A local complement response by RPE causes early-stage macular degeneration

Abstract

Inherited and age-related macular degenerations (AMDs) are important causes of vision loss. An early hallmark of these disorders is the formation of sub-retinal pigment epithelium (RPE) basal deposits. A role for the complement system in MDs was suggested by genetic association studies, but direct functional connections between alterations in the complement system and the pathogenesis of MD remain to be defined. We used primary RPE cells from a mouse model of inherited MD due to a p.R345W mutation in EGF-containing fibulin-like extracellular matrix protein 1 (EFEMP1) to investigate the role of the RPE in early MD pathogenesis. Efemp1^R345W RPE cells recapitulate the basal deposit formation observed in vivo by producing sub-RPE deposits in vitro. The deposits share features with basal deposits, and their formation was mediated by EFEMP1^R345W or complement component 3a (C3a), but not by complement component 5a (C5a). Increased activation of complement appears to occur in response to an abnormal extracellular matrix (ECM), generated by the mutant EFEMP1^R345W protein and reduced ECM turnover due to inhibition of matrix metalloproteinase 2 by EFEMP1^R345W and C3a. Increased production of C3a also stimulated the release of cytokines such as interleukin (IL)-6 and IL-1B, which appear to have a role in deposit formation, albeit downstream of C3a. These studies provide the first direct indication that complement components produced locally by the RPE are involved in the formation of basal deposits. Furthermore, these results suggest that C3a generated by RPE is a potential therapeutic target for the treatment of EFEMP1-associated MD as well as AMD.

Introduction

Age-related macular degeneration (AMD) causes irreversible vision loss in millions of people (1). AMD has a complex etiology, but MD can also be inherited in a Mendelian fashion and affect individuals in early adulthood (2). Inherited MD and AMD share drusen as an early indicator of disease (3–6). Drusen are extracellular deposits of protein and lipid that accumulate between the retinal pigment epithelium (RPE) and Bruch's membrane (BrM) (7–9).

Although the sources of drusen components and the mechanisms of drusen formation are not fully understood, many components of drusen have been identified, among them, complement components such as C3, complement component H (CFH) and C5 (3,10,11). Based in part on these findings, it has been suggested that the formation of drusen is an inflammatory process (3,7,12–15). Consistent with this idea, genetic variants in complement genes are considered to be one of the main risk factors for AMD, but how alterations in complement components contribute to the MD pathology remain to be defined (16–18). In the last decade, many potential therapies for MDs have been designed to target complement components (19). However, it remains unclear if regulation of the complement system locally or systemically is needed for the most effective treatment of MD, or at what stage of disease these treatments can be most usefully applied (13).

Some authors consider confluent basal laminar deposits as the histological definition of AMD (20). Basal laminar deposits are microscopic lesions associated with RPE/choroid dysfunction that also contain activated complement proteins (6,8,20–26). RPE cells are critical for the visual cycle and retinal homeostasis (27). Briefly, RPE cells reduce oxidative stress by the absorption of light, recycle outer segments by phagocytosis, supply nutrients to the photoreceptors and form the outer component of the blood retinal barrier (27–29). RPE cells also secrete most of the extracellular matrix (ECM) components in BrM including matrix metalloproteinases (MMPs) responsible for the turnover of this ECM (30–34). Alteration of the ECM turnover in BrM has been associated with complement activation in late stages of AMD (6,27,35).

The inherited form of MD clinically most similar to AMD is the EGF-containing fibulin-like ECM protein 1 (EFEMP1)-associated MD, caused by the dominant p.R345W mutation in the EFEMP1 (4,36). EFEMP1 belongs to the fibulin family of ECM proteins and is thought to be involved in elastogenesis (37). As for the potential role of EFEMP1 in AMD, it (i) accumulates beneath the RPE overlaying drusen and basal deposits (38–40), (ii) binds the tissue inhibitor of metalloproteinases 3 (TIMP-3) (40), which is associated with another inherited MD and with AMD (41,42), and (iii) binds CFH, with higher affinity for the AMD-associated variant (39). Therefore, understanding how the p.R345W mutation in EFEMP1 leads to the RPE pathology in EFEMP1-associated MD will help elucidate the early mechanisms responsible for RPE pathology in AMD.

As described previously, Efemp1^R345W/R345W mice (henceforth denoted Efemp1^R345W) develop extensive basal deposits only in the presence of an active complement system (43–45). To investigate the role of RPE, the complement system and ECM in the early aspects of MD pathogenesis, we developed a culture model using primary RPE cells from Efemp1^R345W mice that recapitulate the formation of sub-RPE deposits in vitro in 2 weeks. Studies performed using this model system show that the secreted mutant EFEMP1 forms an abnormal ECM that stimulates a local inflammatory process, mediated primarily by C3a and altered ECM turnover, causing the formation of basal deposits. This is not only a valuable model to study the early mechanisms responsible for deposit formation in AMD, but also a unique system to discover new targets and test drugs that might inhibit the formation of deposits.

Results

RPE cells from Efemp1^R345W mice recapitulate deposit formation in vitro

Efemp1^R345W mice develop basal deposits between the RPE and BrM and have been characterized as a model to study deposit formation in MD (43–45). To investigate the molecular events associated with the RPE pathology in early MD, we developed a primary RPE culture model from 2-month-old Efemp1^R345W mice. Culture procedures were optimized in order to reproduce in vitro the properties of RPE cells in vivo. By 72 h, RPE cells were present as a confluent monolayer of hexagonal bi-nucleated cells (Fig. 1A). Polarization of the cells was demonstrated by (i) transepithelial electrical resistance (TER) that reaches over 200 Ω cm² by 72 h and increases to nearly 300 Ω cm² at 1 week (Fig. 1B), (ii) expression of the tight junction-specific protein zonula occludens 1 (ZO-1) at 1 week (Fig. 1C) and (iii) transmission electron microscopy (TEM) that showed the presence of apical microvilli, basal infoldings, pigment granules (Figs 1D and 2A–C) and tight junctions between adjacent cells (Fig. 2C). Other cell types such as fibroblasts in the RPE cell cultures were not identified by morphological examination or immunostaining of ZO-1 (Supplementary Material, Fig. S1). Additionally, RT-PCR for the markers Von Willebrand factor (Vwf), cluster of differentiation 68 (Cd68), cluster of differentiation 163 (Cd163) and rhodopsin (Rho) ruled out the presence of endothelial cells, macrophages and photoreceptors.

Figure 1.

Confluent primary RPE cell cultures grown on transwells. (A) Light microscopy of confluent RPE cell culture after 1 week. (B) TER (Ω cm²) as a function of time in culture (n = 100, mean ± SEM). (C) Staining of ZO-1 after 1 week in culture. (D) TEM of RPE cells on transwells. RPE, retinal pigment epithelium; TEM, transmission electron microscopy.

Figure 2.

RPE cells from Efemp1^R345W mice recapitulate deposit formation in vitro. TEM (A–F) and SEM (G–I) of polarized RPE monolayers grown on transwells for 2 weeks show apical microvilli, basal infoldings and tight junctions (white arrows) (A–C). (B) A cellular projection growing through a pore of the insert (black arrow; notice that the pore appears wider than those shown in flatmounts due to the sectioning process). Thick deposits (white arrow) spread across the bottom side of the insert in KI cultures (E), whereas very thin cellular projections are barely visible in WT (D) and KIC3KO (F). SEM from bottom flatmounts shows thin cellular projections (arrows) in WT (G and J) and KIC3KO (I and L), whereas pores (white arrowheads) can be seen between cellular material (J and L). KI cultures present overlapping layers of deposits (H), which completely cover the pores of the insert (K). (J–L) are higher magnifications from black squares in (G–I), respectively. Scale bar A–F, 5 µm; G–L, 10 µm. KI, Efemp1^R345W; KIC3KO, Efemp1^R345W:C3^−/−. Thickness of the insert = 10 µm.

Of note, as observed for other RPE culture systems, RPE cells cultured on transwells developed cell projections that passed through the pores to the bottom surface of the transwells (46–49) (Fig. 2D–F). Using scanning electron microscopy (SEM), we observed that the cell projections generated by wild-type (WT) cells were a mixture of thin, filamentous projections and flat processes which covered only a portion of the bottom surface of the transwell (Fig. 2G and J). Efemp1^R345W cells showed the same cell projections as WT cells at 1 week (Supplementary Material, Fig. S1). However, at 2 weeks, the bottom surface of transwells with Efemp1^R345W RPE cells was completely covered by thicker more amorphous deposits (Fig. 2H and K). The deposits did not exhibit the shape or structure of cell projections seen in WT cultures, but was spread across the whole transwell surface forming overlapping layers of disorganized, clumped material. Since this was thicker than the material produced by WT cells and it covered more of the insert, it was also noticeable by TEM on transverse sections (Fig. 2E). The RPE cells from Efemp1^R345W mice thus recapitulate in vitro the formation of basal deposits that was observed in vivo via the production of the deposits (43–45).

The role of complement in the formation of deposits was tested in cultures of primary RPE cells from homozygous double-mutant Efemp1^R345W:C3^−/− mice. Interestingly, Efemp1^R345W:C3^−/− RPE cells did not develop deposits on the bottom side of the insert, but showed normal cellular projections similar to WT (Fig. 2I and L). This suggests that the formation of deposits by Efemp1^R345W RPE cells is complement-dependent, as was observed for the formation of basal laminar deposits in vivo in Efemp1^R345W mice (45).

Sub-RPE deposits share features with basal laminar deposits

Based on the TEM images, the apical side of the cultured RPE cells appeared normal, and thus, the projections and the deposits formed by Efemp1^R345W RPE cells were likely extensions derived from the basal sides of the cells. Although structurally different, sub-RPE deposits share features with basal laminar deposits in humans and mice (25,45). The bottom side of the transwells was mounted as flatmounts for immunostaining to analyze the composition of the deposits. Staining with Hoechst dye demonstrated that the projections were not migrated cells, as no nuclei were detected on the bottom surfaces of the transwells. The deposits contained increased amounts of EFEMP1, collagen I, elastin, TIMP-3, fibrillin-1 and fibronectin relative to the projections of the WT RPE cells (Fig. 3A). Interestingly, these are major constituents of BrM and basal laminar deposits (25). Increased levels of ECM proteins found in the deposits were not associated with an increased production of their mRNAs (Supplementary Material, Fig. S2). These results suggest that sub-RPE deposits mimic deposits in Efemp1^R345W mice, and is primarily comprised of normal ECM proteins that are present in abnormal amounts (45). The increased levels of EFEMP1 protein observed in the deposits were not associated with increased production or secretion of the mutant EFEMP1^R345W by the Efemp1^R345W cells compared with WT cells (Fig. 3B–D).

Figure 3.

EFEMP1^R345W expression and secretion. (A) EFEMP1, TIMP-3, elastin (ELN), Col1a1, fibronectin (FN) and FBN-1 were detected in the deposits by immunostaining of the bottom side of the inserts mounted as flat mounts. (B) mRNA levels of Efemp1 measured by qRT-PCR (n = 12 WT, 11 KI and 7 KIC3KO). (C) Western blot (apical media) and (D) ELISA (apical and basal media) of WT, KI and KIC3KO cultures at 2 weeks (n = 4 per genotype) showed that WT and mutant EFEMP1 are secreted to the same degree. Scale bars, 50 µm. Data represented as mean ± SD.

Complement is locally produced by RPE cells

The finding that an intact complement system is required for the formation of deposits by Efemp1^R345W RPE cells in culture suggests that the RPE cells produce complement themselves. To evaluate this possibility, we measured the expression of complement by the RPE cells in vitro. We detected mRNAs for complement components belonging to the three complement pathways in the RPE cells grown in the absence of fetal bovine serum (Supplementary Material, Fig. S2). Levels of the mRNAs for these complement components were similar in WT and Efemp1^R345W RPE cells with the exception of C3, which was 2-fold higher in the Efemp1^R345W cells [t-test, P = 0.0299, n = 18 WT and 13 knock-in (KI); Fig. 4A]. Increased C3 was also demonstrated by ELISA, with C3 detected in media being 3.8-fold higher in Efemp1^R345W cultures compared with WT cultures (t-test, P = 0.0387, n = 5 per genotype; Fig. 4B).

Figure 4.

C3 expression and secretion. (A) mRNA levels of complement components in WT and KI RPE cultured for 2 weeks showed that C3 was increased in KI. (B) ELISA demonstrated that C3 was increased in KI media. (C) Immunostained flatmounts showed that C3 was more intense in deposits of KI cells; C3a was only detected in KI cultures, whereas CFH, MAC and C3d were visible in both WT and KI cells. Scale bars, 50 µm. *P < 0.05, t-test. Data represented as mean ± SD.

Consistent with the increased production of C3 by the Efemp1^R345W RPE cells, increased levels of C3 were detected in the deposits produced by these cells (Fig. 4C). Of note, the activation product of C3, C3a, was detected in deposits but not in normal cellular material produced by WT cells (Fig. 4C). In contrast, CFH, C3d and membrane attack complex (MAC) were detected on the bottom surface in both WT and Efemp1^R345W cultures (Fig. 4C). CFH and C3 appeared to be co-localized with EFEMP1 in deposits (Supplementary Material, Fig. S2). Taken together, these data support the idea that complement components produced by the RPE are also activated locally.

Secreted EFEMP1^R345W and C3a act in a non-cell autonomous fashion to stimulate the production of deposits

Having established that RPE cells from Efemp1^R345W mice produce deposits in culture, we next used this in vitro system to investigate the mechanisms by which the mutant EFEMP1 protein causes deposit formation. We hypothesized that factors secreted by the Efemp1^R345W cells rather than intracellular stress were responsible for deposit formation. To test this hypothesis, we performed a media exchange experiment, in which WT RPE cells were grown in conditioned media (CM) from Efemp1^R345W cells, and vice versa for 2 weeks. As shown in Figure 5A, WT cells grown in CM from Efemp1^R345W cells developed deposits. On the other hand, Efemp1^R345W cells grown in CM from WT cells, or for which media were changed every day, showed a substantial inhibition of deposit formation (t-test, P = 0.0337 and P = 0.0130, respectively; Fig. 5A).

Figure 5.

EFEMP1^R345W and C3a and the formation of sub-RPE deposits. Percentage of deposits (percentage of SEM images positive for deposits divided by the total number of images taken of the whole insert) observed in (A) WT RPE cells grown in CM from KI, (B) WT cells grown in fractions of CM from KI (>100, 100–30, 30–10 and 10–3 kDa), (C and D) WT and KI cultures treated, respectively, with different doses of rmEFEMP1^R345W (C) and EFEMP1-neutralizing antibody (D). (E and F) WT and KI cultures treated, respectively, with different doses of rmC3a (E) and C3a-neutralizing antibody (F). (C–F) Best-fit values curves show the percentage of deposits versus concentration of recombinant protein (C and E) or neutralizing antibody (D and F). All data represented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001 compared with no-treated controls by the t-test. KI, Efemp1^R345W; IC₅₀, half maximal inhibitory concentration; EC₅₀, half maximal effective concentration.

These results demonstrate that factors secreted by the Efemp1^R345W cells are necessary and sufficient to stimulate deposit formation by WT cells. To help identify these potential candidate factors, we fractionated the CM from Efemp1^R345W cells by molecular weight. WT RPE cells grown in either the 3–10 or 30–100 kDa fractions of CM from Efemp1^R345W cells made deposits after 2 weeks, whereas no deposits were observed in WT cells cultured in the 10–30 or >100 kDa fractions (Fig. 5B). Based on molecular weight, the best candidate proteins that could be responsible for deposit formation were EFEMP1 (55 kDa) and C3a or C5a (9 kDa each). We therefore tested the ability of these proteins to stimulate formation of deposits by WT RPE cells.

Cultured WT RPE cells treated with recombinant mouse EFEMP1^R345W protein (rmEFEMP1^R345W; 10, 50 and 100 ng/ml) for 2 weeks formed deposits in a dose-dependent manner (Fig. 5C), with an EC₅₀ of 11.4 ng/ml. Conversely, the formation of deposits was inhibited in a dose-dependent manner in cultures of Efemp1^R345W RPE cells treated with EFEMP1-neutralizing antibodies (0.01, 0.1, 1 and 10 µg/ml) for 2 weeks. At high concentrations of the antibody, the formation of deposits was completely inhibited (Fig. 5D).

Similarly, cultured WT RPE cells treated with recombinant mouse C3a (rmC3a; 5, 10, 50 and 100 ng/ml) formed deposits in a dose-dependent manner (Fig. 5E) with an EC₅₀ of 54.8 ng/ml. Furthermore, the formation of deposits was inhibited in a dose-dependent manner in cultures of Efemp1^R345W RPE cells treated with C3a-neutralizing antibodies (0.0001, 0.001, 0.01, 0.1 and 1 µg/ml) for 2 weeks. At high concentrations of the antibody, mutant cells did not show any deposits (Fig. 5F).

In contrast, cultured WT RPE cells treated with recombinant mouse C5a (5, 10, 50 and 100 ng/ml) for 2 weeks did not form deposits. These data demonstrate that secreted EFEMP1^R345W acts in a non-cell autonomous fashion, likely mediated in part by C3a but not by C5a, to stimulate deposit formation.

EFEMP1^R345W protein increases extracellular aggregates in Efemp1^R345W RPE cultures

We next asked how the secreted mutant EFEMP1^R345W protein stimulates deposit formation. EFEMP1 levels were increased in both BrM of Efemp1^R345W mice and the deposits produced by Efemp1^R345W RPE cells, and thus, one possibility is that incorporation of the mutant protein into the ECM results in an abnormal ECM (45). One metric of ECM status is the level of extracellular protein aggregates; we therefore stained the bottom flatmounts from WT and Efemp1^R345W RPE cells with the Proteostat protein aggregation dye (Enzo Life Sciences, Farmingdale, NY, USA), designed to detect aggregation of a broad range of proteins (50,51). As shown in Figure 6, increased levels of protein aggregates were detected in the deposits produced by Efemp1^R345W cells (t-test, P = 0.0357). The EFEMP1 signal co-localized with the aggregated protein in the deposits of Efemp1^R345W RPE cells, which was also observed in WT cells (Fig. 6A and B). This is consistent with a recent study, showing that some aggregation is normal for EFEMP1 (37).

Figure 6.

Formation of aggregates in EFEMP1^R345W RPE cell cultures. (A) Aggregates stained with Proteostat (Enzo) co-localize with the EFEMP1^R345W protein in the deposits. Confocal images, scale bar, 50 µm. (B) Quantification of fluorescent signal obtained from Proteostat and EFEMP1 staining using the Image J software. Data represented as mean ± SD. **P < 0.01. KI, Efemp1^R345W; KIC3KO, Efemp1^R345W:C3^−/−.

Decreased MMP-2 activity is associated with the formation of deposits

BrM is an ECM that is regulated in part by MMP/TIMP-3 activity (30–32). Since EFEMP1 interacts with TIMP-3 (40), it is possible that in addition to promoting protein aggregation, EFEMP1^R345W could also contribute to abnormal ECM structure and deposit formation by reducing ECM turnover. To determine if EFEMP1^R345W alters ECM turnover, we measured MMP activity in Efemp1^R345W cultures. As shown in Figure 7A, the MMP-2 activity was decreased in media from Efemp1^R345W or double-mutant Efemp1^R345W:C3^−/− RPE cells compared with WT (ANOVA, P = 0.047 and P = 0.0062). This appears to be due to an effect of the EFEMP1^R345W protein on MMP-2 activity, as Mmp-2 mRNA levels were not significantly different from WT cells (Fig. 7B).

Figure 7.

Efemp1^R345W cultures: ECM turnover and interleukin secretion. (Aa) MMP-2 activity decreased in apical and basal media from KI and KIC3KO RPE cultures at 2 weeks (n = 5 per genotype). (Ab) Representative zymography gel showing MMP-2 activity in apical (a) and basal (b) media of WT, KI and KIC3KO RPE cultures. (B) mRNA levels of MMP-2 in WT, KI and KIC3KO RPE cells at 2 weeks (n = 10 WT, 11 KI and 4 KIC3KO). (C) MMP-2 activity decreased by rmEFEMP1^R345W in media from WT and by EFEMP1-neutralizing antibody in media from KI after 2 weeks. (D) MMP-2 activity decreased by rmC3a in WT media but not by C3a-neutralizing antibody in KI cultures after 2 weeks. (F and H) IL-6 (pg/ml) and (E and G) IL-1B measured in apical and basal media of WT and KI cultures treated with rmEFEMP1^R345W, anti-EFEMP1 (E and G), rmC3a or anti-C3a (F and H). All data represented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by ANOVA. KI, Efemp1^R345W ; KIC3KO, Efemp1^R345W:C3^−/−; rmC3a, recombinant mouse C3a; rmEFEMP1^R345W, recombinant mouse EFEMP1^R345W; anti-C3a, neutralizing antibody for C3a; anti-EFEMP1, neutralizing antibody for EFEMP1.

To determine if the decrease in MMP-2 activity observed in the Efemp1^R345W cells was due to the EFEMP1^R345W protein itself, we treated WT RPE cultures with increasing concentrations of rmEFEMP1^R345W, which resulted in a dose-dependent inhibition of MMP-2 activity (Fig. 7C). Conversely, Efemp1^R345W cultures showed a rescue of MMP-2 activity after being treated with EFEMP1-neutralizing antibody (Fig. 7C). These results also suggest that EFEMP1^R345W can directly modulate the activity of MMPs.

Since C3a can stimulate the production of deposits by WT RPE cells, we also evaluated the effect of C3a on MMP activity. As observed for EFEMP1^R345W protein, rmC3a also decreased MMP-2 activity in the media of WT cells in a dose-dependent manner (Fig. 7D). However, neutralization of C3a in cultures of Efemp1^R345W cells did not increase the activity of MMP-2 (Fig. 7D).

These data suggest that decreased MMP-2 activity and the associated alteration in ECM turnover are mechanisms by which deposits are formed. To determine if decreased ECM turnover can result in the formation of deposits, we treated WT RPE cells with ARP100, a specific inhibitor of MMP-2 (52). After 2 weeks the cells did not show any deposits, although MMP-2 activity was effectively inhibited (Supplementary Material, Fig. S3), and thus decreased MMP-2 activity alone is not sufficient to induce deposit formation by WT RPE cells.

Interleukins contribute to the process of deposit formation

Since C3a can stimulate the production of interleukin (IL)-1B and IL-6, cytokines that are likely involved in the pathogenesis of RPE in MD, we evaluated the potential roles of these interleukins in deposit formation by Efemp1^R345W RPE cells (53–55). IL-1B did not show a significant change in WT culture media after the addition of rmC3a; however, IL-1B decreased after neutralization of C3a in Efemp1^R345W media (Fig. 7E). IL-6 levels increased following the addition of rmC3a in WT cultures and significantly decreased in Efemp1^R345W media after the addition of C3a-neutralizing antibody (Fig. 7F). In contrast, no change in IL-1B or IL-6 concentration was observed after the addition of rmEFEMP1^R345W in WT cultures or neutralization of EFEMP1^R345W in Efemp1^R345W cultures (Fig. 7G and H).

To determine if interleukins could also contribute to decreasing the activity of MMPs, we tested the MMP-2 activity in WT media after the addition of either IL-1B or IL-6. The MMP-2 activity did not show a significant change after the addition of rmIL-1B or rmIL-6 to WT RPE cultures; however, MMP-2 activity increased after the addition of antibodies for IL-1B or IL-6 in Efemp1^R345W (Supplementary Material, Fig. S3). This result indicates that interleukins could also contribute to deposit formation, at least in part by inhibiting MMP-2 activity in Efemp1^R345W cultures. Interleukin expression was shown to be C3a-dependent and not EFEMP1-dependent; thus, interleukins may contribute to the inflammatory process once complement is active.

Working model for basal deposit pathogenesis

Based on the data described above, we have generated a working model of basal deposit formation in EFEMP1-associated MD (Fig. 8). The novel aspects of the model include the central roles of C3a and defective ECM turnover in deposit pathogenesis.

Figure 8.

Working model for deposit pathogenesis. EFEMP1^R345W has a dual role in the RPE pathogenesis: (i) EFEMP1^R345W forms aggregates that tend to bind other ECM proteins, generating an abnormal ECM structure, and (ii) EFEMP1^R345W inhibits MMP-2 activity and alters the turnover of the ECM, also resulting in abnormal ECM structure. The abnormal ECM structure exposes neoepitopes that induce increased local activation of complement, resulting in the release of C3a and subsequent formation of deposits. C3a also stimulates IL-1B and IL-6 release, which further inhibits MMP-2 activity. This mechanism results in a chronic inflammatory situation that is exacerbated by the presence of interleukins and reduced MMP-2 activity. ECM, extracellular matrix; RPE, retinal pigment epithelium; MMP, matrix metalloproteinase.

Discussion

A role for the complement system in AMD was suggested by genetic association studies in 2005, but how alterations in the complement system contribute to this disease remain to be defined, in part due to the lack of good models (16–18,56). The studies reported here, facilitated by the use of RPE cells from a genetic model of MD, provide several new insights into the pathogenesis of MD. First, the data demonstrate that a local complement response by the RPE is responsible for the earliest lesions observed in MDs, basal deposits. Second, this complement response is atypical since it is mediated by C3a, but not by C5a or the MAC. Third, the increased activation of complement in this system occurs in response to abnormal ECM, generated by extracellular aggregates containing EFEMP1^R345W and reduced ECM turnover due to inhibition of MMP-2. Fourth, increased production of C3a also stimulates the release of cytokines such as IL-6 and IL-1B, which appear to have a role in deposit formation, albeit downstream of C3a. These results demonstrate the central role of the complement system in MD, and have important implications regarding the development of therapies to treat early MD.

A number of investigators have suggested that drusen results from a chronic inflammatory response due to RPE or choriocapillaris injury, but the specific mechanisms that connect RPE pathology, complement activation and drusen formation are unknown (3,7,12–15,22,23,57). Use of Efemp1^R345W RPE cells made it possible to investigate the specific role of the RPE in early MD pathogenesis. The Efemp1^R345W RPE culture system recapitulates the formation of sub-RPE deposits in vitro in only 2 weeks, compared with months in Efemp1^R345W and other gene-targeted mice (43,56). Furthermore, the deposits generated by the Efemp1^R345W RPE cells share components with basal laminar deposits and drusen (25,38,45,58). So far, this is the most accurate model in reproducing in vitro the in vivo features of basal deposits. Formation of wide-spaced collagen or APOE-rich deposits had been reported for ARPE-19 and human RPE cells, however only after challenge with retinal homogenates or human serum (59,60). The Efemp1^R345W RPE cells provide the first model of genetically induced sub-RPE deposits in vitro, and demonstrate a new approach to study the mechanisms responsible for the RPE pathology in inherited and age-related macular degenerations.

In the pathogenesis of MDs, RPE cells are thought to modulate the activation of immune cells by producing complement components and cytokines (13,60–62). Our results show that RPE cells go beyond this and act as immune cells, creating a local immune response in the absence of other cell types. RPE cells from the Efemp1^R345W:C3^−/− mice do not produce deposits in vitro, confirming the dependence of deposit formation on an intact complement system (45). Our studies provide the first direct indication that complement components produced locally by the RPE are involved in the formation of basal deposits.

The local immune response by the RPE is atypical since is not based on MAC or C5, but rather a direct action of C3a. Of note, C3 was the only complement component found to be increased in the Efemp1^R345W RPE cultures, and C3a the only complement activation product detected exclusively in deposits. We are investigating which of the three complement pathways (classical, lectin or alternative) is involved with the initiation of complement activation. One possibility is that the initial generation of C3b can be explained by tick-over (63), where the hydrolysis of C3 occurs at physiological level and forms the C3bBbP convertase (64). According to this theory, the C3bBbP complex from the alternative pathway can be deposited on ECM in a stable manner, regardless the complement pathway that was primarily activated (65,66). Alternatively, it is possible that abnormal matrix generated by EFEMP1^R345W would favor the deposition of C3bBbP in a stable manner (64). In either case, the C3BbP convertase will generate more C3b creating a positive feedback loop that will result in the increased release of the anaphylatoxin C3a. C3a has been detected in basal deposits and drusen, and reported to alter proteasome function in RPE and to promote choroidal neovascularization through VEGF release in late AMD (7,12,67–70). However, C3a has not been previously associated with early stages of MD pathogenesis. Thus, we propose a new role of C3a as an inflammatory molecule that is also capable of stimulating the production of abnormal ECM by RPE cells (71,72). This new role of C3a could explain the lack of responses of patients with AMD to treatments targeting C5 (19,73). Furthermore, these results suggest that therapies targeted at C3a in RPE could play an important role in treating AMD.

The primary defect in the EFEMP1-associated MD is the production of mutant EFEMP1^R345W protein (36,38,43–45). In contrast to prior studies using cell lines that overexpressed EFEMP1^R345W, we found that the mutant protein was efficiently produced and secreted by Efemp1^R345W RPE cells and acts in a non-cell autonomous fashion to stimulate deposit formation (38,74–76). We did observe that the secreted EFEMP1^R345W protein forms extracellular protein aggregates that create an abnormal ECM, consistent with the accumulation of the mutant protein in BrM of Efemp1^R345W mice and patients (38,45). The protein aggregates may also contribute to making the ECM abnormal via binding other ECM components, and impairing ECM homeostasis (50,58,77,78).

Abnormal ECM may also be a shared feature between inherited and age-related MDs. A study suggested that altered ECM composition in BrM causes impaired RPE function and contributes to the development of AMD (79). The pathology in EFEMP1-associated MD is caused by a single mutation versus the combination of genetic predisposition and life-long environmental influences that are thought to occur in AMD. Despite these differences in etiology, increased complement activation is a common feature in both diseases (45). Thus, EFEMP1-associated MD represents a useful model for studies of disease pathogenesis, and can help to elucidate the mechanisms involved in the inflammatory process of AMD.

We found that both EFEMP1^R345W and C3a decrease the activity of the MMP-2, a metalloproteinase required for normal turnover of the ECM in BrM. Loss of MMP-2 favors the accumulation of ECM proteins and is thought to contribute to the development of sub-RPE deposits in vitro and in vivo (59,80). Furthermore, alterations in MMP-2 activity can contribute to perpetuate the inflammatory process (35).

We also found that C3a participates in regulating the expression of the interleukins IL-1B and IL-6 by RPE cells. The pro-inflammatory cytokines are also thought to affect RPE function and to play a role in AMD pathogenesis (71,72). Inhibition of IL-1B and IL-6 repressed the formation of deposits in vitro, but recombinant IL-1B or IL-6 was not sufficient to cause deposit formation without the action of C3a. Thus, interleukins may contribute to the inflammatory process once complement system activation is increased by other stimuli.

In summary, studies using the novel Efemp1^R345W RPE cell model system suggest that basal deposit formation is a local inflammatory process, which occurs in response to abnormal ECM (78,81–83). The novel finding that C3a locally produced by RPE cells plays a key role in early MD pathogenesis implies that complement-modulating therapies could be applied locally and addressed specifically against C3a in patients with early macular disease. This could avoid disruption of the complement system systemically, and potentially prevent progression of disease to vision-threatening sequelae such as geographic atrophy or choroidal neovascularization.

Materials and Methods

A detailed description of Materials and Methods can be found in Supplementary Material.

Mice

The guidelines of the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research were followed. Efemp1^R345W/R345W mice, made and characterized previously, were generated from het × het crossings (43). WT littermates generated during het × het crosses of Efemp1^+/R345W mice were used as controls. Mice were also generated from homozygous crossings Efemp1^R345W/R345W × C3^−/− mice, generated and characterized previously (43,45,84,85). None of these mice contained the rd8 mutation (85).

Mouse RPE cell isolation

RPE cells were harvested from 10-week-old mice following CO₂-induced euthanasia. Eyes were dissected and RPE cells were collected by trypsin digestion. A very detailed protocol is available in Supplementary Material.

RPE cell cultures

Cells were cultured at 37°C in 5% CO₂ under a humidified atmosphere changing media twice a week. Serum was removed at least 72 h before performing any experiment. TER of the RPE cultures were measured using an epithelial voltohmmeter (57,86).

Culture treatments

CM from Efemp1^R345W to WT

WT and Efemp1^R345W RPE cells were isolated and cultured in parallel in duplicate. Apical and basal supernatants were collected from Efemp1^R345W cultures every 24 h and added into the counterpart (apical and basal) WT cultures, while fresh media was added daily to the mutant cultures. Serum was depleted in all the cultures for the last 72 h. Supernatants collected every 24 h were kept and frozen at −80°C for further experiments. After 2 weeks in culture, cells were fixed and coated for SEM.

Fractionated CM from Efemp1^R345W to WT

WT and Efemp1^R345W RPE cells were isolated and cultured in parallel. After the first 72 h, apical and basal supernatants from Efemp1^R345W cultures were collected every 24 h and fractionated using Amicon filters (3, 10, 30 and 100 kDa; Millipore, Billerica, MA, USA) under sterile conditions. Volumes were brought to 140 µl for the apical chamber and 600 µl for the basal chamber of the transwells with media. WT RPE cells were cultured in 3–10, 10–30, 30–100 or >100 kDa fractions from Efemp1^R345W CM. Small molecules (<3 kDa) were discarded. Each fraction was added daily to the counterpart WT culture during 2 weeks per duplicate with the last 72 h in the absence of serum.

Treatment with recombinant proteins and neutralizing antibodies

Primary RPE cell cultures were treated with different doses of recombinant proteins or neutralizing antibodies every 48 or 72 h, respectively, for 2 weeks. For treatments with rmC3a or neutralizing antibody for C3a, cells were serum starved after the first 72 h and grown without serum for another 11 days. Recombinant proteins, neutralizing antibodies and inhibitors are listed in Supplementary Material.

Recombinant protein FLAG-EFEMP1-R345W

cDNA from Efemp1^R345W RPE cells was subcloned into Gateway compatible pCAG-FLAG-GFP (expression vector containing FLAG) and transfected into CHO cells using Lipofectamine 2000 (Life Technologies, Grand Island, NY, USA). Recombinant protein was purified from supernatants after 72 h using the FLAG-M purification kit (Sigma Aldrich, St Louis, MO, USA).

RNA and protein extraction

After 2 weeks in culture, RPE cells were lysed and RNA and protein were extracted with the DNA/RNA/Protein Mini Kit (Qiagen, Venlo, Netherlands). The quality and quantity of RNA was assessed using the Agilent RNA 6000 Nano Kit Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). All samples had RNA integrity number values between 9 and 10. Supernatants from the upper and bottom chamber were separately removed, and proteins were precipitated with 100% (w/v) TCA and resuspended in a buffer containing 50 mm Tris–HCl pH 8, 100 mm DTT, 10 mm EDTA, 2% SDS and 1 mm PMSF.

Expression of complement components by RPE cells

qRT-PCR was used to analyze mRNA levels. PCR conditions and specific primers can be found in Supplementary Material.

Western blot analysis

Approximately 1 µg of protein was electrophoresed on a 4–12% polyacrylamide gel (Life Technologies) and then transferred to a PVDF membrane using the iBlot System (Life Technologies). Immunoblotting was performed according to the Odyssey System (Li-Cor, Lincoln, NE, USA), using rabbit anti-EFEMP1 antibody at 0.4 µg/ml and incubated overnight at 4°C (SDIX, Newark, DE, USA). Odyssey secondary antibody anti-rabbit was incubated 1 h at room temperature (RT).

Enzyme-linked immunosorbent assay

Apical and basal supernatants from RPE cell cultures were collected after 2 weeks in culture and concentrated to equal volumes through 10 and 3 kDa Amicon filters (Millipore). The fraction over 10 kDa was used to quantify IL-1B, IL-6, EFEMP1, SC5b-9 and C3. The fraction between 3 and 10 kDa was used to quantify C3a and C5a. ELISA kits used to quantify IL-1B, IL-6 and C5a were purchased from R&D Systems (Minneapolis, MN, USA). ELISA kits used to quantify EFEMP1, SC5b-9 and C3a were purchased from MyBioSource (San Diego, CA, USA). C3 was quantified by the ALPCO ELISA kit (Salem, NH, USA).

Characterization of deposits

Deposits were characterized by TEM, SEM and immunofluorescence (87).

1. TEM: Samples were fixed in ½ Karnovsky's fixative for 12 h at 4°C. After fixation, samples were rinsed with 0.1 M sodium cacodylate buffer, post-fixed with 2% osmium tetroxide in 0.1 M sodium cacodylate buffer for 1.5 h, en bloc stained with 2% aqueous uranyl acetate for 30 min, then dehydrated with graded ethyl alcohol solutions and resin infiltrated in ethyl alcohol and Spurr's epoxy resin mixtures (Electron Microscopy Sciences, Hatfield, PA, USA) utilizing an automated EMS Lynx 1 EM tissue processor (Electron Microscopy Sciences). Processed samples were infiltrated in two changes of fresh Spurr's resin for 24 h and then oriented in Spurr's resin, and polymerized within silicone molds using an oven set for 60°C for 24 h minimum. The grids were imaged using an FEI Tecnai G2 Spirit transmission electron microscope (FEI, Hillsboro, OR, USA).

2. SEM: RPE cells grown on inserts were fixed in cold 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) followed by fixation in 1% glutaraldehyde, washed in PBS and then in dH₂O, and dehydrated by serial ethanols, 35, 50, 70, 95, 95 and 100% followed by critical dehydration using the SAMDRI-795 system. After dehydration, inserts were split into two pieces and the top or bottom side was coated with chromium using a Gatan Ion Beam Coater for 10 min. Coated inserts were imaged by an Field Emission Scanning Electron Microscope (JEOL 7401F).

   The percentage of deposit formation was defined as the percentage of SEM images positive for deposits divided by the total number of images taken of the whole insert (# images per equal parts insert). A minimum of 10 images was taken per sample. The experimenter was masked to the sample identity.

3. Immunostaining: Inserts were rinsed in PBS, fixed for 10 min in cold 4% PFA in PBS followed by fixation in 1% glutaraldehyde for 30 min at RT. The inserts were cutoff with a razor blade and stored in PBS at 4°C pending sectioning and immunohistochemical analyses. Inserts were placed face down on slides as flatmounts of the bottom side. For staining, sections were blocked with 1% BSA and incubated with primary antibodies (listed in Supplementary Material) overnight at 4°C. Secondary antibodies labeled with Alexa-488 or Alexa-555 (Life Technologies) were incubated for 1 h at RT. Slides were mounted with Fluoromount G (Electron Microscopy Sciences) and visualized by a TCS SP5 II confocal laser scanning microscope (Leica). RPE bottom flatmounts stained in parallel with the samples but incubated with 1% BSA in the absence of primary antibody were used as negative controls.

4. Protein aggregation assay: Proteostat protein aggregation dye was used the following manufacturer's instructions (Enzo Life Sciences).

5. Quantification of the fluorescent signal: Images were converted to binary format with ImageJ (87). The integrated intensity was measured.

MMP-2 activity

Was measured in media from RPE cultures by zymography? Briefly, 5 µl of equal volume supernatants were loaded onto Novex 10% gelatin gels (Life Technologies). Zymography assays were then performed as per the manufacturer's instructions. Gels were scanned using the Odyssey system (Li-Cor). MMP-2 was identified by molecular weight. Gelatinase activity was quantified using densitometry and the software ImageStudioLite (Li-Cor).

Statistical analyses

Results are expressed as mean ± SD, with P < 0.05 considered statistically significant. Differences between groups were compared using Student's t-test or ANOVA as appropriate.

To calculate the EC₅₀ and IC₅₀ doses in RPE treated with recombinant proteins or antibodies, respectively, we used the GraphPad Prism software. Non-linear regression (curve fit) was performed using the log (concentration) versus the percentage of deposits calculated as described above.

Supplementary Material

Supplementary Material is available at HMG online.

Funding

This work was supported by the Ocular Genomics Institute. The TEM work was supported in part by a National Eye Institute Core grant (P30EY003790).