Characteristics of Retinal Vascular Degeneration and the Expression of Vessel-Related Claudin Proteins in Retinal Degeneration Mouse

Retinal degeneration models are widely used to investigate the pathogenesis of retinitis pigmentosa (RP) in humans. This hereditary retinopathy is characterized by the progressive loss of retinal photoreceptor cells and visual acuity and the eventual blindness [1, 2]. As retinal photoreceptors undergo apoptosis, the metabolic demand of the photoreceptor cells decreases and the ocular microcirculation is altered in RP [3‒6]. The decreased blood flow of retinal and choroidal vessels in patients with RP has been found in some studies [4‒8]. However, the mechanisms of retinal vascular alteration and its association with photoreceptor degeneration need to be further clarified.

Claudins (CLDs) are 20–27 kDa proteins with over 24 isoforms in humans and mice [9, 10]. CLDs are crucial proteins primarily localized at the intercellular junctions of epithelial cells, endothelial cells (ECs), and tumor cells. CLDs can regulate the intercellular transport of water and ions and play a vital role in maintaining cell polarity, cell proliferation, barrier function, and cell signaling [11, 12]. Our previous studies have demonstrated that CLD-1, -2, and -5 specifically express on retinal vessels; and the overexpression of CLD-5 can decrease cell mobility and the sprouting capability of the vessel in vitro and induce new interaction patterns between CLD-5 and CLD-1 or -2 in retinal pigment epithelium cells and human retina ECs [13, 14]. CLD-1, -2, and -5 are implicated in the development and maintenance of the blood-retinal barrier (BRB), as well as the preservation of ECs’ structural integrity [13, 15‒17]. The expression of CLD-5 upregulated by JAM-A, a component of tight junctions, can reduce endothelial permeability [18]. Notably, CLD-3 is the sole known isoform of CLD that interacts with CLD-1, -2, and -5 through extracellular loops and then forms heterodimers [19]. It has been proved that elevated intraocular pressure leads to a significant decrease in the expression of CLD-3 and CLD-5, along with the up-regulation of interleukin 1-β and eNOS as well as the injury of BRB [20]. Our previous studies have also found that CLD-3 mainly locates at the retinal ganglion cells (RGCs) and the downregulation of claudin-3 can impede retinal vessel development in mice [13, 21], and CLD-3 is also required for normal development of the neural retina and retinal vessels in zebrafish [22]. Therefore, CLD-3 can also be considered a vessel-related CLD. These findings collectively unveil the importance of CLDs that can affect the development and stabilization of retinal vessels and retinal nerves.

The trait of rd1 (Pde6β^rd1/rd1) mice is caused by a recessive mutation of Pde6β, which encodes the rod β-subunit of cyclic guanosine monophosphate phosphodiesterase [23, 24]. The photoreceptor development is initially normal in rd1 mouse, but all rod photoreceptors start to degenerate at the postnatal 2 weeks; then the cone photoreceptors gradually degenerate [23]. As the rd1 mouse shares the pathophysiological features with RP disease, the rd1 mouse is a valuable tool for studying RP [25]. The study aimed to investigate the characteristics of retinal vascular degeneration and the expression of vessel-related CLDs and analyze the correlations between the degeneration of retinal vasculature and retinal neurons in rd1 mice.

All animal experiments adhered to the ARVO Statement for the use of animals in ophthalmic and vision research and were approved by the Institutional Review Board of Zhongshan Ophthalmic Center (Guangzhou, China). The rd1 mice (FVB/rd1) and C57BL/6 wild-type (WT) mice were purchased, respectively, from Vital River Laboratory Animal Technology Co. Ltd. (Beijing, China) and Southern Medical University (Guangzhou, China). All animals were kept in specific pathogen-free animal laboratory of Zhongshan Ophthalmic Center (Guangzhou, China) with standard feeding conditions. The retinas of two groups were collected at postnatal day 3 (P3), P5, P8, P11, P13, P15, P18, and P21 for the following experiments.

To evaluate retinal vascularization and vascular distribution, retinal flatmounts and retinal frozen section were determined by isolectin B4 (IB4) fluorescence staining at P3, P5, P8, P11, P13, P15, P18, and P21 as previously described [21, 26, 27]. After cardiac perfusion with 30 mL phosphate-buffered saline and 20–30 mL 4% (v/v) paraformaldehyde, the eyes were enucleated and fixed in 4% (v/v) paraformaldehyde for 45 min at 4°C, then dissected to make retinal flatmounts and retinal frozen section. The retinal flatmounts were permeabilized with 0.5% Triton X-100 in phosphate-buffered saline for 10 min, then blocked with 20% (v/v) fetal bovine serum (Gibco, USA) for 1 h at room temperature, and then stained with vascular marker Alexa 568-conjugated isolectin B4 (1:200, Invitrogen, Carlsbad, CA, USA) and anti-ki67 primary antibody (1:100, Cell Signaling Technology [CST], Danvers, USA) to detect proliferation at 4°C overnight [26]. The retinas were then labeled with Alexa Fluor^® 488-conjugated secondary antibody (1:1,000, CST) for 1 h at room temperature. Vascular distribution and tip cells were identified under a confocal microscope (LSM 980, Carl Zeiss AG, Oberkochen, Germany). Adobe Photoshop (Adobe, Inc., San Jose, CA, USA) and ImageJ (National Institute of Health [NIH], Bethesda, MD, USA) were used to analyze the retinal vasculature.

Mouse eyes enucleated at P3, P5, P8, P11, P13, P15, P18, and P21 were fixed and embedded in optimal cutting temperature compound (Sakura Finetek USA, Inc., Torrance, CA, USA) at −80°C and sectioned as previously described [26, 27]. The retinal cryosections were stained with Alexa^® 568-conjugated IB4 (1:200), anti-CLD-2 antibody (1:250, Abcam, Cambridge, MA, USA), anti-CLD-3 antibody (1:100, Invitrogen), and anti-RBPMS antibody (1:200, GeneTex, Inc., Irvine, CA, USA) to detect the RGCs at 4°C overnight. The retinas were then probed with anti-rabbit Alexa Fluor 488 secondary antibody (1:1,000, CST, Danvers, MA, USA) and then incubated with 4′,6-diamidino-2-phenylindole (1:2,000, Invitrogen) for 15 min to visualize the cell nuclei as previously described [27]. Sections were viewed under a confocal microscope. ImageJ software was used to examine the distributions of the retinal vessels.

Retinal proteins at P3, P5, P11, P15, P18, and P21 were extracted by radioimmunoprecipitation assay buffer (Beyotime, Shanghai, China) with protease (Absin, Shanghai, China) and phosphatase inhibitors (Absin, Shanghai, China). The protein concentration was measured by Pierce^™ BCA Protein Assay Kit (Thermo Fisher Scientific, USA) according to the manufacturer’s instructions. Western blotting was performed on SurePAGE and Bis-Tris gels (GenScript Biotech Corp., Biotech Corp., Piscataway, NJ, USA). The primary antibodies used were anti-β-actin (1:5,000), anti-VEGFA (1:1,000), anti-VEGFR2 (1:1,000) (all from CST), anti-CLD-1 (1:1,000) (HUABIO, Woburn, MA, USA), anti-CLD-2 (1:1,000, Abcam), anti-CLD-3 (1:1,000) (Invitrogen, USA), and anti-CLD-5 (1:1,000, Abcam). Anti-rabbit IgG HRP-linked antibodies (1:20,000, CST) were used to incubate membranes. Signals were detected with ChemiDoc^™ MP Imaging System (BioRad, USA). Quantification of the protein expression was measured by ImageJ. All experimental steps were conducted as previously described [13, 26, 27].

Retinal RNA at P3, P5, P8, P11, P13, P15, P18, and P21 was extracted with TRIzol reagent (Invitrogen) and synthesized the cDNA with a cDNA Synthesis Kit (TaKaRa Bio Inc., Kusatsu, Shiga, Japan). Synthesis of cDNA and RT-qPCR was performed as previously described [27, 28]. Information of the PCR primers was summarized in Table 1. GAPDH was used as the reference gene to normalize the total mRNA levels.

Eyeballs from rd1 and WT mice at P5, P8, P11, P15, P18, and P21 were fixed in FAS eyeball fixative solution (Servicebio, Technology Co. Ltd., Wuhan, China). Eye cups were embedded in paraffin, dehydrated with ethanol and xylene, and sectioned. The retinal paraffin sections were stained with hematoxylin and eosin and sealed with neutral glue as previously described [27]. HE-stained sections were used to analyze the retinal thickness of the regions, which is 350 μm adjacent to the optic nerve. The thickness of the inner part of the retina (IPR) and outer nuclear layer (ONL) and the number of RGCs were measured at the same region [29‒32]. IPR included the outer plexiform layer, inner nuclear layer, inner plexiform layer, ganglion cell layer, and nerve fiber layer (NFL) [30]. Images were scanned and measured with TissueFAXS (TissueGnostics Asia Pacific, Beijing, China).

Cell apoptosis was quantified in retinal frozen sections of P8, P11, P15, P18, P21 rd1 mice and WT mice using TdT-mediated dUTP nick-end labeling (TUNEL) In Situ Cell Death Detection Kit (Roche Diagnostics, Risch-Rotkreuz, Switzerland) as previously described [21].

Two independent masked investigators performed all counting and measurements in duplicate. All data are expressed as means ± standard error of the mean. Each experiment was repeated at least three times. The data were analyzed and the graphs were plotted with GraphPad Prism v. 8.0 (GraphPad Software, La Jolla, CA, USA), ImageJ (NIH), and SPSS v. 23.0 (IBM Corp., Armonk, NY, USA). One-way analysis of variance and Spearman’s correlation coefficients were calculated. A p value <0.05 was considered statistically significant.

There was no significant difference in the Ki67-positive ECs between rd1 mice and WT mice at P3, which is the early developmental stage of superficial vascular plexus (SVP). SVP of both groups was extremely active and growing toward the peripheral retina, and overall growth progress of SVP was slower in rd1 mice compared to WT mice. The WT mice had more Ki67-positive ECs than the rd1 mice at P5. At P8, the SVP termini formed a closed loop in WT mice; however, the rd1 mice had more Ki67-positive ECs than WT mice (Fig. 1a, c).

The number of tip cells in the SVP was not significantly different between the WT and rd1 mice at P3 (Fig. 1b, d). The rd1 mice had fewer tip cells than WT mice at P5, while the rd1 mice had more tip cells than the WT mice at P8 (Fig. 1b, d).

Throughout SVP development, the retinal vessel density of the WT mice was always denser than that of the rd1 mice (Fig. 1b, e). At P11, there was no difference in the vessel density of SVP between the WT mice and the rd1 mice. However, this similarity disappeared at P13. During P15 to P21, the retinal SVP of the rd1 mice whose network remained relatively intact was slightly less dense than that of the WT mice (Fig. 2a, b).

In both groups, the retinal deep vascular plexus (DVP) was not fully developed by P11, and there was a noticeable distance between the blood vessel network termini and the retinal periphery. Moreover, it was slightly less dense in the rd1 mice than in the WT mice. At P13, the vessel density of DVP in the WT mice was slightly but not significantly higher than that in the rd1 mice. In the rd1 mice, the retinal SVP and DVP rapidly degraded from P15 to P18, and the DVP was very sparse at P21. In contrast, the DVP in the WT mice remained stable from P13 to P21 (Fig. 2a, c). There was a significant difference in the vessel density of SVP and DVP between the WT and rd1 mice from P15 to P21 (Fig. 2b, c).

The retinal vascular distributions in the posterior pole, mid-peripheral, and peripheral parts of the retinal layers were compared using IB4 fluorescence staining of the retinal frozen sections at P8, P11, P13, P15, P18, and P21. At P8, the retinal blood vessels of the WT mice and rd1 mice were mainly localized at the superficial layer of the retinal posterior pole without significant difference. However, a few vessels started to develop into the retinal intermediate layers only in the WT mice. During P8 and P13, the blood vessels communicating between the superficial and intermediate layers were visible and showed no difference in the mice of both groups. In the rd1 mice, the retinal vessel density in the three parts of retina began to decrease from P15 onward, and the retinal SVP and DVP became very sparse at P21 (Fig. 3a–c). The peripheral retinal vasculatures of both groups were nearly equal only at P13. At all other timepoints, the retinal vessel density was higher in the WT mice than in the rd1 mice. By P15, the peripheral retinal vasculature reached stable in the WT mice but rapidly degraded in the rd1 mice (Fig. 3a, d).

Given that nearly 97% of photoreceptors are rods and localize in the ONL of the mouse retina, we evaluated the photoreceptors’ survival rate using the ONL thickness as an index [32, 33]. Examples of HE-stained sections of eye cups were shown in Figure 4a, and regions at 350 µm from the optic disk were enlarged and shown in Figure 4b. As the eye matures, the retinal thickness of WT mice exhibited a reduction in retinal thickness during P5–P8 and eventually stabilized at the end of the developmental period. Conversely, rd1 mice displayed a gradual attenuation and degeneration of retinal thickness from P5 to P21. The fluorescence intensity of RBPMS-positive RGCs in the rd1 mice gradually diminished from P11 onward (Fig. 4c). Here, the number of RGCs decreased in rd1 mice and there were significantly fewer RGCs in the rd1 mice than in the WT mice from P11 to P21 (Fig. 4c, i). This finding corresponded to the results of the HE staining (Fig. 4h).

The retinal thickness and structure were comparably similar in both types of mice before P11. However, the IPR, the ONL, the full-retina thickness, and RGCs in rd1 mice underwent a precipitous degeneration since P15 (Fig. 4d). Total retinal vessel density (including superficial vascular density and deep vascular density) was closely and positively correlated with changes in total retinal thickness, the IPR thickness, and the ONL thickness (r = 0.8883, p < 0.01; r = 0.7036, p < 0.01; r = 0.6893, p < 0.01, respectively) (Fig. 4e–g). Changes in the density of total retinal vessels, SVP and DVP in rd1 mice were found to be positively correlated with the number of RGCs (r = 0.8543, p < 0.01; r = 0.5255, p < 0.05; r = 0.6418, p < 0.01, respectively) (Fig. 4h–J). RGCs underwent degeneration associated with retinal vascular degeneration (Fig. 4h–J).

The expressions of four vessel-related CLD in rd1 mice and WT mice were investigated at different timepoints. The peak mRNA expression of CLD-1 and CLD-2 in rd1 mice was earlier than that of WT mice, but the overall mRNA expression levels of CLD-1 and CLD-2 were lower than those of WT mice. The mRNA expression levels of CLD-3 and CLD-5 in rd1 mice were lower than those of WT mice during the overall developmental period (Fig. 5a–d).

During the retinal development period of WT mice, the protein expression of CLD-1 and CLD-5 showed a gradually increasing trend, while the protein expression of CLD-2 and CLD-3 showed a trend of increase at first and then decrease later (Fig. 5e). The protein expression levels of CLD-1, CLD-2, CLD-3, and CLD-5 in the retinas of rd1 mice were significantly lower than those of WT mice, especially at the late development stage when the degeneration of retinal deep blood vessels occurred (Fig. 5e, f). The protein expression of CLD-3 in the rd1 mice was extremely low at P21 (Fig. 5e, h).

Immunofluorescence staining was used to detect the expression of CLD-2 and CLD-3 in retinal frozen sections at P18, which is a period marked by the active formation of the BRB [34]. The co-localization between CLD-2 (green) and vascular marker, IB4 (red), was observed (Fig. 6a), indicating the expression of CLD-2 was in retinal vascular ECs. CLD-3 (green) was not co-stained with IB-4 (red) (Fig. 6a) but was found to co-localize in the NFL (Fig. 6d). These results indicated that CLD-3 was mainly expressed in the mouse RGCs, which was consistent with our previous research [21]. At P18, significant differences in vascular and neural-associated expressions were observed between the two groups. Compared to WT mice, the rd1 mice demonstrated a notable reduction in the expression level of CLD-2 and CLD-3, as well as in the RGCs and the retinal vessels (Fig. 6a–f).

Vascular endothelial growth factor A (VEGFA) and vascular endothelial growth factor receptor 2 (VEGFR2) are closely associated with angiogenesis. Expression of VEGFA showed a gradually increasing trend in the WT mice but a gradually decreasing trend in the rd1 mice. And VEGFA was markedly lower in the rd1 mice than in the WT mice at all subsequent timepoints except at P3 (Fig. 7d, e). Expression of VEGFR2, which mediates VEGF signaling in ECs, was lower in rd1 mice than in WT mice (Fig. 7d, ) from P15 onward.

Apoptosis detected by TUNEL staining was first observed in the ONL of rd1 mice at P8 (Fig. 8a). The most intensive apoptosis occurred in the retinas of rd1 mice at P11 and P15 due to the dramatic degeneration of the ONL, and the number of TUNEL-positive cells decreased at P18 and P21 (Fig. 8a, b). Interestingly, the TUNEL-positive cells were also found in the NFL at P11 (Fig. 8a). In contrast, fewer TUNEL-positive cells in the layers of the retina could be observed in WT mice at all timepoints (Fig. 8a).

Retinal vascular development in normal WT mice typically progresses through three stages [35]. Beginning on P1, vessels sprout from the optic disc head and spread outward. By P8, these superficial vessels reach the peripheral retina and then dive deeper to form the DVP. From P15, the superficial and deep retinal vascular networks start to form and fully establish by P21.

Our study indicated that by P8, retinal vessels in rd1 mice did not fully extend to the peripheral retina and this might result in more tip cells at the SVP in rd1 mice than WT mice. Fewer active vascular sprouts grow in rd1 mice than in WT mice, which might further affect the development of DVP. During P3–P13, the development of SVP and DVP of rd1 mice was slightly delayed but fully completed. However, from P15 onward, the retinal vessels in rd1 mice began to degenerate. By P21, the degeneration of retinal vasculature in rd1 mice was very noticeable. The SVP density of rd1 mice was significantly sparse compared to that of WT mice, and only a few residual branches and broken ends remained in DVP, which could not form a complete deep vascular network structure with normal function and shape. We also found retinal vessel degeneration in rd1 mice starting from small, deep retinal vessel branches in distal region to superficial large vessels in proximal area. This finding was consistent with previous studies [5, 36].

Angiogenesis involves novel blood vessel sprouting [37, 38]. It is commonly accepted that an abnormal hyperoxic state triggers retinal vascular regression. In other words, the characteristic retinal vascular attenuation is considered to be a secondary change occurring after photoreceptor loss [39]. The delayed development of SVP might be due to ocular blood supply insufficiency and mitochondrial stress in the early development period [40]. DVP network degeneration may be due to oxidative stress and blood oxygen supply-demand imbalances. This rd1 model aligns with peripheral-to-posterior pole vessel degeneration, which is observed in RP patients [1, 41]. The highest density of DVP in rd1 mice at P13 suggested the delayed deep vessel development, possibly due to the activation of full photoreceptor, accumulation of cyclic guanosine monophosphate, and reactions of oxidative stress [39, 42]. The vessel degeneration in our study was also consistent with previous studies [43, 44]. Oxidative stress disrupts the interaction between ECs and pericytes, which could lead to the responsiveness of ECs to VEGFA and hinder DVP formation [45‒48]. Therefore, retinal hyperoxygenation induced by an early loss of photoreceptors under the rd1 mutant background suppressed the expression of VEGFA/VEGFR2 and vessel-related CLD, including CLD-1, CLD-2, CLD-3, and CLD-5 in the retina of rd1 mice. CLD-5 is an important tight junction protein that constitutes the BRB, and the change of CLD-5 expression is related to VEGF [14, 48, 49]. The HIF2-alpha/VEGF feedback loop in Müller glial has been confirmed to be induced by severe hypoxia and acts on the growing retinal vasculature [48]. Additionally, there are interplays among CLD-1, CLD-2, CLD-3, and CLD-5 in the retina [14, 20, 50]. We can thus hypothesize that factors such as hyperoxia and decreased VEGFA in rd1 mice might act on retinal glial cells, ECs, and retinal neurons, thereby further affecting retinal neuro-vascular development and ultimately causing retinal neuro-vascular degeneration in rd1 mice. While some scholars have found that patients with RP have primary pathological reductions in retinal blood flow early in life. These manifestations are parts of the primary vascular dysregulation syndrome in RP, which leads to further photoreceptor damage and worsens the course of the disease [6, 51]. Therefore, vascular dysfunction in rd1 mice may be a result of metabolic changes but also possibly a primary alteration due to a genetic mutation. This inference still requires substantial follow-up studies.

We found that during developmental stages, changes in total retinal vascular density and apoptosis were positively correlated with the thickness of the ONL, which was consistent with previous study [5]. We hypothesized that in the developmental stage of rd1 mice, the choroidal circulation might be still unable to regulate apoptosis of the outer retina, retinal inner circulation, especially the DVP, but demonstrate a tolerance and regulatory response to the apoptosis of the photoreceptor cells. Previous studies have confirmed that DVP may satisfy the oxygen needs of photoreceptor cells’ inner segments [41]. Meanwhile, approximately 15% of the blood supply for photoreceptors near the macular fovea comes from the retinal circulation [52‒54]. This regulatory phenomenon of DVP in response to changes in oxygen supply is also observed in diabetic retinopathy and high myopia [55, 56]. The apoptosis of the outer retina occurring in the developmental stages of rd1 mice is sensitive to the oxygen supply from the DVP, but this does not contradict the fact that the main source of blood supply to the outer retinal layer primarily originates from the choroidal circulation [57]. The specific mechanism and the role of retinal blood supply in neurovascular degeneration in rd1 mice is worth an in-depth study.

RGCs are unique among retinal neurons as they directly project their axons into the central nervous system and serve a critical role in visual function [58]. Loss of RGC in rd1 mice at 1 year of age does not exceed 6% of those in WT mice [59]. However, Steven et al. [60] have also found that rd1 mice develop a decreased ganglion cell response later in development. In this study, we proved the loss and degeneration of RGCs was at the late stage of the developmental period in rd1 mice, and the expression of CLD-3 was a significant reduction, which was closely related to RGCs [20, 21]. These findings are consistent with previous studies using disease models of RGC injury [20]. Programmed cell death of RGCs in WT mice begins at P2, continues throughout development, and ends in adulthood [61]. We found that the loss of RGCs was related to changes in SVP and DVP and the degeneration of retinal vessels was earlier than the loss of RGCs in rd1 mice. RGCs derive their blood supply from all layers of the retinal vasculature and their axons run through the retina. RGCs have certain physiological remodeling and compensatory functions, so a therapeutic time window would occur between vascular degeneration and neuro-degeneration in rd1 mice [62]. We proposed that interventions targeting retinal vasculature might bear greater practical significance in the early treatment of retinal degenerative diseases. Our previous research has demonstrated that downregulation of CLD-3 can induce apoptosis of RGCs by suppressing the expression of Bcl-2 and VEGFA [21, 22]. CLD-3 functions as an intermediate junction station for proteins and pathways associated with retinal nerve and vascular development [50, 63]. Our previous study has indicated that the downregulation of CLD-3 in zebrafish might have an impact on the Notch pathway [22]. In this study, we have observed a reduction in CLD-3 levels, concomitant with the downregulation of VEGFA/VEGFR2 in rd1 mice. Therefore, loss of RGC might be due to hypoxia, retinal vessel degeneration, or downregulation of CLD-3 [20, 64]. These needed further in-depth investigations.

This study demonstrated that rd1 mouse had characteristics of retinal vasculature degeneration and retinal neuro-degeneration of ONL and RGCs. rd1 mouse also showed a notable decline in the expression of vessel-related CLD proteins, which might play a pivotal role in neurovascular development. The data indicated rd1 mouse is a valuable animal model of not only retinal neuro-degeneration but also retinal vascular degeneration and remodeling.

This study protocol was reviewed and approved by the Institutional Review Board of Zhongshan Ophthalmic Center (Guangzhou, China), Approval No. Z2021080.

The authors have no conflicts of interest to declare.

This work was supported by the National Natural Science Foundation of China to Yan Luo (81770971, 81371020).

Aoxiang Wang completed experiments and wrote paper. Jinxi Zhou completed image processing. Yiwen Hong, Yishen Wang, and Jianying Pan supported the data analysis. Yamei Cui and Yue Wu reviewed the paper and gave suggestions on it. Yan Luo contributed to the experiment design and paper supervision.

Data are not available due to ethical reasons. Further inquiries can be directed to the corresponding author.