additional 74-day washout phase with DHA. At 4 weeks, >55% of the DHA in the retina was D-DHA rising to >60% at 5 weeks (Figure 2a). At similar time points, D-DHA in the RPE/ choroid/sclera was >80%. Washout in the RPE/choroid/sclera was similarly more rapid than retina. Uptake and elimination followed classic first-order kinetics. Based on the accretion and elimination data, two-month-old mice were fed with diets containing either D-DHA or natural DHA control for 1, 2, 3, and 4 weeks before the IVT injection of iron and saline control ( Figure 2b). In order to better approximate typical DHA doses in human prescription omega-3 supplements (e.g., 1,500 mg/day DHA in Lovaza), the D-DHA and DHA experimental mouse diets were adjusted to 0.25% instead of 0.5% for most of the study. On a 0.25% D-DHA diet, retinal D-DHA levels exceeded 50% at 5 weeks (52.2 ± 1.5% and 55% of total DHA by our GC-and LC-based MS methods, respectively), regardless of whether eyes were injected with iron or control saline (Table 1).

| D-DHA protected against iron-induced retinal autofluorescence (AF) and degeneration
We previously reported IVT iron-induced retinal AF and degeneration (Liu et al., 2021). To evaluate the protective effect of 0.25% D-DHA diet, confocal scanning laser ophthalmoscopy (cSLO) and optical coherence tomography (OCT) were employed for in vivo imaging at 1 week after injections. For the cSLO imaging, both F I G U R E 1 Chemical diagram of the steps in iron-catalyzed lipid peroxidation of phospholipids containing DHA and the formation of CEP. Fe catalyzes hydroxyl radical generation through the Fenton reaction and Haber-Weiss reaction (a). ROS-driven hydrogen abstraction off bis-allylic sites generates free radicals, which rapidly react with oxygen to form lipid peroxyl radicals (b). These newly formed ROS species then abstract bis-allylic hydrogen atoms from the neighboring PUFAs, thus sustaining the LPO chain reaction cycle (c). D-DHA was used in this study (d). DHA peroxidation generates multiple oxidation products including reactive carbonyls such as HHE and HOHA, which can give rise to protein modifications, including OBA, CEP, and MDA adducts (e). The substitution of deuterium for hydrogen atoms inhibits the ratelimiting step of ROS-driven abstraction off bis-allylic sites F I G U R E 2 D-DHA protected against iron-induced retinal autofluorescence and degeneration. Mice were fed with D-DHA for 77 days, followed by a switch to DHA for 74 days for a total of 151 days of feeding. Graph shows %D-DHA in neural retina and RPE-choroid (a). Timeline shows that mice were fed with either D-DHA or DHA for 1 week, 2 weeks, or 4 weeks beginning at 2 months age, then given intravitreal injection of iron in one eye and control normal saline in the other. Mice were continued on their respective diets until their final evaluation (b). cSLO and OCT imaging were performed at 1 week after IVT iron versus saline injection. Graph shows retinal AF area in BAF cSLO images from mice fed with 4 weeks of DHA or D-DHA at 1 week after iron injection (designated 4+1 wk) (c). Representative BAF cSLO images in mice fed D-DHA or DHA for 1 week, given IVT injections, then euthanized a week later (1+1 wk), or fed D-DHA for two weeks, given IVT injections, then euthanized a week later (2+1 wk), etc. (d and e), IRAF cSLO images (f), and horizontal OCT b scans (g) are shown. SLO, scanning laser ophthalmoscopy; OCT, optical coherence tomography; BAF, blue autofluorescence; IRAF, infrared autofluorescence; ONL, outer nuclear layer. Yellow arrows indicate hyper-AF spots induced by iron; red arrows indicate RPE degeneration. Green lines indicate the position and orientation of horizontal OCT b scans in panel e, yellow stars indicate the vortex vein that was used as a landmark for the corresponding position of the OCT scan in IRAF SLO images. N=10 mice/group in c; Error bars indicate mean ± SEM. (**p < 0.01) blue autofluorescence (BAF) to detect bisretinoids and nearinfrared autofluorescence (IRAF) to detect melanin in the RPE and choroid were performed. Fundus AF in BAF images from mice at 1 week after iron injection was quantified by ImageJ Software The extent to which D-DHA replaced natural DHA was determined by LC/MS in pellets of the control diet containing 0.25% natural DHA, the experimental diet containing 0.25% D-DHA, and microdissected samples of neural retina and RPE from mice fed the experimental diet for 4 weeks, given IVT iron or control saline, and then continued on the experimental diet for another week. We confirmed that 22.6% of DHA in the control diet consisted of isotopomers due to natural abundance of carbon-13 (Raefsky et al., 2018) and no D-DHA. Therefore, signals from the 327.2/283.2 transition represent 78.4% of the total DHA. The experimental diet contained no detectable natural DHA, but rather a distribution of deuteriumsubstituted DHA isotopologues ( Figure S1), which is a consequence of the D-DHA preparation method (Smarun et al., 2017;Wang et al., 2021). After corrections were applied for natural abundance of carbon-13 (Raefsky et al., 2018), we determined that the D 10 -DHA isotopologue comprised 45.6% of the D-DHA species, with D 8 -DHA, D 9 -DHA, D 11 -DHA, and D 12 -DHA isotopologues comprising the balance. Therefore, signals from the 337.2/293.2 transition represent 45.6% of the D-DHA. With these corrections applied, D-DHA isotopologues as a percentage of total DHA (i.e., D-DHA +DHA) were determined to be 59.3-60.8% in isolated RPE, and 55.0% in neural retina. There were no significant differences between iron-treated and saline-control eyes (Table 1).

| D-DHA protected retinal function and histologic structure
Electroretinography was conducted on mice fed with D-DHA or natural DHA for 4 weeks to evaluate retinal function. In mice not given IVT injections, dosing with D-DHA caused no significant difference in the rod-b wave, rod-a wave, and cone-b wave amplitudes compared to those fed with natural DHA. Thus, using this measure, incorporation of D-DHA had no impact on retinal function ( Figure 3a).
In mice fed with control DHA, 1 week after iron injection, the rod b-wave, rod a-wave, and cone-b wave amplitudes were significantly decreased in the eyes injected with iron compared with saline con-

| D-DHA prevented the formation of CEP, a unique oxidation product of DHA
Iron-catalyzed peroxidation of phospholipids containing DHA leads to unique carboxyethylpyrrole (CEP) adducts not formed from any other PUFA. CEP has been detected by IHC in human AMD eyes and mouse retinas, including those from mice receiving IVT iron (Crabb et al., 2002;Ebrahem et al., 2006;Liu et al., 2021). To test whether D-DHA could protect against iron-induced CEP formation, mice were fed with D-DHA or DHA for 4 weeks prior to IVT injection of iron or control saline. Cryosections were prepared at 4 hr and 1 week after injections. Co-labeling for CEP and rhodopsin was conducted to assess and localize CEP. At 4 hr after injection, increased immunolabeling for CEP was present in rhodopsin co-labeled photoreceptor outer segments in IVT iron injected eyes of DHA fed mice Immunolabeling for ferritin light chain (L-Ft) was conducted to assess retinal iron levels and localization, since L-Ft protein levels are increased in response to elevated intracellular iron (Song et al., 2014).
At 1 week after saline injection, L-Ft weakly labeled the ganglion cell layer, outer plexiform layer, and inner segment layers (Figure 4e).

| DISCUSS ION
Herein, we tested the hypothesis that inhibition of DHA oxidation might prevent oxidative stress and retinal degeneration in a mouse model with retinal iron overload. We previously reported F I G U R E 3 D-DHA protected retinal function and structure against iron injection. Graphs showing electroretinography amplitudes 4 weeks after dietary dosing of either D-DHA or DHA (a), then re-conducted at 1 week after an intravitreal injection of iron or saline (b). Toluidine blue staining was conducted on plastic sections prepared at 1 week after injections (c and d). Enlarged image (red box1) is from section from mouse fed with DHA diet for 4 weeks then given IVT iron and euthanized a week later. Spider graphs show the mean thickness of each retinal layer. Error bars indicate mean ± SEM of total retinal thicknesses and outer retina thickness (ONL to RPE) in the ventral (inferior) -dorsal (superior) axis at the positions indicated on the x-axis (e and f). Green arrows indicate RPE degeneration. Red arrows indicate cells that have infiltrated the region between the ONL and RPE; these may be myeloid or migrating RPE cells. Two-sample t-tests were performed to compare the total retinal thickness and outer retinal thickness between DHA-Fe group and D-DHA-Fe group at each different location. All statistical comparisons were made using SAS v9.4 (SAS Institute Inc., Cary, NC). No correction for multiple comparisons was performed due to the exploratory nature of this small study. Error bars indicate mean ± SEM. *p < 0.05. Scale bar: 50 μM. N = 8-10/group for electroretinography. N=3/group for retina thickness measures IVT iron-induced retinal AF, oxidative stress, accumulation of carboxyethylpyrrole (CEP), a DHA-specific oxidation product, and photoreceptor degeneration followed by progressive geographic atrophy, replicating features of human AMD (Liu et al., 2021). In this study, we have found that dosing of D-DHA completely protected against all these iron-induced retinal changes.
We found that mice fed with D-DHA for 1, 2, and 3 weeks prior to the iron injection showed a dose-dependent reduction in iron- forming oxidation product (Organisciak et al., 2012). CEP adducts have been found increased in drusen deposits (Crabb et al., 2002) and plasma (Gu et al., 2003) of AMD patients, and elevated in the retinas of rodents after intense light exposure (Collier et al., 2011;Organisciak et al., 2012). Mice immunized with CEP adducts accumulated complement component-3 in Bruch's membrane, drusen deposits underneath the RPE, and RPE degeneration, features of dry AMD (Hollyfield et al., 2008). CEP adducts also stimulated neovascularization in vivo through a VEGF-independent pathway (Ebrahem et al., 2006).
Moreover, D-PUFAs can cross-protect various PUFAs; a membraneincorporated D-PUFA protects other PUFAs in this membrane by terminating the LPO chain reaction (Shchepinov, 2020). For example, the presence of the n-6 PUFA D2-linoleic acid in lipid bilayers down-regulated not just 4-HNE but also 4-HHE formation (Raefsky et al., 2018).
Our previous study showed a threshold protective effect of var- we found that IVT iron-induced damage appears to be less severe with the DHA diet than our previous study in mice fed with "regular chow" (LabDiet 5001) (Liu et al., 2021). We previously observed ironinduced retinal AF throughout the retina in mice fed with LabDiet 5001 at 1 week after iron injection, and "kidney bean" shaped geographic atrophy always occurred at 4 weeks after iron injection (Liu et al., 2021). In the current study, we observed that iron-induced AF was more limited to the superior retina in mice fed with DHA for 4 weeks at 1 week after iron injection ( Figure S2), and the "kidney F I G U R E 6 D-DHA prevented iron-induced acute RPE atrophy and progressive geographic atrophy development. Mice were fed with 4 weeks of D-DHA or DHA before receiving intravitreal injection of iron in one eye and control normal saline in the other. cSLO and OCT images were acquired at 4 weeks after iron or saline injection. Representative BAF cSLO images (a), IRAF cSLO images (b), and OCT scans (c) are shown. Toluidine blue staining was conducted on plastic sections prepared at 4 weeks after injections (d and e). Green lines indicate the positions of horizontal OCT scans. Red arrow indicates region of hyper-AF and hypo-AF lesions in IRAF cSLO image, corresponding to representative atrophic RPEs in OCT scans. Yellow arrows indicate ONL thinning in OCT scans. Red arrows indicate atrophic RPE in OCT scan. Blue arrows indicate pigmented cells between the ONL and RPE layers. These may be myeloid or migrated RPE cells. Representative images are shown from N=4 mice/group. Scale bar: 50 µm bean" shaped geographic atrophy only occurred in some, but not all mice ( Figure S3), which was correlated with the amount of superior region AF one week after iron injection. Additional investigation will be required to determine the basis of the more severe iron-induced retinal degeneration in mice on LabDiet 5001, which differs significantly from the DHA control diet used in this study. Since the only difference between the DHA diet and D-DHA diet used herein is whether DHA is deuterated, our results suggest that D-DHA leads to long-term protection against photoreceptor and RPE degeneration induced by iron overload.

| Ocular D-DHA accretion and elimination
Eleven-week-old C57BL/6J mice were purchased from Jackson

| Iron-induced acute RPE atrophy
Adult male wild-type C57BL/6J mice (Stock No.000664, Jackson Labs) were housed in standard conditions under cyclic light (12 hr:12 hr light-dark cycle). Mice had ad libitum access to water and food. Beginning at 2mo of age, mice were placed on the AIN93G diet described above, supplemented with 0.25% D-DHA or DHA (control diet) for 1, 2, 3, or 4 weeks prior to the intravitreal injection. The complete composition of the diets is shown in Table S1.
Mice were given an intravitreal injection of 1 μl 0.5 mM ferric ammonium citrate diluted in 0.9% NaCl (saline) (MP Biomedicals LLC) or 1 μl of saline as control. Intravitreal injections were performed as previously described (Hadziahmetovic et al., 2011). Mice were continued on respective diet until their final evaluation. All housing and procedures were performed according to the NIH Guide for the Care and Use of Experimental Animals and approved by the University of Pennsylvania Animal Care and Use Committee.

| Mass-spectrometry
Three types of mass spectrometric analysis were performed to confirm repeatability of results. In the first, lipids were extracted from retinas by a modified Folch method (CHCl 3 /CH 3 OH, 2:1), derivatized to fatty acid methyl esters (FAME), and analyzed by high-resolution capillary gas chromatography and specialized chemical ionization tandem mass spectrometry as discussed previously (Wang et al., 2020). Baseline resolved DHA and D-DHA total ion signals were integrated, and the proportions of D-DHA/total DHA were calculated.
The second type of mass spectrometric analysis was performed on samples of the control and experimental diets, as well as microdissected neural retina and RPE from animals on the experimental diet. Lipids were extracted and saponified as described previously (Axelsen & Murphy, 2010) and analyzed by ESI-LC/MS on a 4000 QTrap (Sciex) operating in enhanced negative mode over an m/z range of 320-345 and a scan rate of 250 /sec. This analysis verified that the control diet contained DHA but no detectable D-DHA, while the experimental diet contained an array of D-DHA isotopologues but only trace amounts of natural DHA ( Figure S1).

| In vivo imaging system
Mice were given general anesthesia and placed on a platform.

| Electroretinography
Mice were dark adapted overnight and anesthetized with the same procedure. The electroretinograms were recorded with an Espion E3 system (Diagnosys LLC) with a ganzfeld Color Dome stimulator as previously described (Liu et al., 2021). All electroretinographies were performed at the same time of day.

| Tissue preparation and immunofluorescence
Immunofluorescence on cryosections was conducted as described previously (Hadziahmetovic et al., 2011). Primary antibodies used: were acquired with an epifluorescence microscope (Nikon 80i microscope, Nikon), and analyzed using NIS-Elements (Nikon).

| Plastic sections and Toluidine Blue Staining
Plastic sections (3 μm) were cut in the sagittal plane. The third eyelid was used for the orientation when embedding eyecups.
Toluidine blue staining on plastic sections was used to evaluate retinal morphology as previously described (Hadziahmetovic et al., 2011).

| RNA extraction and quantitative RT-PCR
Neural retina tissues were isolated as previously described (Hadziahmetovic et al., 2011). Gene expression changes in the neu-

| Statistical analysis
Statistical analyses were performed using GraphPad Prism 6.0 (San Diego, CA). One-way analysis of variance (ANOVA) was performed, and post hoc analysis was employed using Tukey-Kramer testing when differences were observed in ANOVA testing (p < 0.05).
Mean ± SEM was calculated for each group. PHA, HGP, GJ, and JTB performed fatty acid analyses. JC contributed essential reagents. YL and JLD drafted the manuscript, and all authors provided critical review of the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.