Subretinal macrophages produce classical complement activator C1q leading to the progression of focal retinal degeneration

All procedures were performed in accordance with the Association for Research in Vision and Ophthalmology (ARVO) Statement for Use of Animals in Ophthalmic and Vision Research. The study was approved by the Australian National University Animal Experimentation Ethics Committee (Ethics ID: A2014/56 and A2017/41). Complement component 1, q subcomponent, alpha polypeptide knock out (C1qa^−/−) mice were used for this study (C1qa < tm1Mjw>). C57BL/6 J mice were used as wild-type (WT) controls. Animals were born and raised in dim (5 lx) cyclic light conditions in 12:12 h cycle, and house in individual vented cages (IVCs). Food and water was provided in constant supply and cages changed on a weekly basis. Adult mice aged between postnatal day (P) 60-90 were used for all experiments. Equal numbers of male and female mice were used throughout the study to avoid any gender biases. Animals with no exposure to photo-oxidative damage were used as dim-reared controls.

To induce photo-oxidative damage (PD), mice were housed in Perspex boxes coated with a reflective interior, and exposed to 100Klux of natural white LED for 1, 3, 5 and 7 days, with free access to food and water, as described in our previous methodology [25]. Some animals were recovered in dim light conditions and were collected at 8, 10 and 14 days following the onset of PD. Each animal was administered with pupil dilator eye-drops twice daily during PD (Atropine Sulphate 1%w/v eye-drops; Bausch and Lomb, NSW, Australia). Electroretinography (ERG) was used to measure mouse retinal function in response to full-field flash stimuli under scotopic conditions as described previously [25]. Animals were euthanized with CO₂ prior to tissue collection. Eyes were collected and processed for cryosections or RNA extraction, as previously described in our publications [12, 25].

C1q neutralisation strategies included local pre-treatment, local post-treatment and systemic treatment following the paradigms listed below. Briefly, a neutralising antibody to C1q, ANX-M1 and a non-specific isotype control IgG monoclonal antibody (mAb) were provided by Annexon Biosciences (CA, USA). Monoclonal antibody M1 binds and neutralises C1q thereby preventing the activation of the classical complement pathway. ANX-M1 anti-C1q antibody or an IgG isotype control antibody were formulated in endotoxin-free 0.1 M phosphate-buffered saline (Thermo Fisher Scientific, MA, USA). All experiments were performed on adult C57BL/6 J mice under double-blind conditions. Intravitreal injections were performed as described in our previous publication, with a volume of 1 μl injected into each mouse eye [26].

100 mg/kg of ANX-M1 anti-C1q antibody or an IgG isotype control antibody was administered intraperitoneally on day 0 (prior to the PD procedure), day 4 (during the PD procedure) and on day 8 (the day after completion of the PD procedure) in order to sustain complement inhibition. Hemolytic assay and retinal assessment were performed at day 12 (N = 10/group) as per our previous publications [11, 27].

A terminal deoxynucleotidyl transferase (TdT) dUTP nick-end labelling (TUNEL) assay (Roche Diagnostics, Basel, Switzerland) was used to detect photoreceptor cell death. Retinal cryosections were used for the TUNEL assay according to our previously described protocols, with minor modifications [28, 29]. Sections were counterstained with the DNA-specific dye bisbenzimide (1:10000; Sigma Aldrich, MO, USA) for visualisation of cell nuclei.

To detect and localise specific proteins, immunohistochemistry was performed on retinal cryosections according to previously published methodology, with minor modifications [12, 30]. A list of primary antibodies is provided in Table 1.

TUNEL⁺ and IBA1⁺ cells were quantified along the full length of retinal cryosection in duplicate sections per animal. For photoreceptor cell death, only TUNEL⁺ cells in the ONL were counted. IBA1⁺ cells in the outer retina (including subretinal space, ONL-RPE) were taken across the cryosection. For each section, number of TUNEL⁺ or IBA1⁺ cells was recorded (superior to inferior) in increments of 0.5 mm across the full length of retinal cryosection. Total cell counts were averaged from at least 2 sections per animal with 6 animals for each experimental group.

To quantify photoreceptor loss in retinal cryosections, the number of rows of photoreceptor nuclei was counted within the retinal lesion site (superior retina, 1 mm away from the optic nerve head). Five measurements were taken per retina in duplicate per animal.

Fluorescence in retinal cryosections was visualised and imaged using a laser-scanning A1⁺ confocal microscope (Nikon, Tokyo, Japan). Images were acquired using NIS-Elements AR software (Nikon). Negative control slides (no primary antibody) were imaged with each immunolabelled slide to determine specificity of staining and set up the threshold for fluorescence intensity for comparison to positively-stained slides. Images were processed and assembled into panels using Photoshop CS6 software (Adobe Systems, CA, USA).

Mouse retinal microglia and macrophages (CD11b⁺) were isolated using a fluorescence-activated cell sorter (BD FACSAria II; BD Biosciences, NJ, USA), using previously described protocols with minor modifications [25, 30, 31]. Mouse retinas were pooled from 2 animals (4 retinas) for each sample. Cells were stained using an anti-mouse CD11b-Phycoerythrin (PE) conjugated antibody (#101207, 1:500; Biolegend, CA, USA). FACS-isolated cells were then used immediately for RNA extraction.

RNA extraction and purification was performed on retinas and isolated CD11b⁺ cells using a combination of TRIzol reagent (Thermo Fisher Scientific) and an RNAqueous Total RNA Isolation Kit (Thermo Fisher Scientific) as described in our previous publication [32]. cDNA was prepared from 500 ng of each RNA sample using a Tetro cDNA Synthesis Kit (Bioline Reagents, London, UK) according to the manufacturer’s protocol.

Gene expression changes were measured via qPCR using Taqman hydrolysis probes (listed in Table 2) and Taqman Gene Expression Master Mix (Thermo Fisher Scientific). Each qPCR was run using a QuantStudio 12 K Flex instrument (Thermo Fisher Scientific) at the Biomolecular Resource Facility (JCSMR, ANU). Analysis was performed using the comparative cycle threshold method (ΔΔC_t) which was normalised to the expression of Gapdh and Actb reference genes, as established previously [33, 34].

Retinas were collected into Cellytic M buffer (Sigma-Aldrich) containing a Protease Inhibitor Cocktail (Sigma-Aldrich). Western blotting was performed on retinal protein lysates according to previously described methods with minor modifications [35, 36]. 20 μg of denatured protein was loaded onto a 4-20% Mini-Protean TGX Precast Protein gel (Bio-Rad, CA, USA) followed by semi-dry transfer to a nitrocellulose membrane. A list of primary antibodies used for Western blot are provided in Table 1. A secondary antibody-peroxidase conjugate was used for visualisation (Bio-Rad). The protein was visualised with chemiluminescence using a Clarity Western ECL kit (Bio-Rad) and images were captured and analysed using a Chemidoc MP with Image Lab software (Bio-Rad). The expression of the protein of interest was normalised to GAPDH.

The expression of C1qa in the murine retina was studied across the time course of photo-oxidative damage (1, 3, 5 and 7 days of photo-oxidative damage) in wild-type (WT) mice using qPCR (Fig. 1). Retinal C1qa gene expression was upregulated in the retina during the course of photo-oxidative damage, compared to dim-reared control retinas (Fig. 1a). C1qa expression was significantly elevated at 5 and 7 days (P < 0.05, Fig. 1a), but not at 1 or 3 days of photo-oxidative damage. Temporal expression of C1qa in the retina coincided with the focal loss of photoreceptors (TUNEL⁺ cells) in the ONL (Fig. 1b-f, arrows). TUNEL⁺ photoreceptor cell death reached peak numbers at 7 days of photo-oxidative damage (P < 0.05, Fig. 1a, f). Temporal expression of retinal C1qa also correlated with IBA1⁺ microglia/macrophage recruitment to the ONL and subretinal space of the outer retina, peaking at 5 days photo-oxidative damage (Fig. 1g-l), with significance detected at both 5 and 7 days for both C1qa expression and IBA1⁺ cell recruitment (P < 0.05, Fig. 1g-l). Double-labelling with CD68, a lysosomal protein, demonstrated a phagocytic phenotype of IBA1⁺ macrophages in the subretinal space after photo-oxidative damage (Fig. 1m). After 7 days of photo-oxidative damage, the IBA1⁺ macrophages were highly immunoreactive for CD68; the majority of the subretinal population contained multiple CD68⁺ cellular compartments (Fig. 1m). Double-labelling of TUNEL and IBA1 demonstrated that IBA1⁺ macrophages were in close proximity and direct contact with multiple TUNEL⁺ apoptotic cells in the subretinal space and ONL at 7 days of photo-oxidative damage (Fig. 1n).

Classical complement pathway knockout mice (C1qa^−/−) and WT were exposed to photo-oxidative damage to assess independently the functional significance of C1qa (classical pathway) on retinal degeneration. To control for any underlying effects, we thoroughly assessed C1qa^−/− mice raised in low light conditions for any histopathological or physiological changes, compared to WT (Additional file 1: Figure S1). There were no significant differences in the amplitude of a-wave, b-wave and cone responses in the C1qa^−/− mice compared to WT. There was no significant change in the expression of neurotrophic factor Cntf and the stress marker Gfap. A reduction in expression was detected in Fgf2 in C1qa^−/− retinas (P < 0.05, Additional file 1: Figure S1).

After 7 days of photo-oxidative damage similar levels of photoreceptor death, indicated by numbers of TUNEL⁺ cells, were observed in C1qa^−/− and WT retinas (P > 0.05, Fig. 2b-d). Measures of retinal morphologies including thickness of the photoreceptor layer at the focal lesion spot and towards the lesion edge were comparable (P > 0.05, Fig. 2e-f). Numbers of IBA1⁺ macrophages found in the outer retina of C1qa^−/− and WT animals were also comparable following 7 days of photo-oxidative damage (P > 0.05, Fig. 2g-i). Western blots for complement C3d protein showed an increase in retinal C3d protein following 7 days of photo-oxidative damage compared to dim-reared controls (P < 0.05, Additional file 2: Figure S2). However, we showed no difference in the abundance of C3d protein in C1qa^−/− retina and WT at 7 days of photo-oxidative damage (Fig. 2j); C3d protein expression, normalised to the loading control, did not significantly change in C1qa^−/− and WT (P > 0.05, Fig. 2k).

ERG analysis showed no differences in a-wave, b-wave, or cone responses in C1qa^−/− mice compared to WT (P > 0.05, Fig. 2l-n). Expression of complement genes in the classical pathway (C2, C4a, Serping1), however, were significantly lower in C1qa^−/− retinas compared to WT (P < 0.05, Fig. 2o). Cfh expression (alternative pathway) was significantly elevated in C1qa^−/− retinas at 7 days of photo-oxidative damage (P < 0.05, Fig. 2o), however, there was no significant change in the expression of Cfb. Using a photo-oxidative damage model, we have previously shown that these complement genes are upregulated in the retina following photo-oxidative damage compared to dim-reared controls [11].

To assess an effect of the classical pathway on photoreceptor damage beyond the standard 7-day photo-oxidative damage time course, the impact of C1qa gene ablation over an extended period after photo-oxidative damage was determined (days 8-14, Fig. 3a). Retinal expression of C1qa was significantly elevated in WT mice between day 8 (one day after the end of photo-oxidative damage) and day 14 (P < 0.05, Fig. 3b) and accompanied by elevated levels of TUNEL⁺ photoreceptor cell death (P < 0.05, Fig. 3b). Photoreceptor cell death was significantly reduced in the C1qa^−/− compared to WT at day 14 (P < 0.05, Fig. 3c, d), and the ONL was significantly thicker (P < 0.05, Fig. 3e). The ONL of the superior retina in C1qa^−/− mice was better preserved at the lesion site (0.5 mm) and at 1.0 mm away from the optic disc at day 14, indicated by the presence of significantly more rows of photoreceptors in the ONL compared to WT (P < 0.05, Fig. 3f).

The retinal macrophage response in C1qa^−/− mice was attenuated at day 14 (Fig. 3g-h). Fewer IBA1⁺ macrophages were present in the ONL of C1qa^−/− at day 14 compared to WT (Fig. 3g), and the numbers of IBA1⁺ macrophages infiltrating the ONL and subretinal space were significantly less than WT (P < 0.05, Fig. 3h). Western blots of complement C3d in C1qa^−/− and WT retina showed no difference in the abundance of complement C3d deposition in C1qa^−/− retina compared to WT at day 14 (P > 0.05, Fig. 3i-j).

ERG analyses showed clear differences in retinal function at day 14 between C1qa^−/− and WT (P < 0.05, Fig. 3k-n). The a-wave response and b-wave response amplitudes to flash intensities ≥0.6 log cd.s/m² were significantly greater in C1qa^−/− retina compared to WT (P < 0.05, Fig. 3k, l). Differences in ERG response characteristics between C1qa^−/− and WT were significant across increasing flash intensities, with the most pronounced differences observed at 1.9 log cd.s/m² (P < 0.05, Fig. 3m). There were no differences in the cone responses between the two groups (P > 0.05, Fig. 3n). Comparison of ERG responses at day 7 of photo-oxidative damage and day 14 (7 days after photo-oxidative damage) indicate improved retinal function in C1qa^−/− animals at 14 days (P < 0.05, Fig. 3o). At day 14, C1qa^−/− retina had significantly better a-wave responses (~ 100 μV) and b-wave responses (~ 200 μV) than WT at 1.9 log cd.s/m² (P < 0.05, Fig. 3o). At day 14, even at low flash intensities, the a-wave amplitudes in C1qa^−/− retinas were significantly greater than in WT retinas (Fig. 3o).

To determine whether local inhibition of C1 activation could mitigate the progression in retinal atrophy, we used an antibody against C1q (anti-C1q; ANX-M1, Annexon Biosciences), to block the classical complement cascade. Intravitreal ANX-M1, but not its immunoglobulin G (IgG) isotype control, administered immediately after the 7-day photo-oxidative damage procedure (day 7), protected the photoreceptors against continuous degeneration at day 14 (Fig. 4). Post-damage, ANX-M1 significantly reduced the number of TUNEL⁺ cells in retinas at day 14, compared to IgG control antibody (P < 0.05; Fig. 4b). Comparison of retinal thickness showed that the ONL was significantly thicker (P < 0.05; Fig. 4c) in the ANX-M1 group. Negligible effects on the number of IBA1⁺ microglia/macrophages in the outer retinal space and ONL were observed in mice treated with anti-C1q antibody, compared to IgG control (P > 0.05; Fig. 4d). ERG analyses showed that the response characteristics for rod a-wave and b-wave were significantly better in the ANX-M1 treated animals (P < 0.05; Fig. 4e, f), with the ANX-M1 cohort demonstrating significantly higher amplitudes (a- and b-wave) at the highest flash intensities compared to IgG controls (P < 0.05; Fig. 4g). Cone responses were also significantly higher in the post-treatment ANX-M1 groups (P < 0.05; Fig. 4h).

ANX-M1 administered before photo-oxidative damage showed no functional or morphological protection from damage in the retinas at day 14 compared to isotype control (Fig. 5). Retinal morphology (Fig. 5b-c), IBA1⁺ microglia/macrophage numbers (Fig. 5d) and ERGs (Fig. 5e-h) showed no significant differences between the treatment groups (P > 0.05; Fig. 5b-h), indicating that pre-damage treatment via intravitreal injection to neutralise C1q was ineffective.

The results from animals treated systemically are presented in Fig. 6. A hemolytic assay using mouse serum from ANX-M1-treated and control mice showed that by day 12, serum complement activity was reduced by > 50% compared to the IgG control group, following the systemic delivery paradigm of ANX-M1 anti-C1q antibody (P < 0.05; Fig. 6b).

The impact of systemic C1q neutralisation on retinal morphology was not significant. No differences in the number of TUNEL⁺ cells, ONL thickness or IBA1⁺ microglia/macrophage numbers were observed in mice treated systemically with the ANX-M1 anti-C1q antibody (P > 0.05; Fig. 6c-e). Similarly, there were no significant functional differences in a-wave, b-wave or cone response at day 12, between the ANX-M1 anti-C1q antibody and IgG control antibody groups (P > 0.05; Fig. 6f-i).

To further assess the potential mechanism involved in modulation of the retinal inflammation observed in C1qa^−/− mice, we investigated the role of inflammasome activation in photo-oxidative damage. C1q is known to activate NLRP3 inflammasome and production of IL-1β in mouse bone-marrow derived macrophages (BMDMs) primed with carboxyethylpyrrole adducts (CEP), which are by-products of oxidative damage in dry AMD [21]. Immunolabelling of NLRP3 demonstrates co-localisation with F4/80⁺ macrophages in the subretinal space of WT mice following photo-oxidative damage (Fig. 7a-b). However, NLRP3⁺ or F4/80⁺ cells were not detected in the ONL, or in other retinal layers. The NLRP3⁺ F4/80⁺ macrophages detected in the subretinal space of WTanimals were distributed at the focal retinal lesion and towards the lesion edges, with the rounded amoeboid morphology of activated macrophages (Fig. 7a, c). No NLRP3-expressing cells were detected in the C1qa^−/− retinas, although F4/80⁺ macrophages were present in the subretinal space (Fig. 7d). A negative control that omitted the primary antibodies showed no specific staining for NLRP3 or F4/80 (Fig. 7e).

Expression of inflammasome-related genes was assessed in C1qa^−/− and WT retinas at 7 days of photo-oxidative damage, using qPCR. Nlrp3, Casp-1 and Casp-8 expression were significantly decreased in retinas of C1qa^−/− mice compared to WT at 7 days of photo-oxidative damage (P < 0.05, Fig. 7f). Additionally, expression of the pro-inflammatory cytokine genes Il-1β and Il-18 were also significantly lower in C1qa^−/− mice compared to WT (~ 1500%, Il-1β; ~ 800%, Il-18; P < 0.05, Fig. 7f).

FACS isolation was used to analyse expression of inflammasome-related genes in retinal CD11b⁺ cells, isolated from C1qa^−/− and WT mice following photo-oxidative damage (Fig. 7g). Primary myeloid cells from C1qa^−/− retinas had a significantly reduced expression of Il-1β and Casp-1 compared to WT retinas (P < 0.05, Fig. 7g), but there was no detectable change in the expression of Il-18, Casp-8 or Nlrp3 (P > 0.05, Fig. 7g).

The presence of IL-1β protein in C1qa−/− and WT retinas was used to confirm inflammasome expression at 7 days of photo-oxidative damage (day 7; Fig. 7h-i) and post-damage (day 14; Fig. 7j-k) using Western blot. There was no significant difference in the abundance of IL-1β proteins observed, between C1qa^−/− and WT retinas at day 7 (P > 0.05, Fig. 7i). At day 14 post-photo-oxidative damage, the IL-1β protein band in C1qa^−/− retinas was significantly less intense compared to retinas from WT mice (P < 0.05, Fig. 7k).

In dim-reared mouse retinas (Fig. 8a-i), C1q protein or F4/80⁺ macrophages were not detected (Fig. 8a). In photo-oxidative damaged mouse retinas, C1q protein was co-localised with amoeboid, F4/80⁺ subretinal macrophages at the focal retinal lesion (Fig. 8b-d) and at the edges of the photoreceptor lesion (Fig. 8e-g). Negative controls showed no specific staining for either C1q or F4/80 in the mouse retinal sections (Fig. 8h). Absence of C1q immunoreactivity was confirmed in F4/80⁺ macrophages in the outer retina after 7 days of photo-oxidative damage in C1qa^−/− mice (Fig. 8i). Immunohistochemistry for the C1q protein in normal retina and AMD-affected retinas (Fig. 8j-r) showed no C1q-expressing cells in the normal retina (Fig. 8j). In the AMD-affected retinas, C1q proteins were detected adjacent to subretinal drusen (Fig. 8k), and in the subretinal space at the lesion edges (Fig. 8l-n). Double immunolabelling demonstrated that most of C1q-expressing cells were immunoreactive for macrophage/microglia marker IBA1 (Fig. 8o-q). Negative controls showed no specific staining for C1q or IBA1 in the human AMD section (Fig. 8r).

C1q is associated with pathological features of age-related neurodegenerative diseases such as amyloid-β accumulation in Alzheimer’s Disease (AD), glaucomatous damage in a model of glaucoma and drusen deposition in AMD [37‐39]. Evidence from other studies indicates that non-canonical functions of C1q influence local inflammation in the central nervous system (CNS), such as by promoting the dynamic shift of astrocytes to a more pro-inflammatory state which is associated with AD and Huntington’s Disease [40]. Moreover, C1q is involved in the elimination of synapses in a mouse model of AD, and blockade of C1q significantly preserved synaptic function [41, 42]. The ablation of C1qa further prevented the progressive loss of retinal ganglion cells in a mouse model of glaucoma [43]. However, a recent study demonstrated that numerous complement components, including C1q, were important in maintaining inner nuclear layer (INL) retinal integrity, and that the absence of C1q in the developing retina reduced retinal function in the aging retina [44]. Others have reported a neuroprotective influence of C1q on the survival of cone photoreceptors in Rho−/− mice and elucidated that C1q was involved in cone cell preservation in hereditary retinopathies [45], indicating a role of C1q in normal retinal development.

Our present findings suggest that C1q facilitates the progression of late-stage retinal atrophy, following the onset of retinal damage. Studies in our laboratory have previously shown upregulation of classical complement components C1s, C4a and C2 in retinas following photo-oxidative damage [11], but their specific roles have not been described. Other studies on the pathological effects of C1qa in neurodegeneration have been implicated in mouse models of retinal ischemic injury [24], and glaucoma [43]. The findings presented here show that initially (days 1-7 of photo-oxidative damage) the absence of C1qa has no definitive impact on retinal degeneration induced by photo-oxidative damage; however, over an extended timeframe, sustained expression of C1qa is a driver of progressive loss of photoreceptors. Rohrer et al. (2010) suggest an involvement of the activation of the classical complement in the development of choroidal neovascularisation, in the late stage of AMD in a laser-induced rodent model of choroidal neovascularisation, since the alternative complement is not sufficient to produce the pathology [15]. Consistent with this concept, the data from the current study demonstrates that localised post-treatment with ANX-M1 C1q inhibition protects against retinal degeneration during the progression phase of damage (7-14 days), but not earlier during the onset of retinal degeneration. The findings suggest a potential role of C1q in mediating the progression of retinal degeneration.

C1q stimulation activates the NLRP3 inflammasome and the secretion of IL-1β in CEP-primed BMDMs [21]. NLRP3 inflammasome activation in AMD has been proposed to occur in response to several stimuli including drusen components, RPE and complement proteins, oxidative stress, oxidative by-products and DNA [21, 46, 47]. In this study, in the absence of classical pathway activation, there was a reduced NLRP3 inflammasome expression in a specific population of cells - the subretinal microglia/macrophages.

This finding is consistent with previous studies showing that C1q can activate NLRP3 in BMDMs of humans, and in mice after immunization with CEP-adducted mouse serum albumin, to model a dry AMD-like pathology [21]. Other evidence has demonstrated the capability of RPE cells to induce NLRP3 inflammasome expression and propagate photoreceptor toxicity in the presence of blue-light photo-oxidative damage [48]. We have previously shown that in photo-oxidative damage, inflammasome expression and IL-1β production were observed in retinal degeneration, where IL-1β was found to be expressed by microglia and macrophages at the site of photoreceptor degeneration [49]. Our findings from the current study show significant down-regulation of IL-1β in isolated retinal macrophages from C1qa^−/− retinas compared to wild-type controls, which correlated to a more preserved photoreceptor population at 14 days. The findings indicate that the reduced activity in the classical pathway dampens NLRP3 inflammasome expression along with pro-inflammatory IL-1β secretion, resulting in protection of the photoreceptor population.

It is of interest that we detected no significant difference in C3 expression by Western blot between WT and C1qa^−/− mice at day 14. This may indicate that the effector functions enabled by the classical complement pathway may be independent of hydrolysis of C3 [38, 50, 51], which is predominantly amplified by the alternative complement pathway [11] in photo-oxidative damage models of retinal degeneration. Our findings indicate that one of the non-canonical mechanisms of complement activation is associated with inflammasome activation, as indicated by decreased NLRP3 labelling in C1qa^−/− microglia. Further, C1q can trigger a rapid enhancement of phagocytosis upon recognition of apoptotic cells, independent of the classical component pathway activation [52]. For these reasons, we hypothesise that both inflammasome activation and enhanced phagocytosis are mediated by C1q, explaining why we see protection of retinal function in the C1qa^−/− mice, even though no changes in C3 were detected at 14 days. These findings highlight the intricate association between the classical pathway and inflammasome activation, although the precise role of C1q in inflammasome activation and other non-canonical functions, such as phagocytosis, warrants further investigation.

Retinal microglia/macrophages express complement components in rodent retinas in response to aging [31]. Here, we demonstrate that subretinal microglia/macrophages also express C1q in focal retinal degeneration. Our data show that subretinal microglia/macrophages in the degenerating retina can express CD68⁺ lysosomal protein aggregates, and contain multiple TUNEL⁺ apoptotic cells, demonstrating a phagocytic phenotype. It is likely that C1q expression may be found in these phagocytic macrophages of the subretinal space, as we have detected C1q expression co-localising with subretinal microglia/macrophages in the degenerating retina after photo-oxidative damage. It is plausible that C1q is secreted to bind and recognize defective synapses at a stage of disease prior to overt neurodegeneration, as reported in models of neurodegenerative diseases in the brain and retina [38, 43], however this was not observed directly in this study.

In normal and diseased brain tissues resident microglia express many-fold higher levels of C1qa, −b and –c, the three protein subunits of C1q, compared to macrophages and other myeloid cells [53, 54]. The capacity of microglia to synthesize C1qa has been confirmed in rodent brain [55], however, in the context of AMD the close interdependence and proximity of the retina with Bruch’s membrane/RPE/choroid confounds the argument that microglia are the key synthesizers of C1q. Our findings demonstrate that C1q proteins are localised near drusen and in the subretinal microglia/macrophages of human AMD patients whose RPE atrophy was prominent. The delayed appearance of these C1q-expressing subretinal microglia/macrophages, coincides with RPE and choriocapillaris breakdown [56] suggesting a population derived from the choriocapillaris. However, a recent study investigating models of RPE breakdown suggests that subretinal populations are predominantly resident microglia, owing to their rapid mobilisation from the inner retina to the site of injury [57].

Our findings with the local delivery of the C1q neutralisation suggest that locally-derived C1q contributes to the progression of retinal atrophy. Coupled with the association of C1q expressed by myeloid cells with neuroinflammation and age-related neurodegeneration [58], our findings support the idea that locally administered immunotherapy targeting the classical complement pathway may be effective in slowing the progression of retinal degeneration. We did not detect any ocular inflammation specifically associated with anti-C1q delivery in these experiments. This suggests that administration of an anti-C1q antibody is unlikely to have additional risks to those normally associated with intravitreal injections (reviewed in [59]). Further investigations are required to explore the mechanisms underlying the protective effect of local C1q neutralisation.