Hyperactivation of retina by light in mice leads to photoreceptor cell death mediated by VEGF and retinal pigment epithelium permeability

Abstract

Light toxicity is suspected to enhance certain retinal degenerative processes such as age-related macular degeneration. Death of photoreceptors can be induced by their exposure to the visible light, and although cellular processes within photoreceptors have been characterized extensively, the role of the retinal pigment epithelium (RPE) in this model is less well understood. We demonstrate that exposition to intense light causes the immediate breakdown of the outer blood-retinal barrier (BRB). In a molecular level, we observed the slackening of adherens junctions tying up the RPE and massive leakage of albumin into the neural retina. Retinal pigment epithelial cells normally secrete vascular endothelial growth factor (VEGF) at their basolateral side; light damage in contrast leads to VEGF increase on the apical side--that is, in the neuroretina. Blocking VEGF, by means of lentiviral gene transfer to express an anti-VEGF antibody in RPE cells, inhibits outer BRB breakdown and retinal degeneration, as illustrated by functional, behavioral and morphometric analysis. Our data show that exposure to high levels of visible light induces hyperpermeability of the RPE, likely involving VEGF signaling. The resulting retinal edema contributes to irreversible damage to photoreceptors. These data suggest that anti-VEGF compounds are of therapeutic interest when the outer BRB is altered by retinal stresses.

Figure 1

Light-induced photoreceptor degeneration by apoptosis. Balb/c mice were exposed to 5000 lux of white light during 1 h. Compared with retina unexposed to light (a), cresyl violet staining of mice, respectively, analyzed 24, 36 and 48 h after LD (b–d) showed that LD-induced photoreceptor cell death progressed with time. (e) Cytoplasmic-free nucleosomes in the retina were quantified using an ELISA-based cell death detection kit at the different time points indicated. One-way ANOVA (Dunnett's post hoc test) analysis revealed that the number of free nucleosomes released after LD were significantly higher 24, 36 and 48 h after the lesion compared with animals unexposed to LD (P<0.001 for all time points studied). Each bar represents the mean value±S.D. (n=6 per group). ONL: outer nuclear layer, INL: inner nuclear layer, scale bar: 50 μm

Figure 2

VEGF-triggered RPE permeability is an early event in the light-damage model. Confocal analysis of tight- and adherens-junction markers performed on flat-mounted RPE revealed that RPE permeability is an early event in the LD model. Whereas the beta-catenin expression delineates the contour of RPE cells revealing their hexagonal shape on control unexposed to light (a), the expression of this marker was localized in the cytoplasm of RPE cells 24 h after LD (b). Similar results were observed concerning the expression of ZO-1 and N-cadherin markers. Normally located in the contour of RPE cells on control unexposed to light (c–e), both of these markers translocate in the RPE cell cytoplasm 24 h after LD (d–f) (n=6 per group – magnification × 400). Immunostaining of serum albumin, normally found in the choriocapillaris and in the retinal vessels located at the inner side of the retina of control unexposed to light (g and h), showed pronounced extravascular albumin leakage 24 (i and j), 36 (k and l) and 48 h (m and n) after light exposure. After LD, albumin was prominent in the RPE and in the outer nuclear layer (ONL). Albumin positivity was also found, to a lesser extent, in the outer plexiform layer (OPL), in the inner nuclear layer (INL), in the plexiform layer (IPL), in the ganglion cell layer (GCL) and, finally, in the inner limiting membrane (ILM), suggesting that plasma leakage originates from the outer BRB (n=6 per group—magnification × 400, scale bar: 100 μm in (g–n)—Green: albumin and blue: DAPI). (o) ELISA quantification shows that VEGF level was 1.8-fold and 1.5-fold increased, respectively, 24 and 36 h after LD compared with control unexposed to light in the neural retina (control no LD: 231.6±56.3 pg/mg protein; 24-h post LD: 423.5±79.3 pg/mg protein; 36-hour after LD: 353.8±87.7 pg/mg protein and 48-h post LD: 209.7±45.3 pg/mg protein). Contrariwise, in the eye-cup, VEGF content was significantly decreased 24, 36 and 48 h after LD compared with unexposed controls (control no LD: 508.9±96.2 pg/mg protein; 24-h post LD: 262.9±37.3 pg/mg protein; 36-h post LD: 340.7±88.8 pg/mg protein and 48-h post LD: 262.7±132.4 pg/mg protein). No obvious difference was observed concerning VEGF content in the whole eye, except at 48 h after LD. At this point, VEGF was 1.5-fold decreased compared with the control. Each bar represents the mean value±S.D. (n=6 per group). Statistical analyses: one-way ANOVA and Dunnett's post hoc test: *P<0.05. (p) Western blot analysis for PEDF in the eye-cup showed that PEDF expression was significantly reduced 24, 36 and 48 h after LD. Quantification is presented as relative expression normalized against values of tubulin controls to adjust for protein loading. Results are expressed as mean±S.D. (n=6 per group). Statistical analyses: one-way ANOVA and Tukey's post hoc test: *P<0.05 and **P<0.01

Figure 3

RPE permeability is induced through a VEGFR-2/Src pathway. Western blot analysis showed a three-fold increase in phosphor-Tyr416-Src (P-Src) expression 6 h after LD in the eye-cup (a), as well as in the neural retina (b). The expression of VEGFR-1 decreased significantly and progressively down to 28% of basal level at 48 h after LD (c). At 24 and 36 h after LD, VEGFR-2 expression was decreased by 45 and 35%, respectively, and seemed to return to basal level at 48 h (19% decrease, not significant – d). Quantification results are expressed as percentage of the control group. Protein expression level was normalized against values of tubulin, GAPDH or actin controls to adjust for protein loading. Results are expressed as mean±S.D. (n=6 per group). Statistical analyses: one-way ANOVA followed by Bonferoni's post hoc test: *P<0.05 and **P<0.01 (no LD: control group unexposed to light)

Figure 4

Blocking VEGF in the LD model inhibits RPE permeability and extravascular albumin leakage. The left panel shows the expression of beta-catenin stained on flat-mounted RPE and analyzed on control unexposed to light (a) or 24 h after LD in the following groups: no treatment prior to LD (b), Sham- (c), LV-GFP- (d) and LV-V65-treated animals (e and f). In unexposed control, the interface of cells is clearly delineated by beta-catenin staining (white arrowheads in a). In the control groups (no treatment prior to LD, sham- and LV-GFP-injected), beta-catenin staining shows a clear translocation of the adherens-junction component toward the cytoplasm of RPE cells and is absent at the cell contacts. Contrariwise, in the LV-V65-treated group, beta-catenin staining is maintained at the interface of cells (white arrowheads in e and f), although slightly more diffuse than in absence of exposure to light. Note that GFP expression of LV-GFP (g) and LV-V65 (h) vectors are observed at the RPE level (n=5 per group – magnification × 400 (confocal analysis) – red: beta-catenin and green: GFP). The right panel presents serum albumin staining of mice unexposed to light (i and j), untreated before LD (k and l), Sham- (m and n), LV-GFP- (o and p) and LV-V65 (q and r)-treated groups and analyzed 10 days after LD. Plasma leakage is observed in Sham- and LV-GFP-injected groups. Albumin positivity is evident in the choriocapillaris, RPE, ONL, OPL, INL, IPL and GCL. However, in the LV-V65-treated group, albumin staining is only limited to the outer segments of the photoreceptors, which appeared to be protected by the blockade of VEGF (n=8 per group – magnification × 400, scale bar: 100 μm – red: albumin and blue: Dapi, OS: outer segments, ONL: outer nuclear layer, INL: inner nuclear layer)

Figure 5

Blocking VEGF in the LD model is neuroprotective for photoreceptors. (a) Representative scotopic electroretinogram response for both the a- and b-wave amplitudes. The quantification of the a-wave (b) amplitude at the different stimuli intensities reveals that response of the LV-V65-treated group is intermediate between control unexposed to light and LV-GFP-treated mice responses. Note that LV-V65-treated mice response is significantly higher than LV-GFP-treated mice response for the following intensities: 3 × 10³ (P<0.05), 10 × 10³ (P<0.01) and 25 × 10³ (P<0.05) mcds/m2. Similarly, the quantification of the b-wave amplitude at the different intensities stimuli showed the same result (c). LV-V65-treated mice response was significantly higher compared with LV-GFP-treated mice response for the same intensities: 3 × 10³ (P<0.01), 10 × 10³ (P<0.05) and 25 × 10³ (P<0.05) mcds/m2. Results are expressed as mean±S.E. (n=8 per group); statistical analyses: two-way ANOVA and Bonferroni's post hoc test. This experiment was reproduced with different batches of lentiviral vectors and led to the same results. Transgene expression level quantified by quantitative PCR revealed that functional rescue of photoreceptor correlates with the level of V65 expression for both the a-wave (d) and the b-wave (e) maximal amplitudes (n=10 per group)

Figure 6

Blocking VEGF in the LD model preserved the rod and cone outer segment morphology. Peripherin–RDS expression, normally localized in rod and cone OS of control unexposed to light (a), is completely abolished 10 days after light exposure (b). Peripherin–RDS expression is also absent in the sham-treated retina (c). Although Peripherin–RDS staining is observed in both LV-GFP- (d) and LV-V65 (e)-treated retinas, the OS in the LV-GFP-treated group appeared shorter and disorganized compared with those of LV-V65-treated group. In the LV-V65-treated group, the OS remains well organized and has an appropriate length. Similar result was obtained for the cone OS marker SWL–opsin analyzed 10 days after LD. SWL–opsin staining, normally found in the cone OS of control unexposed to light (f and f′), is completely abolished after light exposure (g and g′). Similarly, the expression of SWL–opsin is absent in sham-treated retina (h and h′). SWL–opsin expression is maintained in LV-GFP- (i and i′) and LV-V65(j and j′)-treated groups; however, in addition to a weaker expression in the LV-GFP-treated group, the protein was partly mislocalized to the axon and cone pedicle (i′) (n=8 per group – magnification × 400, scale bar: 100 μm – red: Peripherin/SWL–opsin and blue: Dapi)

Figure 7

Blocking VEGF in the LD model preserved visual acuity. The optokinetic response (OR) was quantified in the following experimental groups: no treatment prior to LD (a), Sham- (b), LV-GFP- (c) and LV-V65 (d)-treated groups. The quantification of the OR was measured at the beginning of the experiment (prior the injection and LD), 2 weeks after subretinal injection and 10 days after LD. Statistical analysis presented in the figures compared OR before and after light exposures. Results show that the subretinal injection did not affect OR in the different injected groups (b–d), except in the LV-V65-treated group for 0.075 cpd. Contrariwise, 10 days after light exposure, ORs recorded in the control groups (no treatment prior to LD and Sham-injected) were significantly impaired and no difference was observed compared with background value. OR was still present but decreased in the LV-GFP-treated group after LD. Blocking VEGF preserved OR after light exposure in the LV-V65-treated group. Results are expressed as mean±S.E. Statistical analyses: one-way ANOVA and Tukey's post hoc test: n.s.: nonsignificant, *P<0.05, **P<0.01 and ***P<0.001 (n=8 per group)