Identification of Factor H–like Protein 1 as the Predominant Complement Regulator in Bruch’s Membrane: Implications for Age-Related Macular Degeneration

Age-related macular degeneration (AMD) is the leading cause of blindness in the developed world, affecting ∼50 million people worldwide. The prevalence of this condition is predicted to rise as the elderly population expands: in the United States it has been estimated that there will be a 50% increase in the number of affected individuals between 2004 and 2020 (1). This debilitating disease can be subdivided into neovascular (“wet”) and atrophic (“dry”) AMD (2), both of which are usually preceded by the formation of drusen. These aggregates of lipids, proteins, and cellular debris accumulate within Bruch's membrane, a sheet of extracellular matrix (ECM) that separates the retinal pigment epithelium (RPE) from the blood vessels of the choroid. The presence of complement proteins and downstream inflammatory markers in drusen has led to the hypothesis that chronic local inflammation in Bruch’s membrane and surrounding structures, resulting from inappropriate complement activation, has a major influence on the pathogenesis of AMD (3–5).

Genetic alterations are a major risk factor for AMD. Two major loci have been identified: one is on chromosome 10 near the ARMS2/HTRA1 genes and the other is on chromosome 1 involving complement factor H (FH) and the FH-related (FHR) proteins (6). Additionally, genes encoding members of the alternative complement pathway have been implicated, including C3, complement factor I (FI), and complement factor B (FB), thereby providing strong evidence that this pathway is involved in AMD pathogenesis (6–8). The locus on chromosome 1 is complex with multiple haplotypes having been identified that modify AMD risk (9).

Within the chromosome 1 locus the Y402H polymorphism in FH represents a major risk factor for AMD (10–14). The frequency of this risk allele is ∼35% in individuals of European descent and results in a tyrosine being replaced by a histidine residue at position 402 (using the preprotein sequence numbering) (15). The effects of the Y402H polymorphism appear to be mediated locally, as the polymorphism does not promote complement activation in the blood (14), unlike some mutations found in the C-terminal complement control protein (CCP) domains of FH.

The central activating mechanism of the alternative pathway is the covalent deposition of the protein C3b (an opsonin) on all local surfaces. Surface-linked C3b can react with other complement proteins to form an active enzyme, the C3 convertase that is able to produce further (surface-attachable) C3b molecules. This is achieved as the C3 convertase proteolytically processes C3 in blood into fresh C3b molecules, producing at the same time the anaphylatoxin C3a. Insufficient control of the C3 convertase results in massive production of C3b and C3a molecules and a shift of the complement cascade to its terminal lytic pathway. This produces the most potent anaphylatoxin, C5a, and the cell lytic protein complex termed the membrane attack complex; both C5a and the membrane attack complex provide strong inflammatory signals.

However, the presence of FH on host tissues, and its cofactor activity for FI, results in C3b breakdown (resulting in the formation of inactive iC3b), thereby preventing inappropriate complement activation and inflammation (16). FH also exerts decay-accelerating activity, which can assist in the deconstruction of already formed C3 convertases. Whereas cell surface–expressed complement regulators also prevent complement activation on cells, the blood-borne FH is currently the only known complement regulator to bind and confer protection to ECM such as Bruch’s membrane (17). The protective FH is recruited to self-surfaces, at least in part, by binding specific polyanions such as glycosaminoglycan (GAG) chains or sialic acid groups, which are not normally present on potential pathogens. GAGs are long, unbranched polysaccharides made up of repeating disaccharide units that can be variably sulfated: the sulfation pattern can drive specific protein recruitment (18). One ubiquitously expressed GAG, and a major ligand for FH, is heparan sulfate (HS), which comprises both low- and high-sulfated regions (18). These sulfated ligands are essential ECM components responsible for a range of biological processes, including immune homeostasis (19, 20).

FH comprises 20 CCP domains and contains two main GAG-binding regions, in CCP7 (with some contribution from CCPs 6 and 8) and CCP20 (21) (see Fig. 1A): the Y402H polymorphism resides in CCP7 and alters FH binding to sulfated GAGs (22). The 402H disease-associated variant binds significantly less well to human Bruch’s membrane (an important site in AMD pathogenesis) than does the 402Y form when applied exogenously to tissue sections (23). We have demonstrated that the GAG binding region in CCP6–8 is responsible for surface anchoring and hence host recognition in eye structures, including Bruch’s membrane, whereas the CCP19–20 region anchors FH to ECM in the kidney (24).

In this study we investigated the distribution of a naturally occurring truncated form of FH called FH-like protein 1 (FHL-1), which arises from alternative splicing of the CFH gene (25). FHL-1 is identical to FH for the first seven CCP domains before terminating with a unique 4-aa C terminus (see Fig. 1A). Importantly, FHL-1 retains all the necessary domains for function and is also subject to the Y402H polymorphism. Previous studies have demonstrated FHL-1 expression by RPE cells (13, 26) and it has been identified in the vitreous of the eye where FH and FHL-1 were reported to be in equimolar concentrations (27). In human blood the concentration of FH is ∼300 μg/ml (28) and FHL-1 ∼50 μg/ml (29), resulting in a molar ratio of ∼2:1.

Both recombinant 402H and 402Y forms of FHL-1 were generated according to the methodology described previously (30). The full-length FH protein was purified from human plasma, as described by Hakobyan et al. (31). Commercial Abs used in this study were OX23 (AbD Serotec, Kidlington, U.K.) (32) and L20/3 (Hycult Biotech, Uden, the Netherlands), which has been published previously with the clone name C02 (33). Both of these Abs recognize different epitopes on FH (see Fig. 1A). Anti–FHL-1 was generated against the peptide sequence CIRVSFTL (Mimotopes, Clayton, Australia). Anti–FHL-1 IgG was purified using an affinity column: recombinant FHL-1 was coupled to cyanogen bromide–activated Sepharose according to the manufacturer’s instructions (GE Healthcare, Buckinghamshire, U.K.). Briefly, rabbit antiserum was centrifuged at 12,000 × g for 5 min at room temperature to remove particulate matter. The antiserum was then run down the FHL-1 affinity column in PBS, 1 mM EDTA. The column was washed thoroughly with 5 column volumes 0.4 M NaCl to remove weakly bound material. Two column volumes 3 M MgCl₂ was used to elute bound protein. The column was then regenerated by washing with 5 column volumes 0.4 M NaCl and re-equilibrating with 5 column volumes PBS, 1 mM EDTA. Eluted protein was dialyzed into 2 l Milli-Q H₂O for 16 h at 4°C using 10 kDa cutoff dialysis tubing before being further dialyzed into 20 mM NaH₂PO₄, 150mM NaCl (pH 8.0) for a further 16 h at 4°C. Rabbit IgG was purified from the dialysate using protein A–Sepharose, following the manufacturer’s instructions (Sigma-Aldrich, Poole, U.K.). Isolated IgG was run on 4–12% NuPAGE reducing gels to check for purity and tested for specificity to FHL-1 in solid phase assays (see Supplemental Fig. 1).

Details of donor eye tissue used in this study are listed in Table I. Human eyes were obtained from the Manchester Royal Eye Hospital Eye Bank after removal of the corneas for transplantation. Our research adhered to the tenets of the Declaration of Helsinki. In all cases, there was prior consent for the eye tissue to be used for research, and guidelines established in the Human Tissue Act of 2004 (U.K.) were followed. Except in the case of donor tissue used for the staining of drusen, none of the other donors had a history of visual impairment or eye disease.

For the donor eye pairs used for immunohistochemistry, PCR analysis, and Western blotting, one globe was designated for RPE cell isolation and Bruch’s membrane enrichment. These eyes were opened by making three incisions into the eyecup and flattening out the tissue. The vitreous and neurosensory retina were removed, RPE cells were harvested with gentle scraping, and RNA was isolated (see below). The Bruch’s membrane was enriched by removal of the sclera and choroid. The Bruch’s membrane was washed multiple times with PBS and either frozen for analysis by Western blotting or used in Ussing chamber diffusion experiments (both methodologies are described below).

The second globe from each donor was used to prepare frozen tissue sections and was fixed in 4% (v/v) formaldehyde for 2 h. Eyes were then processed and serial sections of 5 μm were prepared using a cryostat, as described previously (23, 34).

Tissue sections were stained for the presence of endogenous FH or FHL-1 using methods described previously (23). Briefly, tissue sections were incubated with chilled (−20°C) histological grade acetone (Sigma-Aldrich) for 20 s before thorough washing with PBS. Tissue sections were blocked with 0.1% (w/v) BSA, 1% (v/v) goat serum, and 0.1% (v/v) Triton X-100 in PBS for 1 h at room temperature. After washing with PBS, tissue sections were incubated with Ab combinations of either 10 μg/ml mix of L20/3 and anti–FHL-1 or of OX23 and anti–FHL-1 for 16 h at 4°C. Sections were washed and Ab binding was detected using fluorescently labeled secondary Abs (Life Technologies, Paisley, U.K.). An equal 1:5000 dilution mix of Alexa Fluor 488–conjugated goat anti-rabbit (to detect anti–FHL-1) and Alexa Fluor 594–conjugated goat anti-mouse (against OX23 or L20/3) was added to sections for 2 h at room temperature. Finally, DAPI was applied as a nuclear counterstain (at 0.3 μM for 5 min) prior to mounting with medium (Vectashield; Vector Laboratories, Peterborough, UK) and application of a coverslip.

In some experiments the enzymatic pretreatment of tissue was performed as described previously (23). Briefly, 20 U/ml each of heparinase I, II, and III (all from Flavobacterium heparinum, Sigma-Aldrich) in PBS was applied to tissue sections for 1.5 h at 37°C; this was performed after fixation in acetone and followed by washing with PBS prior to the blocking step.

Images were collected on a snapshot widefield microscope (Olympus BX51) using a ×40/0.30 Plan Fln objective. Microscopy images were captured using a CoolSNAP ES camera (Photometrics) via MetaVue software (Molecular Devices). To prevent bleed-through of color from one channel to the next, specific band pass filter sets were used for DAPI, FITC, and Texas Red. All images were handled using ImageJ64 (version 1.40g; http://rsb.info.nih.gov/ij).

Samples were run on 4–12% NuPAGE Bis-Tris gels (Life Technologies) at 200 V for 60 min and transferred onto nitrocellulose membranes at 80 mA for 2.5 h using semidry transfer apparatus in transfer buffer (25 mM Tris, 192 mM glycine, 10% [v/v] methanol). The membranes were blocked in PBS, 10% (w/v) milk, and 0.02% (w/v) BSA for 16 h at 4°C before the addition of an OX23 and anti–FHL-1 Ab mix, each at 0.5 μg/ml, in PBS with 0.2% (v/v) Tween 20 (PBST) for 2 h at room temperature. Membranes were washed twice for 30 min in PBST before the addition of a 1:2000 dilution of IRDye 680RD–conjugated goat anti-mouse and IRDye 800CW–conjugated goat anti-rabbit (LI-COR Biosciences U.K., Cambridge, U.K.) for 2 h at room temperature, protected from light. Membranes were washed and protein bands were visualized using a LI-COR Odyssey infrared imaging system and Image Studio software.

Total RNA was isolated from human RPE cells isolated from the donors listed in Table I, using standard extraction and purification methods. Briefly, RPE cells were homogenized in 1 ml cold TRI Reagent (Life Technologies) per 30 mg tissue using a TissueLyser II bead mill (Qiagen). After homogenization, RNA was isolated following TRI Reagent/BCP disruption and phase separation. RNA was further purified by absorption to an RNeasy mini spin column (Qiagen) with on-column DNAse I treatment. RNA purity and concentration, as measured by a NanoDrop spectrophotometer, were determined by absorbance at 230, 260, and 280 nm. cDNA was synthesized using a Transcriptor High Fidelity cDNA synthesis kit (Roche Diagnostics, Burgess Hill, U.K.) with 1 μg RNA and primed with oligo(dT)₁₈ primer according to the manufacturer’s instructions.

Target gene sequences were obtained from GenBank and PCR primers to specific targets were designed with Primer3 software (http://www.ncbi.nlm.nih.gov/tools/primer-blast/) and are shown in Supplemental Table I. Primers for the RPE cell–specific genes (Bestrophin-1 and RPE65) were also included to ensure purified cDNA was indeed obtained from RPE cells. cDNA was amplified using the following PCR protocol: 95°C for 5 min followed by 40 cycles of 95°C for 10 s, 60°C for 15 s, and 72°C for 20 s, followed by a melting program. The integrity of the PCR reactions was verified by detection of a single band of the correct size by agarose gel electrophoresis. The same experiments were also performed on pooled liver cDNA (Sigma-Aldrich).

The macular region of enriched Bruch’s membrane isolated from donor eyes (described above) was mounted in an Ussing chamber (Harvard Apparatus, Hamden, CT). Once mounted, the 5-mm-diameter macular area was the only barrier between two identical compartments (see Fig. 3A). Both sides of Bruch’s membrane were washed with 2 ml PBS for 5 min at room temperature. Human serum (Sigma-Aldrich) was diluted 1:1 with PBS and 2 ml was added to the Ussing compartment representing the choroidal side of Bruch’s membrane. After 1 min when no leaks were detected into the second compartment (which would indicate a compromise in membrane integrity), 2 ml PBS alone was added to the second compartment and the Ussing chamber was left at room temperature for 24 h with gentle stirring in each compartment to avoid generating gradients of diffusing proteins. Samples from each chamber were analyzed by gel electrophoresis. Gels were either stained with Instant Blue stain (Expedeon, Harston, U.K.) for 60 min at room temperature or subject to Western blotting (see above).

In some experiments the above protocol was repeated using 100 μg/ml FH purified from human plasma (31) on three donor Bruch’s membranes (see Table I). After 24 h the entire PBS compartment, where FH would diffuse into, was collected and StrataClean beads (Agilent Technologies, Cheadle, U.K.) used to pull all proteins out of solution. The entire content of the PBS compartment was analyzed by gel electrophoresis as described above. Similarly, to ascertain whether the AMD-associated polymorphism altered the diffusion properties of FHL-1, 100 μg/ml recombinant 402H or 402Y forms of FHL-1 were separately tested on three donor Bruch’s membranes (see Table I). The protein content of each Ussing chamber compartment was analyzed by gel electrophoresis and band densities were normalized and compared with a 0 h, 100 μg/ml FHL-1 sample.

The fluid-phase cofactor activity of the FHL-1 402H and 402Y variants were measured by incubating FHL-1 (either form), C3b, and FI together in a total volume of 20 μl PBS for 15 min at 37°C. For each reaction, 2 μg C3b and 0.04 μg of FI were used with varying concentrations of FHL-1 ranging from 0.0125 to 0.8 μg per reaction. The assay was stopped with the addition of 5 μl 5× SDS reducing sample buffer and boiling for 10 min at 100°C. Samples were run on a 4–12% NuPAGE Bis-Tris gel at 200 V for 60 min to maximize the separation of the C3b breakdown product bands. The density of the 68-kDa iC3b product band was measured using ImageJ64 (version 1.40g; http://rsb.info.nih.gov/ij) and used to track C3b breakdown efficiency of the FHL-1 proteins. For both forms of FHL-1, averaged data from three separate experiments were used.

The heparin-binding characteristics of the FHL-1 (402H and 402Y variants) proteins were analyzed using microtiter plate-based assays, where either heparin or one of its selectively desulfated derivatives (all from Iduron) was immobilized noncovalently on allylamine-coated heparin-binding plates (BD Biosciences, Oxford, U.K.) as described previously (22, 35). The selectively desulfated heparin samples used in this study were 2-O–desulfated, 6-O–desulfated, and N-desulfated heparin. HS from either porcine mucosa (Iduron) or bovine kidney (Sigma-Aldrich) was also tested under the same conditions.

All GAGs were diluted in PBS and immobilized at 1 μg/well in a volume of 100 μl/well overnight at room temperature. Plates were blocked for 90 min at 37°C with 300 μl/well 1% (w/v) BSA in assay buffer (20 mM HEPES, 130 mM NaCl, 0.05% [v/v] Tween 20 [pH 7.3]). This standard assay buffer was used for all subsequent incubations, dilutions, and washes, and all steps were performed at room temperature. FHL-1 protein was incubated with the immobilized GAGs for 4 h. After washing, bound protein was detected by the addition of 100 μl/well 0.5 μg/ml OX23 Ab and incubated for 30 min followed by washing and a 30-min incubation in 100 μl of a 1:1000 dilution of alkaline phosphatase–conjugated anti-mouse IgG (Sigma-Aldrich). Plates were developed using 100 μl/well 1 mg/ml disodium p-nitrophenylphosphate solution (Sigma-Aldrich) in 0.05 M Tris-HCl, 0.1 M NaCl (pH 9.3). The absorbance values at 405 nm were determined after 10 min of development at room temperature and corrected against blank wells (i.e., those with no immobilized GAGs).

The heparin-binding properties of the 402H and 402Y FHL-1 variants were compared by affinity chromatography on a heparin affinity column in which 30 mg heparin (Iduron) was coupled to 1.5 ml cyanogen bromide–activated Sepharose (GE Healthcare) in 0.1 M NaHCO₃, 0.5 M NaCl (pH 8.3) using the manufacturer’s protocol. Before sample loading, the column was equilibrated in 5 ml PBS (Sigma-Aldrich). Purified recombinant protein (100 μg) was loaded onto the column in a total volume of 5 ml PBS. The column was washed with 4 column volumes PBS before bound protein was eluted with a linear salt gradient of 130 mM–1 M NaCl over 20 ml by mixing elution buffer (PBS, 1 M NaCl) with equilibration buffer at a flow rate of 1 ml/min; fractions (1 ml) were collected throughout the protocol.

To determine whether FHL-1 was present in human Bruch’s membrane we employed the Abs OX23 (which recognizes both FH and FHL-1) (32), L20/3 (previously published as clone C02 and recognizes only FH) (33), and a specific anti–FHL-1 Ab that we generated (Fig. 1A). The anti–FHL-1 and L20/3 Abs were shown to be specific for their intended targets and exert no cross-reactivity when tested against recombinant FHL-1 or FH purified from human serum (Supplemental Fig. 1A–C). Fluorescent staining of six separate human maculae (Table I) identified FHL-1 throughout Bruch’s membrane (Fig. 1B). In contrast, full-length FH protein was identified on the choroidal side of Bruch’s membrane (with particular accumulation in the choriocapillaris), and small amounts were present in patches on the RPE-facing side, but staining was not seen within Bruch’s membrane (Fig. 1C, 1D). The choroidal stroma stained for (full-length) FH, and there was weak staining for FHL-1, especially near Bruch’s membrane. FHL-1 staining could be ablated by preabsorbing the Ab with recombinant protein (Supplemental Fig. 1D, 1E), thereby providing further evidence that the staining identified endogenous protein. Additionally, as expected, we demonstrated colocalization of anti–FHL-1 and OX23 Abs throughout Bruch’s membrane (Supplemental Fig. 1F). Analyses of drusen from four donors (Table I) demonstrate staining for endogenous FH in the periphery of the lesions whereas staining with FHL-1 appeared throughout the lesions (Fig. 1E).

We performed Western blots on extracts from isolated human Bruch’s membrane and probed them with the anti–FHL-1 and OX23 Abs. We identified an ∼49-kDa band that migrated to the same position as recombinant FHL-1 protein control (Fig. 1F). This was identified by both OX23 and anti–FHL-1, thereby confirming that this band was FHL-1 and not one of a number of known tryptic fragments of FH (25, 36). The OX23 Ab was, however, unable to detect a 155k-Da species corresponding to FH in the extracts.

We investigated mRNA expression of a number of complement genes by RPE cells isolated from donor tissue (for a list, see Supplemental Table I). We found that most genes involved in an alternative pathway complement response were transcribed, including C3, FB, FI, FHL-1, and FH (Fig. 2A).

Next we investigated whether FH and FHL-1 could contribute to the protection of the Bruch’s membrane/RPE interface by diffusing across Bruch’s membrane from the blood supply. Isolated Bruch’s membrane was sandwiched between two compartments of an Ussing chamber (see Fig. 3A) as previously described (37), and the diffusion of FH and FHL-1 from human serum across the membrane was investigated. After 24 h only an ∼49-kDa band was identified in the diffusate by Western blot using OX23 (Fig. 3B). Dual staining of this band with OX23 and anti–FHL-1 confirmed that this was indeed FHL-1 traversing the Bruch’s membrane (Fig. 3B). In contrast, FH (155 kDa) was not detected. The experiments were repeated using purified FH, and after 24 h the entire protein content of the diffusate chamber was concentrated and analyzed, and we were unable to detect any FH protein having crossed Bruch’s membrane (Fig. 3C). To ascertain whether the Y402H AMD-associated polymorphism affected FHL-1 diffusion, we repeated the experiments using purified 402H and 402Y forms of FHL-1 (Fig. 3D, 3E). In the case of both variants, equilibrium was reached, although a reduction in the calculated protein recovery for the 402Y form was observed (an average of 73% over three separate experiments opposed to 92% for the 402H form).

We assessed whether the AMD-associated Y402H polymorphism would alter the ability of FHL-1 to regulate complement activation on host surfaces by catalyzing the FI-mediated breakdown of C3b to iC3b. When run on reducing gradient gels, the α- and β-chains of C3b can be clearly separated (Supplemental Fig. 3A). When C3b is incubated with FI and either the 402Y or 402H form of FHL-1, the proteolytic cleavage of the C3b α-chain by FI yeilds bands at 68 and 43 kDa (38), the appearance of which can be used as a measure of FHL-1 cofactor activity. In the present study, we noted no difference in the ability of the two FHL-1 forms to break down C3b (see Supplemental Fig. 3B). This result was predictable because the C3b/FI binding regions and the Y402H polymorphism site are at opposite ends of the FHL-1 protein (Fig. 1A) (39, 40).

Although the Y402H polymorphism does not affect the ability of FHL-1 to inactivate C3b, it does reside in the protein’s only surface recognition domain in CCP7 (Fig. 1A). Previously using FH and recombinant proteins representing the CCP6–8 region of FH, the Y402H polymorphism has been shown to affect GAG binding (22, 23). In this study, we investigated the binding of the FHL-1 402H and 402Y proteins to selectively desulfated heparin. The heparin was desulfated at the 6-O, 2-O, and N positions (see Fig. 4A), and although this did not affect the binding of the 402Y form of FHL-1, desulfation did significantly affect the binding of the AMD-associated 402H form (Fig. 4B, 4C).

Heparin is, however, a convenient highly sulfated model of the HS that is found in ECM. HS contains large regions with low sulfated disaccharides and overall has much less sulfation than heparin (Fig. 4D) (18). The experiments with selectively desulfated heparin demonstrated that the disease-associated 402H variant of FHL-1 relies heavily on highly sulfated GAG sequences and therefore is likely to bind less well to HS than heparin. To confirm this, solid-phase binding experiments with two sources of HS (porcine mucosa and bovine kidney) were performed and these confirmed that the 402H form of FHL-1 binds significantly less well than does the 402Y form to HS (Fig. 4E, 4F). Conversely, the 402H form of FHL-1 bound better than did the 402Y form to highly sulfated heparin (Supplemental Fig. 2), which agrees with the influence of sulfation availability on 402H binding, but also highlights the unsuitability of heparin as a physiological model for FHL-1/HS interactions.

The removal of HS from macula tissue sections by pretreatment with a heparinase I/II/III mix pretreatment reduced the signal seen for endogenous FHL-1 (Fig. 5B), indicating that HS is indeed one of the main ligands anchoring FHL-1 to the Bruch’s membrane. Interestingly, enzymatic treatment of drusen did not appear to alter detectable levels of FHL-1 within the druse, but did remove the distinctive staining of FH around the edges of the lesions (Fig. 5C, 5D).

Since 2005, a large body of work has focused on alterations in FH as major genetic contributors to AMD pathogenesis (6, 41–44) and the importance of the alternative complement pathway. This has been supported by evidence of complement proteins being present in AMD macular tissue, including markers of dysregulation (3, 22, 23, 45). Previous studies have concluded that FH, as a blood-borne protein, is the only complement regulator that would confer protection to ECM such as Bruch’s membrane. Genetic studies have not distinguished between FH and FHL-1 (given that they share the same gene), and although FHL-1–specific Abs have been made in the past (46) they have not been used to probe eye tissue (27).

In this study, we show that FHL-1 is likely to be conferring greater protection to Bruch’s membrane than does FH, whereas the latter is the predominant form protecting the ECM of the choroid (Fig. 1C, 1D). Furthermore, we confirm that the RPE cells transcribe genes involved in complement activation and regulation on Bruch’s membrane, and our data demonstrate that FHL-1 is locally transcribed (Fig. 2). As well as a local contribution to FHL-1 accumulation in the Bruch’s membrane, we have demonstrated that the 49-kDa protein could passively diffuse across Bruch’s membrane from the choroidal vasculature, whereas the 155-kDa glycosylated FH protein cannot (Fig. 3B, 3C). This latter observation supports the previous finding that the Bruch’s membrane confers a size limit on proteins able to diffuse passively across it, the size of which decreases with age (37). The AMD-associated Y402H polymorphism does not affect the diffusion of FHL-1 across Bruch’s membrane (Fig. 3D, 3E), but differences in the amount of recovered protein between the 402H and 402Y forms may be indicative of more 402Y FHL-1 binding to Bruch’s membrane in the Ussing chamber as it passes through compared with the 402H form.

Small patches of FH were observed on the RPE-facing side of Bruch’s membrane (Fig. 1D), which are likely to originate from RPE cells. Local synthesis would also explain the distinctive FH labeling on the surface of drusen, which may be unable to penetrate into them, whereas the smaller FHL-1 can penetrate into the drusen (Fig. 1E). Interestingly, the FH staining around drusen is removed by heparinase treatment (Fig. 5), suggesting that the protein is retained on the surface of drusen by binding HS chains. The lack of heparinase effect on FHL-1 staining within drusen suggests that the protein is either interacting with a currently unidentified ligand, or it is trapped among the plethora of components that make up these hallmark lesions of AMD. Similarly, it may be possible that the heparinase itself is not able to penetrate the tight matrix of the druse. Whether FHL-1 retains activity in this environment is unclear.

We show that the FHL-1 within Bruch’s membrane is immobilized there largely through interactions with HS. In other recent work we have shown that there is a marked decrease in the levels of HS GAGs in Bruch’s membrane with age (47). This decrease in HS levels in Bruch’s membrane coupled with the poorer binding of the 402H form of FHL-1 could explain why AMD is a disease of aging. Individuals with the 402H polymorphism may, as they age, be unable to localize sufficient FHL-1 to Bruch’s membrane so that its protective effects are lost and the complement cascade is activated with damaging consequences that ultimately lead to AMD. Furthermore, the dominant role of FHL-1 in complement regulation at this site provides, to our knowledge for the first time, an explanation of why the GAG-binding region in CCP7 was found to be vital for protein localization in the eye (23), whereas CCPs 19–20 of FH are more important in the kidney (24, 33, 48). The absence of full-length FH in Bruch’s membrane (and thus the CCPs 19–20) may have resulted in evolutionary pressure to express HS species capable of recruiting FHL-1 via its CCP7 domain.

Although our work suggests that FHL-1 is a major regulator of complement in Bruch’s membrane, there is genetic evidence that FH is also important. A highly penetrant mutation R1210C in the C terminus of FH is a strong risk factor for AMD and is thought to coincide with a 6-y earlier onset of the disease (9). The R1210C form of FH is exclusively found covalently bound to albumin in plasma (49), which would affect the proteins mobility, and although albumin binding has no effect on GAG binding, it does alter FH binding to C3b (49, 50). As such, it is likely this mutation hampers FH tissue penetration and surface protection. Furthermore, most AMD patients with the R1210C haplotype have the FH/FHL-1 402H polymorphism on the other allele (9). It may be the case that any contribution conferred by FH to immune regulation in the macula is perturbed by the R1210C mutation and amplifies an already imbalanced immune homeostasis (conferred by linked genetic factors such as the Y402H polymorphism).

Our findings also have implications for understanding how the five FHR proteins contribute to immune homeostasis in the eye. Variations in the genes encoding the FHR proteins are associated with alterations in AMD risk (41, 42), and the FHR proteins can compete with FH binding to C3b and/or GAGs, and even to form a novel C3 convertase (51). As such, they are fast being considered newly identified positive regulators of the alternative pathway of complement (41, 52–54). However, many of these biochemical studies have assumed that the FHR proteins would be competing with FH as the main regulator, which in the context of Bruch’s membrane is not the case, and so their competitive properties with FHL-1 may well differ.

The unique FHL-1 C terminus means that it is necessary to investigate the binding characteristics of known FH ligands such as C-reactive protein (55, 56) and markers of oxidative stress (such as malondialdehyde) (57) in relationship to the Y402H polymorphism. Furthermore, the results shown in the present study may have implications for immune homeostasis in other tissues and in other diseases where FH struggles to access ECM structures or extracellular debris, whereas FHL-1 can. An example is the brain lesions of Alzheimer disease patients (58), and it is of note that changes in HS sulfation are also associated with this condition (59). The identification of FHL-1 as the main complement regulator at a key site in AMD pathogenesis shapes our understanding of the molecular biology of the disease. Only by elucidating the precise biochemistry behind complement regulation at the macula alongside understanding genetic regulation of the implicated proteins can we hope to design successful therapeutic strategies for this debilitating disease.

We thank Prof. Paul Barlow (University of Edinburgh, Edinburgh, U.K.) and Prof. Anthony Day (University of Manchester, Manchester, U.K.) for advice, support, and direction during the early stages of this work. Also, we thank Dr. Isaac Zambrano (Manchester Eye Bank, Manchester Royal Eye Hospital, Manchester, U.K.) for assistance with obtaining the donor eye tissue used in this study.

The authors have no financial conflicts of interest.