Age‐related macular degeneration: A disease of extracellular complement amplification

Age‐related macular degeneration (AMD) is a major cause of vision impairment in the Western World, and with the aging world population, its incidence is increasing. As of today, for the majority of patients, no treatment exists. Multiple genetic and biochemical studies have shown a strong association with components in the complement system and AMD, and evidence suggests a major role of remodeling of the extracellular matrix underlying the outer blood/retinal barrier. As part of the innate immune system, the complement cascade acts as a first‐line defense against pathogens, and upon activation, its amplification loop ensures a strong, rapid, and sustained response. Excessive activation, however, can lead to host tissue damage and cause complement‐associated diseases like AMD. AMD patients present with aberrant activation of the alternative pathway, especially in ocular tissues but also on a systemic level. Here, we review the latest findings of complement activation in AMD, and we will discuss in vivo observations made in human tissue, cellular models, the potential synergy of different AMD‐associated pathways, and conclude on current clinical trials and the future outlook.

are pigmentary changes of the RPE, and the accumulation of extracellular debris, termed drusen, 5 which locate between the basal membrane of the RPE and the inner collagenous layer of BrM. 6 Latestage AMD is characterized either by geographic atrophy, defined by progressive and irreversible vision loss due to degeneration of the RPE and photoreceptors, or by neovascular AMD, defined by the invasion of newly formed blood vessels into the retina, which tend to leak and thereby cause rapid vision loss. Neovascular AMD can be treated with inhibitors targeting vascular endothelial growth factors (VEGF), but for the more common geographic atrophy no treatment exists as of today. 5 AMD is a multifactorial disease, and various risk factors in addition to age have been identified: environmental factors like smoking and diet, but also genetic variation influences the disease risk substantially. 7,8 In the largest genome-wide association analysis conducted so far, an association of 52 individual variants was identified. 7 F I G U R E 1 Retinal anatomy. The retina is a complex structure consisting from posterior to anterior of the following layers: outer nuclear layer (ONL), outer plexiform layer (OPL), inner nuclear layer (INL), inner plexiform layer (IPL), and ganglion cell layer (GCL). In the healthy retina (top left panel) the RPE and BrM form the outer blood-retinal barrier. Accumulation of extracellular debris below the RPE leads to drusen formation (top right panel), and during AMD progression these drusen tend to expand and coalesce leading to geographic atrophy (bottom left panel) or trigger the invasion of newly formed blood vessels called neovascular AMD (bottom right panel). Drusen formation is accompanied by pigmentary changes of the RPE, and macrophage accumulation on the choroidal side. In the late stages, microglia migrate to the outer retina.
These variants cluster in four main pathways: ECM remodeling, lipid metabolism, angiogenesis, and the complement system. The two strongest loci associated with AMD are located in the age-related maculopathy susceptibility/HtrA Serine peptidase 1 (ARMS2/HTRA1) locus on chromosome 10, and in the complement factor H (CFH) locus on chromosome 1. 7 In addition to the CFH locus, variants in the complement genes complement factor I (CFI), complement factor B (CFB), complement component 3 (C3), complement component 9 (C9), and vitronectin (VTN) have been identified. 7 Overall, common risk variants in or near complement genes account for more than 50% of the genetic risk in AMD. 8 The complement system is an ancient part of the innate immune system, classically known as a first-line defense system against pathogens. The complement system can be activated by various epitopes via the classical and lectin pathways ( Figure 2); in addition, constant low-level activation of the alternative pathway, or amplification loop, due to spontaneous tick-over of C3 to C3(H 2 O) primes the complement system for rapid activation. 9 Factor B (FB) binds to C3(H 2 O) and is then cleaved by Factor D (FD), forming the alternative pathway C3 convertase C3(H 2 O)Bb, which further cleaves C3 molecules into the anaphylatoxin C3a and C3b. C3b, again, is bound by FB, generating more C3 convertases and thereby feeding into the amplification loop ( Figure 2).
Additional binding of C3b to C3bBb generates the complement component 5 (C5) convertase C3bBbC3b, which cleaves C5 into C5b, releasing the anaphylatoxin C5a. C5b initiates the sequential binding of the complement components C6, C7, C8, and multiple C9 molecules to form the membrane attack complex (MAC). 9 Upon successful assembly on a target cell, the MAC disrupts the membrane thereby potentially causing lysis of the target cell. 10 The release of the anaphylatoxins C3a and C5a attract circulating polymorphonuclear cells to the site of complement activation and stimulate them to release pro-inflammatory cytokines and stimulate mast cells to degranulate. C3b itself is a strong opsonin, and the deposition of multiple C3b molecules on a surface essentially labels that surface for destruction by phagocytosis by the F I G U R E 2 Schematic overview of the complement system. The complement system can be activated via three different pathways: The classical pathway, the lectin pathway, and the alternative pathway. All three merge at the central component C3, and thereby feed into the self-amplification loop, which enables rapid generation of a strong inflammatory response upon activation. The terminal pathways include formation of the C5b-9 complex, also called membrane attack complex (MAC), which upon successful assembly can disrupt target membranes (left panel). The inactivation prevents aberrant complement activation by irreversibly degrading C3b (right panel). macrophage that join the other massed immune cells: This leads to tissue remodeling including ECM turnover. 11 The ability of rapid activation, however, makes tight regulation of complement activation crucial to prevent host tissue damage. One central regulator is Factor H (FH), which inhibits complement activation at multiple points in the activation cascade: It competes with FB binding to C3b, has decay-accelerating activity for the C3bBb complex, and acts as a co-factor for Factor I (FI) mediated degradation of C3b. 12 As the major soluble FI co-factor, FH plays a significant role in the regulation of complement activation in the fluid phase, but importantly is the only co-factor that mediates complement regulation on acellular surfaces, such as a hosts' ECM 12 : a function that plays a central role in diseases such as AMD in which ECM remodeling is arguably a driver of disease.
In addition to FH, alternative splicing of the complement factor H (CFH) gene results in generation of the smaller transcript Factor H like 1 (FHL-1), which shares many functions with full-length FH. 13 Next to FH and FHL-1, the CFH locus also contains the complement factor H related (CFHR) genes 1-5 encoding for the Factor H related (FHR) proteins 1-5. These proteins show a high homology with FH, which includes the domains required for ligand and cell surface recognition. But, strikingly, the FHR proteins lack the regulatory domains needed for co-factor function, and while the exact FHR protein functions still need to be uncovered, binding of FH ligands without regulatory activity leads to competition with FH and consequential de-regulation of complement activation. 14 Given the delicate balance between activation and inactivation of the complement system, it is not surprising that alterations disrupting this balance have been associated with AMD, but also with other diseases like atypical hemolytic uremic syndrome, systemic lupus erythematosus, and rheumatoid arthritis. 15 In this review, we will describe the role of the alternative pathway in AMD (Section 2), discuss the most recent reports of cellular models with emphasis on the CFH locus and the interaction with the RPE (Section 3), and evaluate recent developments in clinical trials (Section 4).

| THEALTERNATIVEPATHWAYINAMD
The genetic association of the complement system with AMD was first established for the common non-synonymous variant p.Tyr402His in CFH in multiple studies. [16][17][18][19] This association was later confirmed and expanded with the complement components CFI, C3, CFB, and C9 and VTN. 7 Interestingly, seven rare variants in complement genes were individually associated with AMD risk, of these were four non-synonymous variants (CFH p.Arg1210Cys, CFI p.Gly119Arg, C9 p.Pro167Ser, and C3 p.Lys155Gln), and two non-coding variants and one synonymous within the CFH locus (rs148553336 upstream of CFH, rs35292876 in exon 17 in CFH, and rs191281603 in intron 3 of the complement factor H related 5 [CFHR5] gene). 7 For carriers of the variants CFI p.Gly119Arg, C9 p.Pro167Ser, and C3 p.Lys155Gln, a younger age at disease onset has been reported. 20 In addition to the individual associations of rare and common variants in complement genes, a burden of rare variants in CFI and CFH was identified. 7 Identifying individual associations of rare variants with disease is often limited by statistical power, which makes functional analysis of individual variants necessary to classify rare variants. In the following, we describe complement levels in AMD patients, also in context of genetic variation.
For in depth reviews of the genetic association of the complement system in AMD, we refer to recent reviews. 21,22 In light of the strong genetic predisposition in genes encoding complement components, it is not surprising that levels of individual complement components and complement activation markers were extensively studied in serum and plasma samples of AMD patients.
Also, to a lesser extent, measurements in local tissue such as aqueous and vitreous humor have been reported.  [24][25][26][27] Within these fragments, Ba and C3d are markers of chronic complement activation, and when stratified for AMD stage, the CNV subgroup has the lowest C3d levels when compared to GA and early AMD. 24 Comparable findings were reported in a subsequent study with a larger study population. 28 Indeed, studies investigating the contribution of genetic risk alleles to late-stage AMD progression have found that patients suffering from GA mostly carry risk on their chromosome 1 risk loci (involving the CFH/CFHR genes), 29 and patients carrying risk mostly on their chromosome 10 risk allele progressed faster to late-stage disease and predominantly ended up with CNV. 30 When comparing the different AMD stages, the highest systemic complement activation levels, determined by the C3d/C3 ratio, were observed in patients with intermediate AMD or geographic atrophy. 28 Interestingly, C3a levels were significantly elevated in patients with neovascular AMD that did show only a partial response to anti-VEGF therapy compared to patients that showed a response. 31 Strikingly, in line with this observation, local C3a concentrations were significantly elevated in patients after anti-VEGF treatment that relapsed within 100 days compared to those that did not show signs of recurrence within the same time. 32 Interestingly, recently an association between the poor response to anti-VEGF treatment and the CFH p.Tyr402His variant has been reported. 33 These findings indicate that complement activation levels and the genotype in complement genes are also relevant in context of anti-VEGF therapy, and dual treatment of complement inhibition and anti-VEGF treatment might be suitable in some patients with neovascular AMD. Within the complement system, the CFH locus shows the strongest association with AMD risk, 7 consequently FH and FHR levels especially in context of the common variant leading to the p.Tyr402His change (odds ratio 1.47) have been extensively studied in AMD patients.
Initially, decreased FH levels in serum of AMD patients were reported, 27,34 and in carriers of the protective, intronic variant CFH rs6677604 increased FH levels in plasma of carriers was observed. 35 The AMD protective variant rs10922109 is in high linkage disequilibrium, meaning inherited together and incredibly difficult to separate, with this variant (D′ = 0.989 in the European, non-Finnish population), and in line with the observations made by Ansari et al, the rs10922109 is also associated with increased FH levels in serum. 36 However, if stratifying based on AMD disease status no differences in systemic FH levels between cases and controls could be identified. 36,37 It should be noted, however, that while the majority of AMD patients show systemic FH levels in the normal range, carriers of rare coding variants in CFH do present with significantly reduced FH levels in plasma. 38 In line with the observation that for most AMD patients FH levels are within the normal range, carriers of the common variant CFH p.Tyr402His present with normal FH levels, 25,36,37,39 but with a significantly increased C3d/C3 ratio. 25,40 For those patients carrying risk purely on the chromosome 10 ARMS2/HTRA1 locus, circulating FH levels were not analyzed, but given the high prevalence of this locus in AMD patients (minor allele frequency of 44% in AMD patients 7 ), it seems very unlikely that this variant would affect FH levels given the lack of observable change in unstratified AMD patients as a whole. Interestingly, carriers of a rare coding variant in CFH presented with elevated levels of the C3 convertase C3bBbP, 38 which is likely due to impaired co-factor function for FI mediated C3b degradation that has been reported for multiple rare variants in CFH. 41,42 Interestingly, while for individuals without rare variants in CFH C3bBbP levels increased with disease stage, rare variant carriers presented with elevated C3bBbP levels independent of disease stage, 38 indicating that rare variant carriers present with higher complement activation prior to disease onset, while individuals without rare variants acquire elevated complement activation throughout the course of AMD progression. This observation might be relevant in context of treatment schemes, but this needs to be further evaluated with prospective studies.
It should be noted, however, that the genetic association in the CFH locus also includes variants in the CFHR genes. The exact function of FHR proteins in immune homeostasis and their role in AMD pathogenesis still needs to be investigated, but recent findings indicate that their relevance for AMD development has been neglected. Deficiency of FHR-1 and FHR-3 or of FHR-1 and FHR-4 caused by a common and a rare deletion, respectively, are protective for AMD, [43][44][45] and elevated FHR-4 level was shown to be associated with increased AMD risk. 39 In line with those findings, reduced serum concentrations of total FHR-1, FHR-2, FHR-3, and FHR-4A and a higher serum concentration of FH were observed in carriers of the protective variant rs10922109, an intronic variant in the CFH gene. 36 It was shown that FHR-4 competes with FH, and FHL-1, binds to immobilized C3b, 14,39 and stabilizes the C3 convertase, thereby inhibiting C3b cleavage and promoting complement activation. 46 An earlier study showed when added in a much higher and non-physiological concentration (400 μg/mL), FHR-3 and FHR-4 showed a weak activity in promoting the interaction between FH and FI in C3b cleavage. However, FHR-4 alone does not have any FI co-factor activity, even when present at 650 μg/ml. 47 Interestingly, when it is applied in 86.6 μg/mL, FHR-4 is able to enhance the cofactor activity of FH but had no detectable influence on the cofactor activity of FHL-1. 46 Considering FHL-1 is the predominant complement regulator in BrM, 48 it is reasonable to assume that systemically elevated FHR-4 is able to bind C3b but without acting as a co-factor for FI, potentially permitting downstream complement activation in BrM.
For FHR-1 and FHR-5, binding to ECM components has been shown, and this binding resulted in increased deposition of C3 fragments on ECM components, but only treatment with FHR-5 increased FB binding to ECM components. 49 These findings suggest not just local complement activation, but also indicate distinct capacities of the different FHR proteins to fine-tune complement activation on the ECM. Of special interest is the observation that FHR-1 and FHR-5 compete with FH binding to ECM components in a dosedependent manner. 49 In addition to BrM and the choriocapillaris, the proteins FHR-2, FHR-4, and FHR-5 have also been identified in drusen of AMD patients. 36,39 Given that AMD patients present with elevated levels of the FHR proteins systemically, 36 39,43,50,51 suggesting that the primary source of FHR proteins is the systemic circulation. The basal side of BrM and the ECM surrounding the choriocapillaris, however, are in direct contact with the systemic circulation, 52 and consequently are particularly susceptible to complement turnover and activation. Any changes to the systemic concentrations of complement proteins, including increased FHR levels in blood, will have a direct effect within this space, particularly if these proteins are known to interact with ECM components. 49 Next to the CFH locus, also genetic variation in the central regulator CFI has been identified. Two variants, the common intronic variant rs10033900 and the rare coding variant p.Gly119Arg (rs141853578), show individual associations with AMD. In addition, a burden of rare variants in CFI has been associated with AMD. 7 In line with the genetic association, low systemic levels of FI increase the risk of advanced AMD, 53,54 and the rare variant p.Gly119Arg in CFI, which has an odds ratio of 5.12, 7 affects FI secretion, 41,53,55-58 thereby leading to impaired degradation of C3b in serum of carriers. 41,55,58 Interestingly, FI plasma levels between AMD patients and controls do not differ significantly, 59 unless individuals are stratified for rare variants. 53,56,57,60 Reduced systemic FI levels have been reported in carriers of various rare variants, such as CFI p.Ala240Gly 53,54,56 and CFI p.Leu131Arg, 41 and while not all rare variants were measured in serum or plasma samples of carriers, analyses utilizing recombinant protein expression indicate that more than 50% of rare coding variants in CFI identified in AMD patients affect FI secretion. 57 Not surprisingly, FI levels correlate with the observed degradation of C3b in patient serum, 41 and, for example, carriers of the variants CFI p.Leu131Arg and CFI p.Gly119Arg with decreased systemic FI levels showed impaired ability for C3b degradation. 41,55 The relevance of systemic FI levels is further underlined by the observations that lower systemic FI levels are associated with younger age at onset and advanced AMD, 53,54,56 and carriers of the CFI p.Gly119Arg variant are more likely to progress into end-stage dry AMD compared to noncarriers. 20 In addition, even though effects of rare coding variants in CFI have been studied systemically only, the concentration of intraocular FI is correlated with systemic FI concentrations, 60 suggesting that rare variants affecting FI systemically are likely to affect ocular FI concentrations in a similar fashion.
Next to an impairment of the central regulators FI and FH, also gain-of-function variants in activating components of the complement system have been reported. In the central component C3, both rare and common variants have been identified in AMD patients, and the strongest association is observed for the common variant C3 p.Arg102Gly, which has an odds ratio of 1.47. 7 Carriers of this variant present with elevated C3d/C3 plasma levels, 61 and increased activity in the hemolytic assay has been reported. 62,63 The second signal in C3 is the rare variant C3 p.Lys155Gln, which has an odds ratio of 3.1, 7 and carriers present with normal plasma levels. 41 For C3 p.155Gln, reduced binding to FH and reduced degradation has been reported in the co-factor assay with FI and FH 64 ; however, these findings were not confirmed in a later study, even though increased hemolytic activity was observed on sheep erythrocytes. 63 In the terminal component C9, the rare variant p.Pro167Ser has been associated with AMD risk in the genome-wide analysis. This variant has an odds ratio of 1.8, 7 and conflicting results have been reported for this variant. Increased C9 levels in serum 65  (ARPE-19) with WT C9 and the p.Pro167Ser variant, however, no differences were detected. 65 In this context, it should be noted that RPE do not secrete C9, but show high concentrations of the C9 inhibitor CD59. 50,51 However, as discussed in the following, multiple studies showed susceptibility of the choroid to C5b-9 deposition.
Even though the C9 p.Pro167Ser variant has not been studied in the context of the choroid, it seems likely that this variants' effects are located on the choroidal side rather than in the retina.
Interestingly, while for systemic sC5b-9 levels both normal and elevated levels have been identified in AMD patients, 26,27,60 indicating a certain level of heterogeneity between patients and cohorts, accumulation of local C5b-9 is substantially increased in the choroid of AMD patients, 67 and is found most strongly on the RPE at the F I G U R E 3 Complement regulation in the retina. In the healthy retina presence of fluid phase and surface-bound regulators like FI, FH, and CD59 prevent aberrant complement activation (left panel). Impairment of these regulators or excess presence of the de-regulating FHR proteins negatively affect this regulatory balance, and excess deposition of C5b-9, C3, and C4 leads to local tissue damage (right panel). Cell types are indicated as in Figure 1, with addition of mast cells indicated in orange. *For simplicity, complement components identified in drusen are not depicted in this figure.
edge of GA lesions. 68 Although complement activation within the retinal space is a common feature of late-stage AMD it is likely to be a consequence of the ongoing disease, while the complement activation in the underlying ECM of the choriocapillaris has been shown to occur in a genetic-risk mediated fashion decades before any disease manifests itself. 69 It seems reasonable, therefore, that if one was to target complement overactivation in AMD, any therapeutic will need to address the runaway amplification loop on both sides of the outer blood/retinal barrier. 70 While in early AMD the C5b-9 deposition is associated exclusively with the choriocapillaris, the deposition only extends to the RPE on advanced stages of dry AMD. 71 These findings strongly suggest that choroidal endothelial cells are affected prior to the RPE in AMD pathogenesis. Furthermore, C5b-9 depositions are more pronounced in the choroid of individuals carrying the CFH risk allele p.Tyr402His, 71,72 indicating that overall impaired complement activation leads indeed to local accumulation of C5b-9 complexes. Unfortunately, C5b-9 deposition in the choroid has not been stratified for other genetic variants, but it seems likely that, for example, the protective effects of reduced systemic C9 levels 73,74 are due to reduced formation of the C5b-9 complex in the choroid.
For the variant C9 p.Pro167Ser, it remains to be clarified whether patients show alterations in the choroid.
Next to increased C5b-9 deposition, also elevated C-reactive protein (CRP) deposition was detected in the choroid and drusen of AMD patients. 75 76 Additionally, it was shown that the CFH p.Tyr402His polymorphism altered the binding of the truncated form of FH, FHL-1, to both CRP and its closely related long pentraxin PTX3. 78 These findings indicate there might be an overall pro-inflammatory environment in choroid of CFH p.402His genotype carriers.
FD is the enzyme that cleaves FB in the C3bB complex into C3bBb (C3 convertase) in the alternative pathway, and thus contributes to the amplification loop and alternative pathway activation. 79 It was implied that lutein supplementation is associated with decreased systemic FD level and downstream activation products such as C3d, C5a, and sC5b-9, 80,81 even though the sample size of included subjects was relatively small. Additional evidence suggests that lutein supplement slows down AMD progression. In fact, varying results on systemic FD levels have been reported. Some suggest that FD plasma levels are decreased in AMD patients, 82 others did not observe a difference between AMD patients and controls, 26 while some even identified elevated FD levels. 24,83,84 The intronic SNP CFD rs3826945 shows a weak association with AMD (odds ratio = 1.11, P = 0.0032), and no altered FD level was reported for carriers of this variant. 84 Complement inhibitors such as decay-accelerating factor (DAF, CD55), MAC-inhibitory protein (CD59), and membrane co-factor protein (MCP, CD46) are present on host cell surfaces to protect them from aberrant complement activation. 85 It has been shown that CD46 and CD59 expression is decreased on CD14-positive monocytes from patients with neovascular AMD. 86 The link between cell surface complement inhibitors and AMD pathogenesis needs to be further investigated, but it is known that FH, and its al- This missing clarity is due in part to the lack of in vivo models that replicate both the full-array of human complement proteins and the unique architecture of the human eye. Commonly used animal models such as mice lack even a macula region in their retina, and in addition present with an alternatively arranged CFH locus: murine FH does not result in the alternatively spliced FHL-1, and instead of five FHR proteins, only two FHR proteins, termed FHR-B and FHR-C, have been identified in murine plasma. 92 Although FHR-B has been shown to bind to human C3b and competes with FH binding comparable to its human orthologues, 93  and CD11B. 96 Interestingly, while RPE in vivo do express the complement regulators CFH, CD46, CD59, and small amounts of CFI, no expression of C3, C5, and CFB was observed by adult RPE in vivo. 51,97 More studies are needed to explore this phenomenon, but it is likely that the RPE modify the expression and secretion of complement was significantly decreased, while simultaneously cytotoxicity and lipid peroxidation were significantly increased. Interestingly, addition of exogeneous FH did not provide protection from these effects. Furthermore, mitochondrial function, based on oxygen consumption rates, was impaired potentially explaining why exogeneous FH did not rescue the observed effects. 100 In contrast, supplementation of recombinant FH to ARPE-19 cells resulted in significantly reduced cell death when adding FH p.402Tyr but not when adding FH p.402His. The protective effect of FH p.402Tyr was further confirmed in iPSC-derived RPE cells. 101 However, direct comparison of the two studies is difficult, because not only different cell lines, but also different sources and concentrations of FH protein were used. In one study, commercial serum-purified FH, likely containing a mix of FH p.402Tyr and FH p.402His, was applied at 1 μg/mL, 100 and in the other purified recombinant FH was applied at 300 nmol/L (46.5 μg/mL). 101 Indicating that the protective effect of FH is poten- showing FHR-3 internalization, and no report on the effect of lower concentrations is given. 103 In the only study in which physiological FH concentrations were applied, 100 no rescue effect of exogenous FH supplementation on hTERT-RPE1 cells deficient in FH was identified. Consequently, the reported observations with exogenously applied proteins should be interpreted with caution and subsequent validation under the appropriate conditions is important before drawing specific conclusions.
While the cellular internalization of FH and FHR-3 needs to be validated using physiological concentrations, FH binding to necrotic cells is well established. 106 In addition, a mini-FH protein containing only CCP6-8 showed binding to a choroidal cell line expressing the ECM glycosaminoglycans (GAGs) heparan sulfate and chondroitin sulfate. Strikingly, the FH p.402His showed 40% reduced binding compared to FH p.402Tyr, 107 showing that differential binding of mini-FH p.402His and p.402Tyr to human retina and choroid can be modeled in vitro. 87,89 This is perhaps more physiologically relevant to FHL-1, which contains the single GAG-binding site in CCP7 where the p.Tyr402His occurs, opposed to the full-length FH protein whose second GAG-binding site (in CCP19-20) contributes to the avidity of GAG binding and often compensates for observed differential GAG binding when testing the CCP7 domain alone. 89 More recently, the binding of FHR-1 to necrotic cells, including the cell line ARPE-19, but not to live cells, has been reported. Conversely, FHR-5 under the same experimental conditions bound to both live and necrotic cells. 108 Given the strong genetic association of the variant p.Tyr402His in FH, it is not surprising that iPSC-derived RPE carrying this variant have been included in multiple studies. High-risk lines are reported to have significantly increased expression of CFI and CFH, but decreased expression of C3, although these differences did not translate through to alterations in protein secretion. 109 Conversely, reduced FI expression levels in iPSC-derived RPE from AMD patients have been reported. 110 Given the natural variation of FI and FH levels on a systemic level (described in Section 2), these differences are likely due to inter-individual variation and should be considered when interpreting differences in expression and secretion between cell lines from different donors. The utilization of CRISPR editing, however, has made isogenic controls more accessible, likely generating more robust gene expression datasets in context of high and low genetic risk for AMD. This is a powerful tool for assessing the effects of specific genetic variants, but stringent quality control experiments are required to ensure no off-target alterations are also introduced through the CRISPR editing, which is a known phenomenon of this gene editing technique. 111 Complementing the data generated by Armento et al., who used ARPE-19 cells, 100 Hallam et al. also observed mitochondrial stress in iPSC-derived RPE cell lines carrying the high-risk variant p.Tyr402His in CFH based on mitochondria being reduced in quantity, but showing increased volume. 109 Impairment of mitochondrial function in both primary RPE and iPSC-derived RPE derived from AMD patients has been confirmed in multiple studies, 110,112,113 but interestingly in a more recent study mitochondrial impairment was linked to the CFH risk genotype at p.Tyr402His independent of the donor's phenotype. 114 In the same study, significant differences in expression and secretion levels of complement components could not be observed due to variability between the cell lines. Overall, comparable to primary cells, high variability between different iPSC lines is observed, and emphasizes the need for either larger sample sizes or isogenic controls. Nevertheless, it should be pointed out that also in vivo AMD patients show a high heterogeneity also in regard to complement concentrations.
Recently, Sharma et al. showed that iPSC-derived RPE carrying the CFH p.402His risk allele showed significantly increased accumulation of lipid deposits, and significantly decreased transepithelial resistance. 115 Additionally, these effects were amplified by treatment with complement competent serum. Furthermore, apical expression of C3aR1 and C5aR1, and downstream activation was shown.
Interestingly, inhibition of C3aR1 and C5aR1 with compstatin and PMX53 significantly decreased lipid deposition formation and increased the transepithelial resistance, indicating that the observed damaging effect of complement competent serum is mediated via C3aR1 and C5aR1 signaling. 115 It should be noted, however, that compstatin is a C3 inhibitor, 116 so not all the observed effects might be due to altered C3aR1 signaling. Interestingly, iPSC-derived RPE cells carrying the CFH p.402His risk allele also showed significantly impaired phagocytosis of photoreceptor outer segments (POS). 115 Conversely, another study utilizing a genetic risk score based on 13 AMD-associated variants, including FH p.Tyr402His, identified no difference in transepithelial resistance nor phagocytic activity. 117 Interestingly, C3 and C4 deposition on POS distal to lesions of AMD patients has been reported, supporting the hypothesis of complement-mediated modification of POS phagocytosis. These findings were confirmed in vitro incubating POS with human serum showing POS susceptibility to C3, C4, and polymerized C9 deposition. Interestingly, the C9 inhibitor CD59, but not the inhibitors CD46 and CD55 were present on POS. 118 Co-cultures consisting of the ARPE-19 cell line and porcine retinas showed C3 deposition on porcine POS after multiple days in culture, this is not observed in porcine retina only, suggesting that the source of C3 are the ARPE-19 cells. In contrast, C4 staining is predominantly located on the basal side of the ARPE-19. 119 Overall, multiple lines of evidence support the specific deposition of C3 and C4 on POS, but the molecular consequences remain to be evaluated.
A co-culture approach, using porcine retina and hTERT-RPE1 cells, has also been used to study the effects of endogenous FH in RPE cells. Silencing of the CFH gene with siRNA led to reduced retinal thickness compared to FH competent controls, and this effect was not rescued by the exogeneous addition of either FH or C3. 120 Overall the co-culture system of porcine retina and human RPE cell lines presents an accessible and cost-effective approach to recapitulate molecular mechanisms in a multicellular system. Disadvantages, however, are the risk of introducing artefacts due to the inter-species co-culture especially in the context of the complement system, in which slight imbalances and various triggers could artificially alter its activation. In addition, the limited culture time in vitro is a major drawback when studying an age-related disease like AMD, where perhaps a longer culturing time would allow for more physiological comparisons.
As discussed in Section 2 above, the choroid contains high levels of FH and C5b-9 complexes. Extending these data, the choroidal endothelial cells have been shown to highly express FH and FHL-1. 121 In fact, in retinas of three donors, CFH expression in the choroid was substantially higher than in RPE. Using both an immortalized choroidal cell line and an iPSC-derived choroidal cell line, the authors showed that reduced CFH expression resulted in elevated C5b-9 deposition on these cells. 121  and CFB in retinal organoids largely corresponding with the expression pattern observed in human retina. 95 Also, successful co-cultures of retinal organoids with RPE have been reported. 124 Taken together, while in the context of the complement system in AMD only limited results have been reported using these models, this will likely change in the near future enabling deeper understanding of AMD pathology.
In summary, in vitro analysis has identified the RPE cell monolayer as a complement-regulatory unit by responding to exogeneous complement stimuli, and also highlighted the role of the CFH p.Tyr402His in mitochondrial dysfunction in the RPE of carriers. The rapid progression in the generation of improved co-culture models and generation of "organ-on-a-chip" models will probably soon extend these findings under more physiological conditions including the use of multicellular systems.
It is important to always remember though that various cell culture conditions and differentiation protocols are used for the vast majority of these different studies, and as such potentially affect the results and possibly explains discrepancies reported between studies. Also, only very limited genotyping has been performed on the different cell lines, and given the strong genetic association in AMD, it seems likely that lack of extensive genotyping will negatively affect the reproducibility of results. Whatever the methodology applied in a study, it would be pertinent to validate one's results, preferably using well-characterized, physiologically relevant, human donor tissues.

| INTER AC TI ONOFTHECOMPLEMENT SYS TEMWITHD IFFERENTPATHWAYS INVOLVEDINAMD
We discussed the need for multicellular structures for the holistic modeling of AMD in vitro. Adding a new level of complexity, the complement system itself is an intricate system posed to react to a broad spectrum of triggers. 9 On top of that, AMD is a multifactorial disease with not just the complement system contributing to disease risk, but also genetic variation in the pathways of extracellular matrix remodeling, lipid metabolism, and angiogenesis has been identified. 7 Consequently, it needs to be addressed whether these pathways all contribute independently or instead in an interactive way to AMD pathogenesis. Recent findings support the latter.
Metabolite analysis in a large genotyped cohort of AMD patients and controls identified significantly decreased levels of verylow-density lipoprotein (VLDL), and significantly increased levels of high-density lipoprotein (HDL) in AMD patients. Interestingly, of 60 metabolites showing an association with AMD, 57 also associated with complement activation levels determined by the C3d/C3 ratio.
Among these were a negative correlation of VLDL with complement activation, and a positive correlation of HDL with complement activation. 125 In addition, binding of FH to HDL particles mediated by apolipoprotein E has been reported. 126 In this context, it should be mentioned that the common variant c.388T > C (p.Cys130Arg) is protective in the context of AMD (odds ratio 0.7), 7 and is associated with reduced ApoE plasma levels. 127 In summary, it might be that increased binding of FH due to elevated HDL levels reduces freely available FH to, for example, protect BrM from aberrant complement activation. However, this is speculative and more insight of the interaction of these components is needed in the context of AMD.
Surprisingly, the common variant c.184G > A (p.Val62Ile) in CFH is associated with decreased serum levels of matrix metalloproteinase-8 (MMP-8), while the p.Tyr402His change is associated with increased MMP-8 levels. 128 MMP-8 belongs to the family of metalloproteinases and is a collagen degrading protease released by neutrophils during inflammatory reactions. 129 Interestingly, when treating neutrophils with activated serum from carriers of the protective variant p.Val62Ile, they showed significantly lower MMP-8 secretion compared to carriers of the major allele p.Val62. 128 Even though to date no direct link between AMD and MMP-8 has been reported, the RPE and choroid do express MMP-8 at low levels. 50 Nevertheless, given that ECM remodeling is known to be associated with AMD risk, 7,130 and that MMP-8 is released during inflammation (again a key event in AMD), it seems reasonable to assume that MMP-8 may well be relevant in AMD, although clearly more research is required to fully address this question.
In line with this, a recent study showed that donors without any physiological signs of AMD, but who carried a genetic predisposition either solely in CFH p.Tyr402His or in Arms2/HTRA1 (rs10490924) presented with significantly elevated mast cells in the choroid, elevated deposition of mast cell-specific proteases in BrM, and increased amounts of denatured collagen within their BrM when compared to donors of similar age, but who carried no genetic risk for disease. 131 These findings add to a growing number of studies that imply a mechanistic link between the two distinct genetic-risk loci for AMD that appear to converge in the early stages of disease.
Although the afore-mentioned study did not investigate the underlying mechanisms behind their observation, it is intriguing to note Overall, there are multiple lines of evidence that the four pathways genetically associated with AMD, namely the complement system, ECM remodeling, lipid metabolism, and angiogenesis, interact with each other, and while knowledge on these interactions so far is limited, the first steps for a more holistic understanding are taken and generation of more complex models as described in the previous section will likely contribute to deepen the understanding of how these four pathways interact in the context of AMD.

| CLINI C ALTRIAL SAIMINGTOREDUCE COMPLEMENTAC TIVATI ONINAMD PATIENTS
The recent renaissance of complement-targeting therapeutic strategies in the AMD space is perhaps not surprising given the swathe of genetic and biochemical data implicating its major role as a driver of disease pathogenesis. 135 Early clinical trials focused on repurposing pre-existing complement inhibitors such as Eculizumab, a C5 inhibitor administered intravenously for the treatment of Paroxysmal nocturnal hemoglobinuria (PNH) and Atypical hemolytic uremic syndrome (aHUS). 136 However, intravenous administration of Eculizumab in AMD failed to show any effect in GA patients during a phase II trial (NCT00935883). 137  Therapeutics in their GEM103 program (NCT04246866: Table 1), but was abandoned mid-way through the subsequent phase II trials (NCT04643886) calling into question this therapeutic approach.
Furthermore, real-world injection data for current intravitreal therapies suggest a distinct lack of appetite in eye clinics around the world for a regularly administered, high resource utilization, therapeutic for a disease as common as dry AMD 140 making this delivery strategy itself, irrespective of therapeutic payload, a tough sell for any commercial venture trying to bring this option to the market.
Despite these concerns, the best complement-targeting therapeutic strategies for dry AMD to date are both administered by regular intravitreal injections (see Table 1). The therapies, termed APL-2 and Zimura that target C3 and C5, respectively, both reported a reduction in GA lesion growth of 29% and 27.8% during their respective phase II trials 141,142 : interim phase III trial data for APL-2 suggest a similar reduction in GA lesion size, of ~22%. Interesting is the striking similarity in results between the two therapies, as only APL-2 targets the amplification loop directly (i.e., reducing both C3a and C5a anaphylatoxins and the deposition of C3b), while Zimura targets further downstream of the cascade at C5. Additionally, both treatments saw an increased incidence of CNV in the treated eyes over sham treatment and an increase in cases of endophthalmitis, 141,142 a devastating ocular inflammatory condition where patients often have to have their infected eyes enucleated. 143 The increased incidence of CNV may be partly due to the pegylated (PEG) formulation of both therapies, as PEG has been used in the past to induce CNV in mouse models. 144 The effect of PEG itself on retinal cells is also brought into But perhaps ironically, the unique anatomy and immune privilege nature of the eye that poses problems for direct drug delivery, also represents an opportunity to utilize the very latest gene-therapy approaches, including the delivery of therapies via Adeno-associated Viruses (AAV). 145 The early development of gene therapies had been fraught with various difficulties, with examples of clinical trials having to stop due to the death of patients. 146 Indeed, major incidents recorded in clinical trials appear to be mediated in no small part by the AAV-delivery itself and not so much the payload. Interestingly, complement activation by the AAV capsid is believed to play a major role in the immunogenicity observed in these cases, 147  any of its primary outcomes (see Table 1).
Another intriguing possible strategy for targeting AMD is to not target the eye at all. Once such strategy is the targeting of FB by IONIS pharmaceuticals (Table 1), whose siRNA against the CFB gene targets the liver, thus lowering circulating levels of FB, is being tested in phase II trials for GA (NCT03815825). This is likely an attempt to address complement turnover in the ECM of the choriocapillaris, although it will presumably alter complement turnover elsewhere in the body too. Similarly, orally delivered small molecules targeting FB (Iptacopan) and FD (Danicopan) are also challenging the locally delivered therapeutic dogma for AMD (see Table 1), although quite how efficacious any of these strategies will be remains to be seen.
The recent discovery of elevated FHR proteins in the blood of AMD patients, and their accumulation in the choriocapillaris where they compete for the binding of FH and FHL-1 to deposited C3b, also makes them viable targets. 36,37,39,151 The FHR proteins are made primarily in the liver, so a similar siRNA strategy targeting the CFHR genes should also result in reduced circulating levels of the FHR pro- First of all, AMD is a multifactorial disease with genetic and environmental factors influencing disease risk, and while the result, that is, the phenotype, is the same, the initiating triggers vary between patients. Recent studies suggest that eventually the different pathways merge and thereby contribute to the same disease outcome, but further understanding is needed as to how exactly this happens on a molecular level. This is especially relevant on context of emerging therapies in regard of patient selection and timing of treatment intervention. Second, the eye is an immune-privileged site and outer blood-retinal barrier forms a unique structure being in itself a passive regulator of complement activation by having selective diffusion properties as to which complement components may or may not enter the retina, and, as discussed in the previous section, these properties may in fact make the difference between success and failure of complement inhibiting therapy in AMD. Tightly connected to this observation is the third point, that even though systemic levels of complement activation have been extensively studied due to easy accessibility of tissue, data on local processes are comparably scarce, especially when looking into studies including genotype information. Unfortunately, it is impossible to obtain longitudinal data on ocular tissue from the same patient, but recent advances in disease modeling utilizing "retina on a chip" approaches hold great promise to fill this gap in the future.

ACK N OWLED G EM ENTS
We Clark are supported by the Helmut Ecker Foundation.

CO N FLI C TO FI NTE R E S T
SJC is an inventor named in patent applications that describe the use of complement inhibitors for therapeutic purposes and the use of circulating complement protein measurements for patient stratification, and is a co-founder and shareholder of Complement Therapeutics, a company which focuses on the development of complement targeted therapeutics, including for AMD. SdJ and JT have no conflicts of interest.

DATAAVA I L A B I L I T YS TAT E M E N T
Data sharing is not applicable to this article as no new data were created or analyzed in this study.