Molecular Vision 2020; 26:679-690
Received 03 February 2020 | Accepted 08 October 2020 | Published 10 October 2020
The last two authors are co-senior authors for this study.
1Department of Optometry and Vision Science, University of Melbourne, Victoria, Australia; 2The John Curtin School of Medical Research, The Australian National University, Acton, Australia; 3The Australian National University Medical School, Acton, Australia; 4Department of Anatomy and Neuroscience, University of Melbourne, Victoria, Australia
Correspondence to: Haihan Jiao, University of Melbourne, Victoria, Australia 3052; Phone: +61390357545; email: firstname.lastname@example.org
Purpose: Dysregulation of the complement cascade contributes to a variety of retinal dystrophies, including age-related macular degeneration (AMD). The central component of complement, C3, is expressed in abundance by macrophages in the outer retina, and its ablation suppresses photoreceptor death in experimental photo-oxidative damage. Whether this also influences macrophage reactivity in this model system, however, is unknown. We investigate the effect of C3 ablation on macrophage activity and phagocytosis by outer retinal macrophages during photo-oxidative damage.
Methods: Age-matched C3 knockout (KO) mice and wild-type (WT) C57/Bl6 mice were subjected to photo-oxidative damage. Measurements of the outer nuclear layer (ONL) thickness and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining were used to assess pathology and photoreceptor apoptosis, respectively. Macrophage abundance and phagocytosis were assessed with immunolabeling for pan-macrophage and phagocytic markers, in conjunction with TUNEL staining in cohorts of C3 KO and WT mice.
Results: The C3 KO mice exhibited protection against photoreceptor cell death following photo-oxidative damage, which was associated with a reduction in immunoreactivity for the stress-related factor GFAP. In conjunction, there was a reduction in IBA1-positive macrophages in the outer retina compared to the WT mice and a decrease in the number of CD68-positive cells in the outer nuclear layer and the subretinal space. In addition, the engulfment of TUNEL-positive and -negative photoreceptors by macrophages was significantly lower in the C3 KO mice cohort following photo-oxidative damage compared to the WT cohort.
Conclusions: The results show that the absence of C3 mitigates the phagocytosis of photoreceptors by macrophages in the outer retina, and the net impact of C3 depletion is neuroprotective in the context of photo-oxidative damage. These data improve our understanding of the impact of C3 inhibition in subretinal inflammation and inform the development of treatments for targeting complement activation in diseases such as AMD.
The activation of an immune cascade known as the complement system is a crucial factor in the etiology of age-related macular degeneration (AMD). Complement contributes to a major arm of the innate immune response and provides a rapid response to a range of immunological challenges [1,2]. Three activating pathways comprise the complement cascade (classical, mannose-binding lectin, and alternative), and all converge on the proteolytic cleavage of C3 to generate an arsenal of inflammatory mediators, including the opsonin C3b and the anaphylatoxin C3a. This process is mediated chiefly by C3-convertases that are assembled from complement constituents such as C2, C4, CFB, and CFD in the presence of noxious stimuli. If left to propagate, the accumulation of C3b triggers cleavage of C5 to produce C5a and C5b, with the latter inducing the assembly of a membrane attack complex (MAC) that binds to cell surfaces, forming transmembrane channels that cause cytolysis or apoptosis of the target cells.
Polymorphisms in complement genes have been reported to account for approximately 70% of the risk for developing AMD, as demonstrated in multiple, independent genome-wide association studies (GWASs) [3,4]. The first reported strong associations between the Y402H polymorphism in complement factor H (CFH) with all forms of AMD, while later investigations also identified risk-conferring variants in other complement genes, including C3, C2, and CFB [3,4]. The current dogma of CFH function is to downregulate the complement cascade, by inhibiting cleavage of C3, limiting the generation of byproducts that spur inflammation. This feature prioritizes modifiers of C3 activation as potential therapeutics for ameliorating complement , although additional molecular studies are required to develop the groundwork needed for effective targeting of complement. Dysregulation of the cascade is also linked to a range of neurodegenerative conditions, including traumatic brain injury, Alzheimer disease, and Parkinson disease .
Complement expression and activation occur locally within the retina under pathological conditions with causative effects that are driven by a combination of genetic and environmental factors, such as photo-oxidative damage (PD) [7–9], smoking, and obesity , and leading to dysregulation of inflammatory cascades . Over the last several years, several investigations have shown that macrophages are a source of complement components, particularly C3, and related regulatory factors [12,13]. These include rodent models of aging , obesity , PD [9,16,17], and most recently, retinitis pigmentosa . The identification of retinal macrophages as a major source of C3 suggests that the abundance of these cells in the outer nuclear layer (ONL) and the subretinal space dictates the scale and duration of local complement activation that leads to cell death. The abundance of mononuclear phagocytes in the outer retina has been a feature of AMD [19,20] and occurs in several retinal degenerative animal models [19,21-24]. Our most recent study indicated that C3 is locally expressed by ONL and subretinal macrophages in postmortem tissues from patients with atrophic AMD, and that total ablation of C3 provides neuroprotection in PD . This investigation and the previous report on mouse laser-induced choroidal neovascularization (CNV)  demonstrated that serum sources of complement are dispensable for inducing complement-mediated attack in PD and CNV, while local inhibition of C3 via RNA interference (RNAi) ameliorates complement deposition in the ONL and the subretinal space. A similar neuroprotective effect has since been observed in an experimental model of glaucoma .
Intriguingly, a recent study by Silverman and colleagues showed that total ablation of C3 suppresses photoreceptor death and that macrophages/microglia exhibit reduced phagocytosis of apoptotic photoreceptors . The researchers’ observations suggest that a degree of local complement expression may be necessary to facilitate apoptotic cell clearance and homeostasis, which is neuroprotective in instances where excessive and inappropriately regulated complement is not present. Recently, we showed that ablation of local C3 reduces photoreceptor death in the context of PD , although changes in the activity of reactive macrophages concerning the C3 ablation in this system are unknown. In the present study, we investigated the impact of C3 on macrophage morphology, abundance, and photoreceptor phagocytosis in the PD model and further explored the protective phenotype. The data show that macrophages of the outer retina (the ONL and the subretinal space) in C3 knockout (KO) mice exhibited a decrease in immunoreactivity for CD68 following PD. In addition, phagocytosis of terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL)-positive and TUNEL-negative photoreceptors by macrophages was suppressed in the C3 KO mice after PD.
All experiments were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Australian National University (ANU) Animal Experimentation Ethics Committee. Mice homozygous for C3 mutation, C3−/− mice (C3tm1Crr/J), and age-matched, isogenic wild-type (WT) littermates, bred on the C57BL/6J background, were used at 3 months of age. Both groups were subjected to PD following our established protocols , with food and water provided ad libitum. The mice were euthanized with cervical dislocation at either 1, 3, 5, or 7 days of PD.
TUNEL was used to quantify photoreceptor apoptosis in retinal cryosections. The sections were permeabilized with 0.01 M PBS (1X; 120 mM NaCl, 20 mM KCl, 10 mM NaPO4, 5 mM KPO4, pH 7.4) and 0.1% Triton-X100 for 10 min at room temperature and then labeled using a TUNEL assay (Roche, Indianapolis, IN) according to the manufacturer’s specifications. Counts of TUNEL-positive cells in the ONL were performed blind to the condition and genotype, and along the full-length of retinal cryosections cut along the vertical meridian (superior-inferior) including the optic disc. The final count for each animal was the average of those obtained at comparable locations in two retinal cryosections.
Retinal cryosections sections were used for immunohistochemical analysis. The primary antibodies are listed in Table 1. Some retinal sections were stained with TUNEL before immunohistochemistry. Cryosections were firstly subjected to heat-induced antigen retrieval using 100% Reveal-it (ImmunoSolution, QLD, Australia) for 1 h at 37 °C, then washed in 0.1 M PBS containing 0.3% Reveal-it, and blocked and permeabilized (10% normal goat serum in 1X PBS with 0.1% Triton X-100 for 1 h at room temperature). Following overnight incubation of the primary antibodies at 4 °C, the retinal sections were washed in 0.1 M PBS and then incubated with secondary antibodies: Alexa-488 conjugated goat anti-rabbit for IBA1 and GFAP, 4.0 h at room temperature and Streptavidin Alexa-594 conjugated anti-biotin for CD68, 1.5 h at room temperature (Thermo Fisher Scientific, NSW, Australia). The sections were then incubated with the DNA-specific dye bisbenzimide (1:10,000, Sigma-Aldrich, NSW, Australia) for 2 min, washed, and then coverslipped using Aqua-Poly/Mount (Polysciences, Warrington, PA).
Immunofluorescence in the retinal sections was visualized using a laser-scanning A1+ confocal microscope (Nikon, Tokyo, Japan), and images were acquired using the z-stack function of the NIS-Elements AR software (Nikon, Tokyo, Japan) under the same acquisition settings. Quantification analyses included retinal morphometric analysis and total counts of immunolabeled cells (GFAP, IBA, and CD68). Analyses were based on the average of two sections per experimental group of animals (n = 5–6 per group).
The immunofluorescence intensity of GFAP and the length of immunolabeled GFAP-positive Müller glia were quantified in the confocal images of the WT and C3 KO mouse retinal sections. Confocal images of GFAP-labeled sections were processed and analyzed using the NIS-Elements AR software. Consistency in the analyses was ensured by employing identical parameters with areas sampled for the analysis, camera, and laser configuration. For each sample, an average of three consistent locations in the superior retina (500, 1,000, and 1,500 μm eccentricity) was used to account for geographic variability. The inferior retina was excluded because GFAP changes in this region were less pronounced for reliable quantification.
For quantification of fluorescence, the areas of interest were marked with a rectangle drawn from the apical surface of the ganglion cell layer (GCL) to the basal surface of the ONL, from which the mean fluorescent intensity for each retina was then collected. The mean GFAP fluorescent intensity was measured from each rectangle across the superior retina, and the average of two sections from each animal was taken (n = 5 per group). The length of the GFAP-positive Müller glia process was taken from within the lesion region of the superior retina, using selection tools from NIS-Elements AR software. The measurements were then averaged for each experimental group (n = 5 per group).
The abundance and morphology of outer retinal macrophages in the WT and C3 KO mouse groups were quantified on retinal cryosections immunolabeled with IBA1 over the time course of PD (n = 5 per group). The amoeboid IBA1-positive macrophages were characterized by the circularity which is defined as perimeter2 / area (P2/A) and validated with the grid-cross analysis method. Higher circularity and a lower number of grid-crossed points per cell indicate a round cell, thus amoeboid morphology, whereas lower circularity indicates ramified morphology. The total number of ramified and amoeboid IBA1-positive cells was separately quantified in the outer retina (the ONL and the subretinal space) across the full length of the retinal cryosections in duplicate.
The phagocytic properties of outer retinal IBA1-positive macrophages following PD were assessed with quantification of immunolabeling for CD68 and their engulfment of photoreceptors. The total number of cells identified as IBA1-positive CD68-positive following immunolabeling on the retinal cryosections was quantified in the outer retina across the full length of the retinal cryosections (superior-inferior) after PD. The number of photoreceptor nuclei engulfed by outer retinal macrophages was determined using sections that were labeled for IBA1, TUNEL, and bisbenzamide. Counts were made of IBA1-positive cells in the ONL that had engulfed either TUNEL-positive or TUNEL-negative cells in each section. All counts were performed in six areas across the full length of the retina in duplicate (n = 6 per group).
For each quantitative confocal datum, the labels of the stained retinal cryosections were masked to collect and analyze the data in a blinded fashion. The data were quantified from an average of three consistent locations in the superior retina (500, 1,000, and 1,500 μm eccentricity), to account for geographic variability.
The retinal cryosections were stained with toluidine blue to assess changes in ONL integrity following PD and then quantified using an established methodology . Briefly, the number of rows of photoreceptor nuclei was counted in six areas of the retinal cryosections (superior-inferior) from each group, the central, mid-periphery and peripheral retina, at 500 μm intervals across each section. In at least two retinal sections per animal, photoreceptor rows were counted per area and then averaged for each experimental group (n = 6 per group).
Statistical analyses were performed in Prism 7 (GraphPad Software, San Diego, CA). Statistical significance was determined by using two-way analysis of variance (ANOVA) and the Student t test. Two-way ANOVA to compare the photoreceptor row counts and parameters of subretinal macrophage reactivity in C3 KO and WT animals with Tukey’s post-hoc tests were performed to validate the comparisons. An unpaired Student t test was used to compare retinal GFAP responses (intensity and process lengths) in the C3 KO and WT mice. An alpha of less than 0.05 (p<0.05) was considered statistically significant.
The number of photoreceptor rows in the ONL was quantified to determine photoreceptor layer thickness after PD between the C3 KO and WT cohorts (Figure 1). The C3 KO and WT mice displayed fewer rows of photoreceptors in the superior retina, which is the focal region of damage in this model  after 7 days of PD (p<0.05, Figure 1A). In the C3 KO cohort, there was statistically significant rescue of photoreceptors in the superior retina compared to the wild-type controls (p<0.05, Figure 1A). Histologically, there was no perturbation in the retinal structure evident in either the WT or C3 KO mice (Figure 1B,C), as reported previously . After 7 days of PD, there was severe thinning and disruption of the ONL in the WT mice. The thinning was not as apparent in the C3 KO cohort compared to the WT cohort (Figure 1D,E). The most pronounced disturbances in the WT retinas were observed in the central (about 500 μm eccentricity) and peripheral locations (about 1,000 μm eccentricity; Figure 1F,G). In the C3 KO retinas, there was some thinning of the ONL in the central location (about 500 μm eccentricity) following photo-oxidative damage, while the ONL was relatively intact in the peripheral retina (1,000 μm eccentricity; Figure 1H,I). In the WT cohort, there was also disruption of the RPE and the Bruch’s membrane complex, while the outer retina featured large, amoeboid cells in the subretinal space (Figure 1J). Visually, there appeared to be no disruption of the RPE monolayer or Bruch’s membrane in the C3 KO animals (Figure 1K).
In conjunction with these data, immunoreactivity for GFAP Müller cells was also assessed in the WT and C3 KO cohorts as a generalized marker of retinal stress . GFAP immunoreactivity was not detectable in the Müller cell processes of either the dim-reared C3 KO or WT mice (Figure 2A,C), with no statistically significant difference detected in the GFAP fluorescence intensity between the two groups (Appendix 1, p>0.05). At 7 days of PD, GFAP immunoreactivity in Müller cells was readily apparent in the WT retinas, although it appeared less abundant in the C3 KO retinas (Figure 2B,D). The GFAP-positive processes extended from the inner retina into the outer nuclear layer in the WT retinas after photo-oxidative damage. These data were quantified through measurement of pixel intensity (Figure 2E) and the average length of the GFAP-positive Müller cells processes (Figure 2F), which were significantly lower in the C3 KO retinas compared to the WT retinas (p<0.05).
The effect of C3 knockout on the migration and morphology of outer retinal macrophages was assessed across the time course of PD at 1, 3, 5, and 7 days (Figure 3A–F). Using immunolabeling for IBA1, the number of macrophages (the ONL and the subretinal space) in the WT mice increased progressively over the time course of PD from 3 days and reached a peak at 5 days (Figure 3G). In the C3 KO mice, however, the number of outer retinal macrophages was found to be reduced at 5 and 7 days compared to the WT retinas (p<0.05, Figure 3G). Morphological analysis of the microglia/macrophages situated within the ONL and the subretinal space was also conducted in the WT and C3 KO cohorts following PD (Figure 3H–J). Quantification of amoeboid versus ramified cells in this region demonstrated that an activated phenotype, characterized by the amoeboid morphology (Figure 3H, arrowheads), was less prevalent in the C3 KO mice after PD compared to the WT mice (p<0.05, Figure 3J). There was also trend in a higher number of ramified quiescent IBA1-positive cells in the C3 KO retinas (Figure 3I, arrowhead), although this trend was not statistically significant (p>0.05, Figure 3J).
The distribution of phagocytosis-related markers in subretinal macrophages and their engulfment of photoreceptors were investigated to assess whether the morphological changes observed in the C3 KO mice (Figure 3J) were accompanied by changes in phagocytosis (Figure 4). Immunolabeling for CD68/macrosialin was used to interrogate phagocytosis and scavenger receptor activity in macrophages of the central nervous system (CNS; reviewed in ). After 7 days of PD, there was an abundance of IBA1-positive outer retinal macrophages that were immunoreactive for CD68 in the WT mice (Figure 4A,C), and these macrophages were more numerous in the superior retina where the focal point of injury is situated (p<0.05, Figure 4C). In contrast, there were far fewer ONL and subretinal macrophages that were CD68-positive in the C3 KO mice at the same time point (p<0.05, Figure 4B,C).
The IBA1-positive cells were colabeled with TUNEL to determine the change in the phagocytosis of photoreceptors by outer retinal macrophages in the WT and C3 KO cohorts (Figure 4D–I). These data were used to ascertain not only the phagocytosis of apoptotic TUNEL-positive photoreceptors but also the phagocytosis of stressed but still viable TUNEL-negative photoreceptors, a process known as phagoptosis . Double-labeling indicated that phagocytosis of TUNEL-positive photoreceptors by IBA1-positive cells was evident at the focal point of the lesion (Figure 4D, approximately 500 μm from the ON) and surrounding regions (Figure 4E,F,G, approximately 1000 μm from the ON) in the WT mice after 7 days of PD. In the C3 KO mice, this phagocytosis was restricted to the central retina, which was the center of the lesion at 7 days of PD (Figure 4H,I). When quantified, approximately 55% of the phagocytosed photoreceptors were TUNEL-positive in the WT retinas at 7 days of PD, while the remaining were TUNEL-negative (Figure 4J). In the C3 KO cohort, there were significantly fewer phagocytosed TUNEL-positive photoreceptors compared to the WT mice (p<0.05, Figure 4J). The number of phagocytosed viable photoreceptors (TUNEL-negative) was also lower in the C3 KO mice compared to the WT mice (p<0.05, Figure 4J). Stratification of TUNEL-negative/IBA1-positive cells in 500 μm increments across the vertical meridian indicated a peak in phagocytosis of TUNEL-negative photoreceptors at the lesion periphery or edge (1,000 μm eccentricity) in the WT mice (p<0.05, Figure 4K). In the C3 KO mice, however, the number of TUNEL-negative/IBA1-positive cells in this region (1,000 μm) was decreased by approximately one third (p<0.05, Figure 4K).
The present investigation sought to determine whether the neuroprotective phenotype of C3 ablation in PD is accompanied by changes in macrophage reactivity within the outer retina. The present study data show that C3 ablation suppressed photoreceptor death in a region-specific manner (in which more photoreceptors in the lesion margin were spared) and was associated with a reduction in the retinal stress marker GFAP. Furthermore, we observed spatiotemporal changes in the reactivity of outer retinal macrophages in C3-deficient mice following PD. First, there was a reduction in the number of macrophages in the ONL and the subretinal space in the C3 KO mice, and a reduction in those exhibiting an ameboid morphology compared to the WT mice. Second, outer retinal macrophages of the C3 KO mice were found to have reduced CD68-positive phagosomes compared to the WT mice. Finally, there was a decrease in the phagocytosis of TUNEL-positive and TUNEL-negative photoreceptors by macrophages in the C3 KO mice.
Previous investigations have shown that subretinal macrophages are predominantly mobilized resident microglia in PD . The subretinal populations come from their niches in the inner retina, switching their functions toward proinflammatory as an attempt to protect the compromised RPE [32,33]. As PD recapitulates some events akin to the atrophic AMD [7,17], the present study data suggest that the central complement component C3 amplifies the subretinal macrophage activity and the progression of retinal degeneration. Targeting the local sources of C3  can be potentially beneficial to mitigate the adverse effects of subretinal macrophages in the degenerating retina. Complement aids in furnishing macrophages with tools for recognizing and opsonizing cellular debris and apoptotic cells , and is associated with progressive retinal degeneration . C3-dependent phagosome expression in macrophages contributes to neurodegeneration , including that observed in the retina . In the present study, the lower number of phagosomes and ameboid macrophages coincides with the reduced phagocytosis following the ablation of C3, and with the mitigated progression of retinal atrophy.
Among the recent literature, the present study results suggest that the contribution of complement and macrophages to retinal degeneration is context-dependent, governed by the nature of the pathological cues encountered. In the rd10 mouse model of retinitis pigmentosa (RP), the recent investigation by Silverman and colleagues indicated that complement aids in the clearance of accruing apoptotic photoreceptors by infiltrating macrophages/microglia. Although the ablation of C3 elicited dampened phagocytosis of photoreceptors in both investigations, the present data report a neuroprotective phenotype of C3 ablation in PD, rather than an overtly deleterious response as shown in RP. Moreover, genetic ablation of C3 accelerates photoreceptor degeneration, subretinal deposits, all of which link to impaired phagocytosis of RPE in aged C3 KO mice . The reason for this discrepancy could lie in the larger extent of degeneration and maladaptive complement activity in the PD model, including activation of the proinflammatory alternative pathway of the cascade. In previous studies, ablation of alternative factors B and D has been shown to suppress local complement and photoreceptor death following PD [15,38]. Other maladaptive contributions of complement have been documented in loss of retinal ganglion cell synapses in the early and moderate stages of glaucoma in DBA/2J mice  and in an inducible rat model of glaucoma .
The reduction in phagocytosis of TUNEL-negative photoreceptors led us to speculate about a potential role for complement in promoting cell death through the phagoptosis of stressed but still-living photoreceptors. Phagoptosis arises from the expression of “eat me” signals of viable cells and has been proposed as a contextual trigger for the destruction of neurons by activated microglia in the CNS (reviewed in [29,30]), including brain ischemia . A maladaptive function of outer retinal macrophages in their elimination of neighboring viable rod photoreceptors has been reported in the rd10 model . Dampened phagocytosis of TUNEL-negative photoreceptors was also reported in the Silverman et al. study despite the deleterious effect of ablating C3, although a causative role of complement in this process is unclear. Nevertheless, the potential link between phagoptosis and complement in retinal degeneration certainly merits more in-depth investigation in future studies.
This investigation showed that ablation of C3 suppressed macrophage activity over the time course of PD, as evidenced by the altered morphology and cell density of macrophages in the outer retina. The ablation of C3 proved to be neuroprotective in PD despite the reduction in photoreceptor phagocytosis, suggesting an overtly maladaptive complement response in the given context. Although targeting complement could be useful in mitigating retinal atrophy , the results also indicate that targeting complement activity should be sought with caution to avoid impacting the phagocytic clearance.
Appendix 1. GFAP immunoreactivity in dim-reared WT and C3 KO retinas display no significance in the fluorescence intensity between the groups (p>0.05).
This study was supported by project grants from the National Health and Medical Research Council (APP1165599; APP1127705), Australian National University Translational Fellowship and Australian Government Research Training Program Scholarship.