Photoreceptor oxidative stress in hyperoxia-induced proliferative retinopathy accelerates rd8 degeneration

To investigate the impact of photoreceptor oxidative stress on photoreceptor degeneration in mice carrying the rd8 mutation (C57BL/6N). We compared the hyperoxia-induced proliferative retinopathy (HIPR) model in two mouse strains (C57BL/6J and C57BL/6N). Pups were exposed to 75% oxygen, starting at birth and continuing for 14 days (P14). Mice were euthanized at P14, or allowed to recover in room air for one day (P15), seven days (P21), or 14 days (P28). We quantified retinal thickness and the length of residual photoreceptors not affected by rosette formation. In addition we explored differences in retinal immunostaining for NADPH oxidase 4 (NOX4), Rac1, vascular endothelium, and activated Mϋller cells. We analyzed photoreceptor oxidative stress using DCF staining in cross sections and quantified NOX4 protein levels using western blotting. C57BL/6N mice in HIPR showed increased oxidative stress, NOX4, and Rac1 in the photoreceptors at P14 and P15 compared to C57BL/6J. In addition, we observed significant progression of photoreceptor degeneration, with significantly accelerated rosette formation in C57BL/6N under HIPR, compared to their room air counterparts. Furthermore, C57BL/6N under HIPR had significantly thinner central retinas than C57BL/6J in HIPR. We did not find a difference in vascular disruption or Mϋller cell activation comparing the two strains in hyperoxia. In HIPR, the C57BL/6N strain carrying the rd8 mutation showed significantly accelerated photoreceptor degeneration, mediated via exacerbated photoreceptor oxidative stress, which we believe relates to Rac1-NOX dysregulation in the setting of Crb1 loss-of-function.


Introduction
The mutation known as retinal degeneration 8 (rd8) is caused by a spontaneous frameshift mutation resulting in a premature truncation in crumbs 1 (Crb1). On funduscopic examination, the rd8 mice show large white spots in the nasal quadrant, progressively spreading along with a slow process of photoreceptor degeneration [1,2]. Histologically, the rd8 mutation is characterized by retinal folding, photoreceptor rosettes, outer nuclear layer loss, and retinal PLOS ONE | https://doi.org/10.1371/journal.pone.0180384 July 3, 2017 1 / 22 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 degeneration [1]. The photoreceptors are 25% shorter in mice with rd8 at 28 days of age, progressing to disorganized outer segments, with shortened inner segments by 10 weeks of age [2]. At five months of age, the photoreceptors lose their distinct, stratified organization, with complete loss of the outer segments in 35-month-old mice [2]. The protein product of the Crb1 gene is localized to the outer limiting membrane (OLM), which is comprised of highly specialized junctions that maintain retinal integrity and polarity. One of these junctions is the subapical region (SAR), where the photoreceptor inner segments and Müller glial cells connect. Crb1 localizes at the SAR on the microvilli of Müller glial cells [3][4][5]. Loss of Crb1 function disrupts the photoreceptor-Müller cell adhesion, which in turn weakens the integrity of the OLM, ultimately contributing to photoreceptor loss and the appearance of the classic rosettes of involuted photoreceptors [2][3][4][5].
In Drosophila, crb has been shown to control epithelial polarity by counteracting phosphoinositol 3-kinase (PI3K) and Rac1, a Rho GTPase [6][7][8][9][10]. Rac1 is important in cell-cell interaction, cell polarity, and cell migration [11]. Crb represses PI3K and Rac1 to stabilize the OLM integrity. It acts as a negative regulator of the Rac-NADPH oxidase (NOX)-dependent production of reactive oxygen species (ROS) in photoreceptors [9]. In a Drosophila model, loss of crb was shown to lead to dysregulation of the Rac1-NOX pathway, causing superoxide overproduction, and destruction of the photoreceptors [9]. Similarly, mice with constitutively active photoreceptor Rac1 have been shown to exhibit an increase in NOX-meditated oxidative stress that leads to photoreceptor degeneration [12]. Histologically, these mice had outer nuclear layer folds, with loss of photoreceptor orientation, a phenotype that simulates the rd8 mutation [12,13].
Given the importance of Crb1 in regulating Rac1 and limiting the production of photoreceptor superoxide, we sought to explore the effect of neonatal hyperoxia and neonatal proliferative retinopathy on mice carrying the rd8 mutation. We used the hyperoxia-induced proliferative (HIPR) model [14], where new-born pups are placed in high oxygen (75%) for 14 days. In the current study, we compared the effects of HIPR in two mouse strains: C57BL/6J (B6J) and C57BL/6N (B6N). These two strains were chosen since the C57BL/6N were homozygous for the rd8 mutation, while the C57BL/6J carried the wild-type Crb1 [15,16]. We hypothesized that neonatal oxidative stress would exacerbate the rd8 phenotype characterized by rosette formation in B6N mice via Rac1 dysregulation. We found that C57BL/6N develop an accelerated retinal degeneration, with loss of about 80% of photoreceptors within the first 4 weeks, compared to their room air raised littermates. These findings demonstrate that Crb1 is important in regulating photoreceptor oxidative stress and may have important translational implications to human disease, particularly infants carrying CRB1 mutations.

Hyperoxia-induced proliferative retinopathy model
This study was approved by the Institutional Animal Care and Use Committee at Northwestern University and all animal procedures were performed in compliance with the Institutional Animal Care and Use Committee at Northwestern University and the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Mice were sacrificed with isoflurane overdose and cervical dislocation.
Two strains of mice, C57BL/6J (000664, Jackson Laboratory, Bar Harbor, ME) and C57BL/ 6N (027, Charles River, Wilmington, MA), were used. The C57BL/6N mice carry the rd8 mutation while the C57BL/6J mice carry the wild-type Crb1 gene [15,16]. We have recently characterized the retinopathy in the HIPR model [14,17,18]. Briefly, newborn pups are placed in a Plexiglas chamber with an oxygen controller (Pro-Ox 110; Biospherix, Lacona, NY) and exposed to 75% oxygen from birth to P14. Dams were rotated from hyperoxia to room air every 24 hours to avoid oxygen toxicity [18]. Mice were either sacrificed after two weeks in hyperoxia (P14) or removed to room air for one day (P15), one week (P21), or two weeks (P28) and then sacrificed ( Fig 1A). Control, age-matched pups were raised in room air. Both male and female mice were used in control and HIPR groups in equal numbers. The HIPR model was reproduced in 42 litters (with corresponding room air controls) over a period of three years. The data presented is collated from experiments with mice in different litters.

Detection of rd8 mutation by PCR
DNA was isolated from tail biopsies and DNA samples were amplified separately for the wildtype and mutant rd8 alleles as previously described, using the mCrb1-mF1, mCrb1-mF2, and mCrb1-mR primers [2]. The PCR reaction was carried out in a 25 μL reaction volume containing 20ng In the hyperoxia-induced proliferative retinopathy model, mice were exposed to hyperoxia for 14 days and either sacrificed or returned to room air for one day (P15), seven days (P21), or 14 days (P28). Room air (RA) control pups were sacrificed at the same time points. Retinal cross sections were stained with hematoxylin and eosin (B, C). Rosettes of photoreceptors were seen in C57BL/6N HIPR mice at P21 and more prominently at P28 (blue arrows). Scale bar, 50 μm (N = 3-5). ONL, outer nuclear layer. INL, inner nuclear layer. IPL, inner plexiform layer.
DNA, 1.25 μM forward and reverse primer, and 2X Master Mix (Thermo Fisher Scientific, Waltham, MA). PCR reactions were performed using the following conditions: denaturation 94˚C for  5 minutes, amplification for 40 cycles at 94˚C 30 seconds, 65˚C for 30 seconds, 72˚C for 30 seconds, and final extension at 72˚C for 7 minutes. The wildtype allele produces a 220 bp fragment and the mutant rd8 allele produces a 244 bp fragment.

Retinal thickness and residual photoreceptor analysis
Enucleated eyes were fixed in 10% neutral buffered formalin for 24 hours or fixed in 4% paraformaldehyde for four hours. Paraffin embedded eyes were sectioned (7 μm), deparaffinized, and hematoxylin and eosin stained. Images were taken with Nikon 80i Eclipse microscope (Nikon, Tokyo, Japan) using a Photometrics CoolSnap CF camera (Photometrics, Tucson, AZ). Retinal thickness analysis was performed using ImageJ software (NIH, Bethesda, MD), as previously reported [19]. Thickness was measured by two masked observers at multiple locations: 100 μm on either side of the optic nerve and in the retinal periphery. The measurements were calculated using distance_between_polylines.java ImageJ plug-in [20]. This tool allows multiple thickness measurements to be automatically calculated and averaged. We calculated the retinal thickness as the average of three to four eyes per time point.
The percentage of residual healthy photoreceptors was determined by dividing the length of visible photoreceptors by the total retinal pigment epithelium length, per retinal cross-section. Two independent masked observers analyzed four to five separate cross-sections (40 μm apart) per eye. The measurements obtained by the two observers were compared to calculate the interclass correlation coefficient (ICC).

Retinal flat mount analysis
Retinas were processed and stained as previously reported [21]. Eyes were fixed in 10% neutral buffered formalin for 24 hours, retinas were dissected and the vitreous removed. Retinal cups were permeabilized for 18 hours with phosphate buffered saline (PBS) supplemented with 0.1% Triton X-100. After blocking in 10% donkey serum, 1% bovine serum albumin (BSA), 0.1% Triton X-100 diluted in PBS for five hours, retinas were immunostained with GS-Isolectin Alexa Fluor 594 (IB4, 1:75 dilution, I21413, Thermo Fisher Scientific) for 18 hours. After several washes, retinas were flattened and mounted with ProLong Gold Antifade reagent (Thermo Fisher Scientific). Flat mounts were masked and quantified by two independent observers. Measurements were made using ImageJ (NIH) to quantify the avascular retinal area (in pixels) as a percentage of the total retinal area.
The sections were then stained for an adhesion receptor restricted to the Müller glial cell end-feet, CD44 [2,23]. Immediately after imaging for DCF, the coverslips were carefully removed and the sections were fixed for 10 minutes in -20˚C acetone. Following air drying and several washes, endogenous hydrogen peroxide was blocked with 3% hydrogen peroxide. After blocking in 10% donkey serum, sections were incubated in rat anti-CD44 (1:100 dilution, 553131, BD Biosciences, San Jose, CA) and then incubated with biotin conjugated donkey anti-rat (1:200 dilution, 112-065-167, Jackson ImmunoResearch). The sections were incubated with streptavidin Alexa Fluor 594 conjugate (1:200 dilution, S-11227, Thermo Fisher Scientific) for one hour and counterstained with To-Pro3 (1:300 dilution, T3605, Thermo Fisher Scientific). Sections were imaged with the Zeiss LSM-510 Meta confocal microscope.

Quantifying fluorescence intensity
We determined fluorescence intensity using three different cross-sections per eye (N = 3 eyes per group) using ImageJ (NIH) [24,25]. The fluorescence was calculated using corrected total cell fluorescence (CTCF) using the following formula: CTCF = Integrated density-(Area of selection X Mean fluorescence of background areas). Reported data are mean ± standard error of mean (SEM).

Statistical analysis
All data were analyzed using SPSS (v.23.0; IBM Corp, Armonk, NY). Results are expressed as mean ± SEM. Mice from different litters (4-5 per group) were analyzed. Group differences were evaluated using ANOVA with Bonferroni's post hoc comparisons. Results were considered statistically significant for p-value < 0.05.

Results
C57BL/6N in hyperoxia display significant progression of rosette formation compared to their room air counterparts, and significantly thinner retina than C57BL/6J mice in hyperoxia To identify morphological differences between C57BL/6J (B6J) and C57BL/6N (B6N) strains, histologic cross-sections were analyzed. In room air controls of both strains, we did not see any evidence of rosette formation at any time point (N = 3-4 per time point). The B6J in HIPR did not display the rosette formation and had continuous intact photoreceptors throughout the retina (Fig 1B), while the B6N HIPR mice revealed rosette structures involving the photoreceptors at P21 and P28 (Fig 1C, blue arrows and S1 Fig). In the B6N strain, the outer nuclear layer and inner nuclear layer were indistinct, precluding any measurements of the outer nuclear layer thickness or nuclei. Thus to quantify the loss of photoreceptors, the length of residual photoreceptors was normalized to the length of the retinal pigment epithelium (RPE) (Fig 2A, left). The ratio of intact photoreceptor (inner and outer segments) to RPE length decreased significantly in B6N HIPR retinas comparing P15 to P21 (99.2 ± 0.8%, ICC 0.98 vs 57.5 ± 5.9%, ICC 0.95, p<0.05) (Fig 2A), decreasing further from P21 to P28 (16.4 ± 1.3%, ICC 0.97, p<0.05) (Fig 2A). In the setting of HIPR, the retinal sublayers became largely disorganized and indistinguishable in rd8 eyes (B6N) at P28 (Fig 1C).
The total retinal thickness was measured near the optic nerve and in the peripheral retina. In B6N mice, retinas were significantly thinner in HIPR than room air controls at P15, P21, and P28 near the optic nerve ( Fig 2B) and only at P21 and P28 in the periphery (p<0.05, Fig  2C). In contrast, 6J HIPR retinas were significantly thinner than room air controls centrally at P14 and P15 (p<0.05, Fig 2B) and in the periphery at P21 (p<0.05, Fig 2C).

Disorganized angiogenesis occurs in both C57BL/6J and C57BL/6N in hyperoxia-induced proliferative retinopathy
We previously reported significant disruption of the retinal vascular development, with persistent hyaloidal vessels in B6J HIPR [14]. Comparing the B6J and B6N strains in HIPR, retinal flat mounts stained with IB4, an endothelial cell marker, were assessed at P21 and P28. In both strains, the retinal flat mounts revealed absent patterned retinal vasculature at P21 and P28. Instead, completely disorganized vasculature were visible surrounding the optic nerve and in the retinal mid-periphery in both strains (Fig 3). Persistent hyaloid vessels were only visible in some flat mounts (Fig 3, red arrowheads) since the hyaloidal vessels are sometimes  inadvertently removed during retinal dissections. There was no qualitative or quantitative difference in these vessels comparing B6J and B6N mice (Fig 3).

C57BL/6J and C57BL/6N hyperoxia mice manifest abnormal localization of vasculature along with Müller cell activation
Based on the presence of extensive angiogenesis with failure of normal retinal vascular development (Fig 3), we wanted to evaluate the exact localization of these vessels in both strains. Retinal cross sections were stained for endothelial cells (IB4) and activated Müller cells (GFAP). In both strains of HIPR, there was complete absence of the retinal vasculature at P14 and P15 (Fig 4). By P21 and continuing at P28 both strains showed endothelial cell staining, appearing as vascular tufts in the inner plexiform layer (Fig 4, arrows). Activated Müller cells were seen in B6J and B6N HIPR mice at P21 and P28 (Fig 4). Neither strains showed any evidence of normally stratified retinal compared to room air controls. C57BL/6N in hyperoxia exhibit higher levels of photoreceptor reactive oxygen species compared to C57BL/6J in hyperoxia We characterized oxidative stress in the retina using dichlorofluorescein (DCF) staining, which fluoresces upon contact with reactive oxygen species (ROS). In a diabetic mouse model study, DCF staining was increased in the photoreceptor inner and outer segments compared to controls [28]. Similarly, we found increased DCF staining in the photoreceptors in both strains in fresh frozen cryosections of HIPR retinas (Fig 5). Since we stained DCF on fresh frozen, unfixed tissue, we counterstained the same tissue, after fixation, with CD44, a marker for Müller glial end-feet for orientation and to explore whether DCF staining colocalized with OLM disruption [2,23]. We found focal disruption of the OLM in B6N HIPR at P15 and P21 (Fig 5B, arrows), but the DCF staining was more diffuse throughout the entire photoreceptor layer (Fig 5B).
In hyperoxia-induced proliferative retinopathy, C57BL/6N have significantly increased NOX4 protein levels at P15 compared to C57BL/ 6J NOX are a major source of ROS in the retina [29,30]. The NOX family of proteins consists of seven isoforms (NOX 1-5, DUOX1 and DUOX2). Since NOX3 is found in the inner ear [31] and NOX5 is not present in rodents [32], we did not study them. NOX1 is located in endothelial cells and vascular smooth muscle cells [33] and is a significant source of ROS in the lung during neonatal lung hyperoxia exposure [34,35]. When we examined NOX1 protein levels in the retina, however, we did not find any statistical differences between the strains (S2A Fig). NOX2 and, more prominently, NOX4 are present in photoreceptors [30]. We have recently shown that NOX2 was elevated in B6J HIPR compared to controls [14]. In the current study, Since the photoreceptors were a major source of ROS in HIPR (Fig 5), we decided to determine NOX4 protein levels and their source in the retina. At all time points, we found significantly higher NOX4 levels in B6N HIPR compared to B6J HIPR (p<0.05, Fig 6A and 6B). Notably at P15, NOX4 was 19.4 ± 6.0 fold change in B6N HIPR compared to 1.1 ± 0.3 fold change in B6J HIPR (Fig 6B). The NOX4 levels in B6N HIPR decreased by P21; though they remained significantly higher than B6J HIPR (p<0.05, Fig 6B). There was stable level of NOX4 protein across the time points in B6J HIPR (Fig 6B), with a stable, low level of NOX4 in photoreceptors similar to room air controls (Fig 6C). NOX4 mainly localized in the photoreceptors in B6N HIPR ( Fig 6C) and was significantly higher than B6N room air controls. In the inner retina, NOX4 localized to the blood vessels and retinal ganglion cell layer in room air controls of both strains (S3 Fig). In HIPR, NOX4 localized to the retinal ganglion cell layer in B6J, compared to more widespread staining in the inner nuclear, inner plexiform, and ganglion cell layers in B6N (S3 Fig).

Rac1 is significantly increased at P15 in C57BL/6N in hyperoxia-induced proliferative retinopathy
The Rho family of GTPases includes Rac1, RhoA, and Cdc42 [36]. We sought to determine the localization of expression of all three GTPases in the retina and to examine strain differences during HIPR. We found that RhoA and Cdc42 labeled the photoreceptors, but there were no clear differences in labeling between the strains at any time point during HIPR (S5 Fig). We therefore sought to focus on Rac1, GTP-bound Rac1 was shown be increased in photoreceptors of Drosophila in the setting of crb knock-down [9], which we hypothesized would impact the photoreceptors in B6N as a result of the rd8 mutation. At all time points, Rac1 was present in the photoreceptors (Fig 7A) and the inner nuclear layer (S4 Fig) in control and HIPR mice. In B6J HIPR, Rac1 staining was not significantly different compared to B6J controls (Fig 7). In contrast, B6N HIPR showed increased Rac1 at P15 in the photoreceptors and the inner nuclear layer compared to B6N room air controls (14,740 ± 1,147 vs 7,113 ± 473 AU, p<0.05) (Fig 7B). By P21 in B6N HIPR, Rac1 levels significantly decreased compared to controls (3,869 ± 620 AU vs 8,349 ± 373 AU, p<0.05, Fig 7B). This decrease in Rac1 most likely correlates with the statistically significant loss in photoreceptors at P21 in B6N due to accelerated rd8 phenotype (Fig 1).

Discussion
Our data provide evidence that neonatal oxidative stress exacerbates rosette formation in B6N mice, which we hypothesize may be related to Rac1 dysregulation. The B6N mice in HIPR showed dramatically accelerated rd8 phenotype with severe retinal disorganization as early as P28. This finding stands in stark contrast to the slow progression of the crb1 mutation in room air raised mice, where the photoreceptor rosettes are not prevalent or functionally significant until around 35 months of age [2]. As early as P15, B6N mice undergoing hyperoxia exposure in HIPR experienced significantly increased oxidative stress with increased ROS (Fig 5), NOX4 (Fig 6), and Rac1 (Fig 7) compared to B6J strain under similar hyperoxia conditions. We hypothesize that the increase in oxidative stress in the B6N strain is due to loss of Crb1 function leading to Rac1 dysregulation, which in turn leads to increased ROS formation and higher NOX4 expression (Fig 8).
Photoreceptors are largely susceptible to oxidative stress because of their high metabolic demand [37]. In Drosophila, crb knock-down showed increased ROS in the photoreceptors compared to controls, which required the presence of Rac1 and NADPH oxidase [9]. When loss-of-function crb mutant Drosophila were treated with apocynin, they experienced improved photoreceptor cell survival [9]. In the setting of diabetic retinopathy, another model of retinal vascular ischemia, photoreceptors have been shown to be an early source of oxidative stress [28]. Taken together, we believe the addition of hyperoxia exposure on the background of Crb1 accelerated the photoreceptor degeneration (Figs 1 and 2A) and increased NOX in the photoreceptors (Fig 6). While increased DCF staining was seen in both strains (Fig 5), NOX4 was significantly increased in the B6N strain (Fig 6), suggesting its potential role in this setting. With the rd8 mutation, Crb1 no longer regulates (inhibits) Rac1. Uninhibited Rac1 then activates NOX2 to produce superoxide. Superoxide is reactive oxygen species (ROS). It is unknown what mechanism regulates NOX4. We hypothesize (dotted line) that in photoreceptors, NOX4 is activated by a Rac1-dependent signaling mechanism. In turn, with loss of Crb1 regulation of Rac1, an increase in NOX4 produces more ROS, leading to photoreceptor damage via oxidative stress. https://doi.org/10.1371/journal.pone.0180384.g008 To understand the mechanism of accelerated rosette formation in HIPR, we explored the expression of downstream molecular targets of Crb1, focusing on Rac1. Rac1 has been shown to play an important role in photoreceptor morphogenesis and survival under stress. In rhodopsin-null Drosophila, the addition of activated Rac1 rescued photoreceptor morphogenesis [38]. In mice, Rac1 was involved in light-induced photoreceptor apoptosis [39,40]. Conditional photoreceptor Rac1 knockout protected the retina from light-induced degeneration [40]. In turn, constitutively active Rac1 in rods lead to severe compromise in photoreceptor polarity and migration, with ultimate photoreceptor apoptosis due to NOX-induced ROS [12,13]. In HIPR, we show that Rac1 was significantly increased at P15 in the B6N strain, localizing to the photoreceptors by immunohistochemistry, which suggests a role for Rac1-mediated acceleration of the rd8-related photoreceptor degeneration in the HIPR model. This finding stood in contrast to the B6J HIPR, where Rac1 levels in the photoreceptors were similar to room air controls at all time points, consistent with functional wild-type Crb1 regulating Rac1 (Fig 7).
Since Rac1 is a member of a highly conserved family of Rho GTPases, which also include RhoA and Cdc42 [36], we wanted to explore whether the other GTPases were affected in the HIPR phenotype. However, we did not find a clear difference in photoreceptor labeling of RhoA or Cdc42 in either strain under HIPR at all time points (S5 Fig). Crb1 forms a complex that helps maintain apical polarity at the subapical adherens junction, which includes Cdc42 [41]. In a rod-specific knockdown of Cdc42 in mice with rd1 and rd10, Cdc42 accumulated in the perinuclear region of the photoreceptors but was not found to play a major role in photoreceptor death [42]. However, neuronal knockdown of Cdc42 during development caused loss of the outer limiting membrane in the peripheral retina and progressive retinal degeneration [43]. In a porcine retinal detachment model, RhoA activity increased within hours after the detachment [44]. Little is known about the mechanism of RhoA activation in the injured retina. Based on our immunostaining, we found that RhoA and Cdc42 were increased in the Müller cells of both strains at P21 under HIPR (S5 Fig). This would suggest that RhoA and Cdc42 increase in Müller cells, whereas Rac1 is more increased in the photoreceptors during HIPR. However, additional studies are needed to further elucidate the specific role of the different GTPases during retinal oxidative stress in other models of retinopathy, as well as in HIPR.
We explored the relative role of the different NOX and found that, similar to other retinal ischemia models [12,45], in the B6N strain but not in B6J, NOX4 contributed to the ROS generation in photoreceptors under HIPR (Fig 6). While NOX2 requires Rac1 for activation, we found increased Rac1 expression only in B6N HIPR mice at P15 (Fig 7). Since we use whole retinal lysates for western blotting, we suspect that the increased NOX2 that we see in both strains may originate from cell types other than the photoreceptors. In a light damage mouse model, NOX2 and NOX4 were responsible for photoreceptor ROS production [45]. In a live retinal explant system, NOX4 was the most abundant NOX in photoreceptors and, with the addition of apocynin, the levels of ROS in photoreceptors were reduced [30]. While the other NOXs (NOX1, 2, 3) require Rac1, NOX4 is thought to not require Rac1 in most cells [46]. However, in mesangial cells, Rac1 was found to stimulate NOX4 ROS production [47]. The mechanism of NOX4 regulation has not been completely resolved in many cell types, including the photoreceptors [48]. Based on our findings, we suspect that NOX4 activation in the photoreceptors is Rac1-dependent, similar to mesangial cells [47] (Fig 8). To determine the exact mechanism of NOX4 activation, further molecular studies are needed that explore in photoreceptor cell culture whether NOX4 activation is dependent upon Rac1 signaling during oxidative stress.
To study other potential modifiers of the phenotype comparing the two strains, we explored inflammatory markers and Müller cell activation. In a study comparing the B6N and B6J strains at 3 months of age, researchers found an increase in inflammatory markers in B6N mice, with no significant difference in gene expression between the two strains [49]. In contrast to these authors, we did not observe a difference in Müller glial cell activation (GFAP staining; Fig 4) It is important to emphasize that there are other genetic differences between the B6N and B6J mice, in addition to the rd8 mutation, that could contribute to the differences in inflammation and response to oxidative stress. Of note, the Nnt gene encodes the mitochondrial surface nicotinamide nucleotide transhydrogenase protein that is responsible for reducing NADP+ and NAD+ to NADPH and NADH. [49] The B6J strain has a deletion in the Nnt gene, causing the protein to be absent [50,51], leading to higher levels of hydrogen peroxide and spontaneous oxidation of NADPH [52], This would lead to potentially enhanced oxidative stress in B6J, which we did not find [51].
Rats carrying Crb1 mutation develop spontaneous intra-retinal angiogenesis in the inner nuclear layer [53]. Based on these findings we expected strain differences in the location and extent of angiogenesis under HIPR (Figs 3 and 4). However, comparing the vascular phenotype in HIPR, we found that both strains showed similar avascular areas and intra-retinal localization of vascular staining (Figs 3 and 4). Similarly, other researchers found no difference in the laser-induced choroidal neovascularization lesion size comparing B6N and B6J [49]. The lack of difference in angiogenesis and Müller cell activation between the two mouse strains in HIPR is intriguing. These two strains differ by only 12 known single nucleotide polymorphisms, making their backgrounds very similar [50,51].
In humans, mutations in the CRB1 gene can cause significant visual impairment, leading to a severe phenotype of retinitis pigmentosa (RP) and Leber congenital amaurosis (LCA) [54][55][56][57][58][59]. CRB1 mutations are detected in 6% of patients with RP [60, 61] and 10-13% of patients with LCA [55, 58,62]. Patients with CRB1 mutations manifest either severe progressive pericentral, visual field loss in adult onset RP or profound loss of vision from birth in LCA [63]. In contrast to these severe findings in human disease, visual function and electroretinography in mice with rd8 are not significantly different from wildtype mice, even at the most advanced stages of retinopathy (> 9 months) [63]. It is not clear why the human CRB1 phenotypes are significantly more severe than in mice. In contrast to rd8 (Crb1 mutants), mice deficient in Crb2 showed more severe, early disorganization throughout the entire retina [64]. Crb1 and Crb2 double knockout mice show severely impaired retinal function and thickening of the retina due to increased cell proliferation [65]. However, humans with CRB2 mutations have rarely been reported with a retinal phenotype [66]. Recently one patient with a CRB2 mutation has been reported, presenting with reduced visual acuity, nystagmus, and irregular retinal pigmentation [67]. This data suggests that in humans, CRB1 mutations alone can cause severe photoreceptor degeneration, while defects in multiple components of the crumbs complex may be required to generate the severe phenotype in mice. In addition, we show that enhanced photoreceptor oxidative stress, in HIPR, leads to significant acceleration of the rd8 phenotype, which may present an opportunity to explore the severe phenotype of LCA seen in human CRB1 mutations.
In summary, our results show that the HIPR model in the B6N mouse strain exposes the photoreceptors to significant oxidative damage, with acceleration of rosette formation and significant degeneration of the photoreceptors by P28, unlike the otherwise mild natural history of this mutation. We found significant differences in ROS, NOX4, and Rac1 between the B6J and B6N substrains in HIPR. The two strains did not show significant differences in angiogenesis or Müller glial cell activation. These findings highlight the importance of understanding strain differences, and alerts researchers to rule out the rd8 (and other retinal degeneration) mutations [1,2,16] as potential confounders in retinal experiments, especially when ischemia and oxidative stress are introduced at a young age. Future studies of HIPR with extended time points as well as functional studies will be important to further explore differences between the strains in this model of severe retinal ischemia and angiogenesis. This model may provide researchers with an opportunity to explore the severe phenotype of Crb1 mutations in the mouse, which may simulate the severe phenotype of human CRB1. The role of oxidative stress in accelerating rd8 phenotype suggests that higher vigilance may be indicated to avoid photoreceptor oxidative stress in infants and children carrying these mutations [68]. Based on our findings, it will be important to explore whether approaches that improve Rac1 regulation could be useful to improve photoreceptor survival in rd8 and, ultimately human CRB1-related retinal degenerations.