Tissue-Specific Host Recognition by Complement Factor H Is Mediated by Differential Activities of Its Glycosaminoglycan-Binding Regions

Complement factor H (CFH), a 155-kDa plasma protein, is the main regulator of the alternative pathway of complement, a key component of the innate immune response. The presence of CFH in tissues (where it is bound to host surfaces) leads to the breakdown of any deposited C3b that would otherwise result in complement amplification (1). Although other proteins exert regulatory activity on host cell surfaces, i.e., CD35, CD46, and CD55 (2), CFH is the only alternative pathway regulator to confer protection to extracellular matrix (ECM) structures such as basement membranes (3, 4). Impaired CFH function is a major contributing factor in the kidney disease atypical hemolytic uremic syndrome (aHUS) (5, 6) and age-related macular degeneration (AMD) (7–13), which is the leading cause of blindness in the western world (14). In AMD, complement dysregulation is associated with drusen formation between the retinal pigment epithelium (RPE) and a multilaminar ECM called Bruch’s membrane (11), resulting in the destruction of the macula and consequent loss of central vision.

CFH recognizes host surfaces through its interactions with certain polyanions, such as the glycosaminoglycan (GAG) chains of proteoglycans that are present on all cells and in the ECM of host tissues (12, 15, 16); GAGs are long, unbranched polysaccharides composed of disaccharide repeats that show a huge diversity in sequence based largely on variable sulfation patterns (see Supplemental Fig. 1). CFH is composed of 20 complement control protein (CCP) domains (17) (also referred to as short consensus repeats) containing two GAG-binding regions corresponding to CCP6–8 and CCP19–20 (18–25). Importantly, mutations and polymorphisms in these GAG-binding regions of CFH have been associated with both AMD (7–13) and aHUS (5, 6), where in some cases they have been shown to affect the interaction of CFH with GAGs (12, 20, 23, 25). For example, the Y402H polymorphism of CFH (26), a major risk factor for AMD (7–10, 27), alters its binding to the GAGs heparin, heparan sulfate (HS), and dermatan sulfate (DS) (12, 20, 23) through a tyrosine-to-histidine substitution at residue 402 within CCP7.

Functional changes within the CFH protein, caused by mutations/polymorphisms, could alter its ability to bind host surfaces. The current understanding is that CCP19–20 are chiefly responsible for host recognition (28), which is consistent with the large number of polymorphisms/mutations in this region associated with aHUS (5, 6). However, with the exception of R1210C (13), mutations/polymorphisms in CCP19–20 are not associated with an increased risk for development of AMD, and the Y402H polymorphism has no effect on the susceptibility of an individual to develop aHUS. We found previously that the Y402H polymorphism had a significant effect on the binding specificity of CFH for sulfated GAGs (20, 23), and that the 402H form of CFH bound less well to the Bruch’s membrane of human macula, where this interaction was mediated, at least in part, by the GAGs HS and DS (12). Poorer binding of the AMD-associated 402H variant to this ECM could lead to reduced complement regulation at this site, resulting in chronic local inflammation and thereby promoting the tissue damage and drusen formation that characterize AMD (3, 4, 12). On this basis we have hypothesized that the two GAG-binding sites have tissue-specific activities, potentially via their recognition of different HS structures. To test this hypothesis, we have compared the GAG-binding specificities of CCP6–8 and CCP19–20, and have investigated their relative contributions to the interactions of CFH with human eye and kidney tissues.

The 402H and 402Y polymorphic variants of full-length CFH protein (flCFH) were purified from the plasma of genotyped donors who were homozygous for either allele, as described by Hakobyan et al. (29); these preparations each had isoleucine at position 62. Recombinant proteins representing the 402H and 402Y variants of CFH and comprising CCP6–8 were expressed in Escherichia coli, refolded, and purified as described previously (20). A recombinant protein comprising CCP19–20 of CFH (designated as residues 1107–1231 in the published CFH sequence, UniProt no. P08603) was expressed in Pichia pastoris as described previously (21); this recombinant protein sequence also includes the nonauthentic tetrapeptide EAEF at the N terminus. All of the recombinant proteins (which come from nonglycosylated regions of CFH) (17) have been shown previously to be correctly folded (20, 21, 23).

The fluorescent labeling of the 402H and 402Y variants of CFH (both CCP6–8 and full-length proteins) with Alexa Fluor 594 and Alexa Fluor 488, respectively, was done as described previously (12). Labeling of the CCP19–20 recombinant protein with Alexa Fluor 488 was carried out in an identical manner, using the same protein labeling kit (Molecular Probes, Paisley, U.K.) in the presence of 10-fold molar excess of heparin dp24 oligosaccharide (Iduron, Manchester, U.K.) to protect the GAG-binding site. The heparin dp24 was removed from the labeled CCP19–20 protein by exhaustive dialysis against PBS supplemented with 1 M NaCl using 10-kDa molecular mass cutoff snakeskin dialysis tubing (Pierce, Cramlington, U.K.) in the dark at 4°C, as described previously (12). The labeling efficiency was calculated to be 1.5 mol fluorophore per mole of protein (the same degree of labeling as for the CCP6–8 proteins) (12). Labeling of the CCP19–20 protein was shown to have no major effect on its heparin-binding properties by affinity chromatography (data not shown). Biotinylation of the CCP6–8 and CCP19–20 recombinant proteins, also performed in the presence of 10-fold molar excess heparin dp24 oligosaccharide (Iduron) to protect the GAG-binding region, was carried out using the same methodology as used previously (20).

Biotinylation of the second international standard (2IS) of heparin (30), chondroitin-4-sulfate (C4S), chondroitin-6-sulfate (C6S), and DS was carried out as described by Clark et al. (20). Biotinylation of hyaluronan (HA) was carried out essentially as described previously (31). In brief, medical-grade HA (Lifecore Biomedical, Chaska, MN) was dissolved overnight in 0.1 M MES, pH 5.5, at a concentration of 5 mg/ml. To 1 ml of this solution, 13 μl of 25 mg/ml 1-ethyl-3-(3-dimethylaminopropyl)carbo-dimide hydrochloride (Sigma, Poole, U.K.) in 0.1 M MES, pH 6.5, and 20 μl 50 mM Biotin-LC-hydrazide (Pierce) in DMSO were added simultaneously and mixed at room temperature for 16 h. The sample was then centrifuged at ∼12,000 × g for 1 min to remove particulate material, and the reaction mixture was dialyzed into H₂O at 4°C for 6 h using 10,000 Da molecular mass cutoff snakeskin dialysis tubing (Pierce).

The heparin-binding characteristics of the biotinylated CCP6–8 (402H and 402Y variants) and CCP19–20 recombinant proteins were analyzed using microtiter plate-based assays, where 2IS heparin (20), or one of its selectively desulfated derivatives (32, 33), was immobilized noncovalently on allylamine-coated BD Heparin Binding plates (BD Biosciences, Oxford, U.K.) as we have described previously (20, 34). The selectively desulfated samples of the 2IS heparin used in this study were 2-O-desulfated, 6-O-desulfated, 2,6-O-desulfated, N-desulfated, and N-desulfated re–N-acetylated. The GAGs were immobilized at 1 μg/well on BD Heparin Binding plates in a total volume of 200 μl/well PBS, overnight at room temperature. Plates were blocked for 90 min at 37°C with 300 μl/well 1% (w/v) BSA in standard 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 at room temperature. Biotinylated recombinant protein was incubated with the immobilized GAGs for 4 h at room temperature. After the plates were washed, bound protein was detected by the addition of 200 μl/well of a 1:10,000 dilution of ExtrAvidin alkaline phosphatase (Sigma-Aldrich, Poole, U.K.) for 30 min at room temperature. After further washing, plates were developed using 200 μl/well of a 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 40 min of development at room temperature and corrected against blank wells (i.e., those that contained no immobilized GAGs).

Alternatively, unlabeled CCP6–8 (either the 402H or 402Y variant) or CCP19–20 recombinant proteins were adsorbed onto the wells of microtiter plates (Nunc Maxisorb, Kastrup, Denmark) at 1 μg/ml in 200 μl PBS for 16 h at 4°C. The ligand-binding characteristics of these immobilized proteins were determined using the biotinylated GAGs 2IS heparin, HA, C4S, C6S, and DS. Plate assays were carried out in 20 mM HEPES, 130 mM NaCl, 0.05% (v/v) Tween 20, pH 7.3, and the level of bound GAG determined as described earlier for the biotinylated proteins. In all these plate assays, data points were determined in duplicate and are reported as n = 4 from two independent experiments.

Details of the human eye and kidney tissues (without AMD or kidney disease, respectively) used in this study are provided in Table I. The staining of human tissue sections with fluorescently labeled proteins was carried out as described previously (12). In brief, tissue sections were incubated with chilled (−20°C) histologic grade acetone (Sigma-Aldrich) for 20 seconds before being thoroughly washed with PBS: every step of the protocol was preceded by a wash step, each time in PBS alone. These were then blocked with 1 mg/ml BSA, 1% (v/v) goat serum, and 0.1% (v/v) Triton X-100 in PBS for 1 h at room temperature. Tissue sections were incubated with the fluorescently labeled variants of flCFH, CCP6–8, or recombinant CCP19–20 (20 μg/ml of each in 100 μl blocking buffer) for 16 h at 4°C: “blank” (i.e., nonstained) tissue sections were incubated with blocking buffer alone. Cell nuclei were visualized using DAPI (0.3 μM final concentration) for 5 min at room temperature. Microscope coverslips were mounted using VectaShield (Vector Labs, Peterborough, U.K.). Endogenous expression of CFH was visualized using the anti-CFH mAb OX23 (35) at 10 μg/ml in blocking buffer. OX23 was incubated with tissue sections for 16 h at 4°C, washed with PBS, and detected using Alexa Fluor 488–conjugated goat anti-mouse IgG (Molecular Probes) at a 1:5000 dilution in PBS for 2 h at room temperature.

In competition experiments, the labeled variants of flCFH (either 402H or 402Y) were added to tissue sections in the presence of either a 1- or 10-fold molar excess of unlabeled CCP6–8 (i.e., the respective 402H and 402Y variant) and/or CCP19–20. In other experiments, kidney tissue was pretreated with specific HS/DS-degrading enzymes (i.e., a mixture of heparinase I, II, and III and chondroitin B lyase, all from Flavobacterium heparinum) (36, 37) before the application of either fluorescently labeled flCFH variant proteins or the anti-CFH Ab OX23; 20 U/ml of each enzyme (Sigma-Aldrich) was incubated with the tissue sections for 2 h at 37°C followed by thorough washing with PBS before blocking.

Images of fluorescently stained human tissue were collected on a snapshot widefield microscope (Olympus BX51) using a 20×/0.30 Plan Fln objective. Microscopy images were captured using a CoolSNAP ES camera (Photo-metrics) via MetaVue Software (Molecular Devices). To prevent bleed through of color from one channel to the next, we used specific bandpass filter sets for DAPI, FITC, and Texas Red. All images were processed and analyzed using ImageJ64 (version 1.40g; http://rsb.info.nih.gov/ij). Relative fluorescent intensities (i.e., of RPE and Bruch’s membrane in the eye samples and glomeruli in the kidney samples) were determined as described previously (12). In the case of human kidney tissues, the freehand tool (in ImageJ64) was used to select 10 individual glomeruli (designated in this study as the tissue within the Bowman’s space) from each donor; the 10 mean fluorescence intensity values were then averaged per donor to compensate for varying blood vessel number per glomeruli. For both eye and kidney tissues, the grayscale images for the red and green channels were overlaid so that an identical region was compared for both wavelengths. Tissue sections from five donor eyes and three donor kidneys were analyzed to obtain mean values (± SEM), and the statistical significance of protein binding was determined using Student t test (in KaleidaGraph v4.0), where p < 0.05 was considered to be significant.

The heparin-binding properties of the CCP19–20 region of CFH were compared with the 402H and 402Y variants of both flCFH and the CCP6–8 region of CFH by affinity chromatography on a 1 ml HiTrap Heparin column (GE Healthcare, Buckinghamshire, U.K.) at room temperature. The column was equilibrated in 5 ml PBS (OXOID). Protein (100 μg recombinant CCP6–8/CCP19–20 proteins or 200 μg flCFH) was loaded onto the column in a total volume of 5 ml PBS. The column was then washed with 4-column vol of PBS before bound protein was eluted with a linear salt gradient (over 20 min) of 0–100% buffer A (1 M NaCl in PBS) in PBS at a flow rate of 1 ml/min; fractions (1 ml) were collected throughout the protocol.

We examined the binding of the CCP6–8 and CCP19–20 regions of CFH to the GAG heparin, a highly sulfated form of HS often used as a model for the HS found ubiquitously in human tissues; both GAGs exhibit a diverse range of sequences (see Supplemental Fig. 1) (3, 38). Affinity chromatography using a HiTrap heparin column revealed that recombinant CCP19–20 protein eluted at a lower salt strength when compared with the 402H and 402Y variants, either in the context of a CCP6–8 construct or flCFH purified from homozygous individuals (Fig. 1). Thus, the CCP19–20 region has the weakest (salt-dependent) binding to heparin. However, it was apparent that flCFH demonstrated better binding to heparin than either CCP6–8 or CCP19–20. This indicates that both of these regions contribute to the strong binding of flCFH to heparin, which is consistent with a recent study (39). The CCP19–20 construct also demonstrated poorer binding to heparin compared with CCP6–8 in microtiter plate assays, where these experiments were conducted in physiological buffer conditions (Supplemental Fig. 2A). Similar analyses with other GAGs indicated that CCP19–20 does not interact significantly with C4S, C6S, DS, or HA (Supplemental Fig. 2B–E); of these GAGs, only DS has been found previously to interact with CCP6–8 (12, 20), and very weak binding has been reported for CCP19–20 with C6S (25).

Given the diversity of heparin/HS sequences (e.g., their wide range of sulfation patterns), we investigated the specificity of CCP19–20 (Fig. 2) to compare this with those of the 402H and 402Y CCP6–8 variants, which we have analyzed previously (20, 23); as before, this used selectively desulfated derivatives of heparin (32, 33) as models of HS. All the partially desulfated preparations of heparin demonstrated poorer binding to CCP19–20 than to the unmodified GAG (Fig. 2), demonstrating that the sulfate groups play a key role in this interaction. From these data, it might be anticipated that highly sulfated forms of HS would be required for optimal binding to the CCP19–20 region of CFH. Thus, the sulfate specificity of the CCP19–20 region is similar to the 402H variant, in that this form of CCP6–8 was more sensitive to removal of sulfate groups compared with 402Y (20, 23); the control data for the CCP6–8 402H and 402Y proteins has been included in this article for completeness and ease of comparison (Supplemental Fig. 3), where these were essentially the same as those obtained previously (20, 23).

However, the specificities of CCP19–20 (Fig. 2) and CCP6–8 402H (Supplemental Fig. 3B), although similar are not identical; for example, N-desulfation/re–N-acetylation of heparin has a smaller effect on CCP19–20 binding compared with the 402H variant, whereas the converse is seen for removal of 6-O-sulfates. On this basis, CCP19–20 would be expected to recognize different HS sequences compared with CCP6–8, which may affect the way CFH recognizes HS in human tissues.

Having demonstrated that the heparin-binding strengths/specificities of the CCP6–8 and CCP19–20 regions differed from one another in experiments using purified GAG preparations, we investigated their relative contributions to binding-site recognition in human tissues. Unlabeled CCP6–8 and CCP19–20 proteins were used as competitive inhibitors for the interaction of fluorescently labeled flCFH (i.e., 402H or 402Y) with human macula tissues from non-AMD donors (age range, 58–81 y; Table I), using a modification of the method described previously (12). CCP19–20 caused little or no inhibition of flCFH binding to the RPE or Bruch’s membrane even at 10-fold molar excess (Fig. 3A); statistically significant effects were seen only for competition of the 402Y variant corresponding to ∼11 and ∼20% reduction in binding to the RPE and Bruch’s membrane, respectively. However, when an equivalent experiment was performed with CCP6–8, a large and statistically significant decrease in binding of flCFH was observed on both RPE and Bruch’s membrane (Fig. 3B); for example, with 10-fold molar excess of CCP6–8, there was 77% inhibition of flCFH binding to the Bruch’s membrane for the 402H variant. These data clearly demonstrate that CCP6–8 make a larger contribution to binding of CFH than CCP19–20. When the CCP6–8 and CCP19–20 proteins were used together as competitors, this led to a greater level of inhibition (Fig. 3C) than that seen for CCP6–8 alone (Fig. 3B). The enhanced competition could be explained by CCP6–8 and CCP19–20 recognizing distinct HS structures within human macula. Overall, these results demonstrate that the CCP6–8 region plays a more prominent role in the interaction of CFH with human macula, although CCPs19–20 likely make some contribution to binding in the context of the flCFH.

Tissue binding experiments revealed that fluorescently labeled CCP19–20 was able to bind directly to the Bruch’s membrane (albeit weakly) and to the RPE (Fig. 4). The pattern of binding for CCP19–20 was similar to that for the 402H variant of CCP6–8, both of which bound less well to the Bruch’s membrane compared with the 402Y form of CCP6–8; all three constructs exhibited essentially identical levels of binding to the RPE (Fig. 4C, 4D). This may be because CCP19–20 and CCP6–8 402H require highly sulfated HS structures for optimal binding (based on the specificities described in Fig. 2 and Supplemental Fig. 3B), of which there may be relatively few in Bruch’s membrane; the broader specificity of 402Y (20, 23) (Supplemental Fig. 3A) would allow it to interact with a wide range of HS sequences within this specialized ECM. Enzyme pretreatment of human eye sections to remove HS caused a small but significant reduction (∼20%) in binding of CCP19–20 to the RPE that was not further reduced by the additional removal of DS (Fig. 5). In the case of the Bruch’s membrane, there was a 55% reduction in the binding of CCP19–20 when HS alone or DS/HS together were removed. This demonstrates that the CCP19–20 region of CFH recognizes HS (but not DS) in human macula (consistent with the GAG specificity described above). However, the effect of enzyme digestion was much less marked on CCP19–20 binding compared with CCP6–8, where removal of DS/HS reduced binding by ≥90% (12), indicating that CCP19–20 also recognize non-GAG ligands; staining with an anti-C3b Ab showed that there was little or no C3b in either the macula or kidney donor tissues used within this study (see Supplemental Fig. 4), indicating that C3b is unlikely to be responsible for the GAG-independent binding seen in this study for CFH. The results described earlier demonstrate that CCP6–8 has a greater role than CCP19–20 in mediating the interaction of CFH with sites in human macula, which is in agreement with previous data showing that flCFH and CCP6–8 have similar patterns of binding on Bruch’s membrane, RPE, and choroidal blood vessels (12).

Given the different HS-binding specificities of the CCP6–8 and CCP19–20 regions (see earlier) and that different tissues have HS with distinct compositions and sulfation patterns (3, 38, 40), we investigated whether the relative contributions of these GAG-binding regions were the same in human kidney; this tissue was chosen because of the association of mutations/polymorphisms in CCP19–20 with aHUS (5, 6). Staining with an anti-CFH Ab showed that the endogenous CFH protein is present throughout the glomerulus, being particularly prominent on the Bowman’s capsule and glomerular basement membrane (GBM; Fig. 6A). A similar staining pattern was apparent when fluorescently labeled CCP19–20 was used as a probe to detect its binding sites (Fig. 6B), whereas neither the 402H nor 402Y variants of CCP6–8 bound significantly to this tissue (Fig. 6C). This observation suggests that the CCP19–20 region plays a more prominent role in the binding of CFH to human glomeruli compared with CCP6–8. When kidney tissue was stained with fluorescently labeled flCFH proteins, both the 402H and 402Y variants bound to similar degrees throughout most of the glomerulus but with somewhat greater binding to the GBM (orange/yellow staining in Fig. 6D); the exception to this was the Bowman’s capsule, where there was a higher level of 402Y flCFH (green) staining. Pretreatment of the kidney tissue with GAG-degrading enzymes led to a ∼30–40% reduction in the fluorescence intensity observed in the glomeruli with the labeled 402H and 402Y flCFH variants (Fig. 7A) and an ∼60% reduction in CCP19–20 binding (Fig. 7B). These data indicate a likely role for HS in the binding of CFH to the glomeruli of human kidneys. However, the relatively small effect of GAG removal on CCP19–20 binding in the kidney (compared with >90% reduction of CCP6–8 binding in the macula) (12) provides evidence that CCP19–20 also interacts with non-GAG ligands in this tissue.

To further assess the relative contributions of CCP6–8 and CCP19–20 on CFH binding to human kidney, we performed competition experiments, essentially as described earlier for eye tissue. The CCP6–8 variant proteins caused no statistically significant inhibition of flCFH binding (Fig. 8B), whereas the CCP19–20 protein caused ∼60% inhibition when present at 10-fold molar excess (Fig. 8A). Competition of flCFH binding, using CCP6–8 and CC19–20 together, had no greater effect than CCP19–20 alone (Fig. 8C). Overall, these data demonstrate that, in contrast with human macula, CCP19–20 play the dominant role in the binding of CFH to sites within human kidney glomeruli, and that, at least in part, this is via its interaction with HS.

In this article, we demonstrate that the two GAG-binding regions of CFH both play a major role in the localization of the protein on host surfaces, but that the relative contribution of these regions is tissue specific; CCPs6–8 predominate in CFH binding to the human RPE and Bruch’s membrane, whereas CCPs19–20 mediate the interaction with human kidney glomeruli. These data correlate well with the evidence that the Y402H polymorphism (in CCP7) is associated with AMD (7–10, 27), whereas mutations/polymorphisms in CCPs19–20 are associated with kidney diseases such as aHUS (5, 6). Interestingly, we have also found that the mechanism underlying the tissue binding of the CCP6–8 and CCP19–20 regions differ, in that the former is mediated primarily via its interaction with sulfated GAGs, whereas a significant component of the latter is GAG independent. Hence, although digestion of HS/DS reduces the binding of CCP6–8 by ≥90% in Bruch’s membrane (12), removal of these GAGs leads to only a ∼50% reduction in the binding of CCP19–20 to both Bruch’s membrane and kidney tissue. However, from our analyses it is clear that ∼50% of CCP19–20 binding, to sites in these tissues, is mediated through its interaction with HS. In this regard, although the GAG-binding specificity of CCP19–20 was shown to differ from that of CCP6–8 in the details of their sulfation requirements, it was more similar to the 402H variant (compared with 402Y) in that it bound best to highly sulfated forms of heparin. Overall, these data suggest that the Bruch’s membrane contains only a relatively small number of appropriately sulfated HS sequences that are able to support the binding of CCP19–20 and the 402H form of CCP6–8. Furthermore, the results of our competition experiments indicate that CCPs6–8 and CCPs19–20 recognize different HS structures in the human eye, allowing high-avidity interaction of the full-length CFH protein by the use of two independent binding sites (i.e., even in the context of the 402H isoform of CFH).

Both AMD and aHUS are conditions thought to result from complement dysregulation brought about, at least in part, by abnormalities in CFH binding/function (6, 11, 12, 41). The results we describe in this article provide evidence that polymorphisms/mutations in different regions of CFH differentially affect its activity in different tissues, and thus predispose to different diseases. Consistent with this, the common Y402H polymorphism is associated with AMD and not aHUS, whereas the majority of amino acid substitutions in CCPs19–20 associated with aHUS have not been linked to AMD (5, 6). One exception to this is the highly penetrant, but rare, R1210C mutation in the C-terminal region of CCP20 that confers high risk for development of AMD (13). Although the effects of this mutation on GAG binding have not been tested in the context of the full-length protein, it does affect the binding of CCP8–20 to heparin (42, 43) and also has a major effect on the interaction of CFH with surface-bound C3b (44). Importantly, the R1210C mutant of CFH has also been shown to form a high m.w. covalent complex with human serum albumin (44), which would have a dramatic effect on the movement of CFH through the Bruch’s membrane (45), as well as masking/altering other functions (44). Thus, the association of R1210C with AMD may be independent of any effect on GAG binding. It should be noted that 60% of individuals who carry the R1210C CFH allele and have AMD also have a 402H CFH allele (13). Based on our description of the importance of CCPs6–8 in the interaction of CFH with sites in normal macula and the effect of the Y402H polymorphism on GAG binding (12, 20), the impaired R1210C function would not be compensated well by the CFH 402H produced from the other allele in these patients.

Another recent study has shown that a major CFH ligand is malondialdehyde (MDA), a common lipid peroxidation product found to accumulate in diseases such as AMD as a result of oxidative stress (46); for example, MDA is found on the surface of apoptotic/necrotic cells and blebs derived from these. Importantly, it was demonstrated that CFH binds to MDA via both CCPs 7 and 20, and that the Y402H polymorphism dramatically reduced binding. Furthermore, it has been proposed that MDA-mediated CFH recruitment inhibits complement activation in regions with drusen buildup, and that reduced binding of the 402H CFH variant would be proinflammatory (46). Thus, although the impaired binding of 402H to Bruch’s membrane could result in local complement dysregulation contributing to disease initiation, as we have suggested previously (4, 12), it could also be the case that the altered binding to ligands such as C-reactive protein (47–49), necrotic cells (46, 47), and oxidized lipids (46) could be important in driving AMD progression once drusen have started to form. Both the work described in this study and that from Weisman and colleagues (46) provide strong evidence that the Y402H polymorphism within the CCP6–8 region of CFH plays a major role in AMD pathogenesis because of the direct effect that this has on its binding functions.

A major finding of our study is that the CCP6–8 and CCP19–20 regions of CFH differentially mediate binding in different host tissues, which occurs, at least in part, through their recognition of distinct GAG populations. It is well established that GAG composition/sequence is dependent on tissue location (3, 38) and, for example, even within the different layers of the macula there is evidence for considerable diversity of GAG structure (40). In this regard, the 402H and 402Y variants (in the context of CCP6–8) recognize different GAG sequences in human Bruch’s membrane (12), and this study demonstrates that the specificity of CCP19–20 for HS in this ECM is different again. As noted earlier, the profound effect of the Y402H polymorphism on GAG binding could explain the functional basis of this polymorphism in AMD. In addition, the interaction of CCP19–20 with HS clearly mediates a significant proportion of CFH binding to kidney GBM. Consistent with this, expression of glomerular heparinase has been shown to increase in patients suffering from dense deposit disease (another kidney condition associated with CFH polymorphisms/mutations in CCP19–20), where the staining of HS in the GBM decreases accordingly (50).

Given the importance of correct surface recognition by CFH on complement regulation, it is not surprising that disruption of this event can result in diseases such as AMD and aHUS. The current widely held view that one main surface-recognition region exists in CFH, and that this resides in the CCP19–20 portion of the protein, may be simplistic and, in the case of certain tissues such as the human macula, incorrect. Knowledge that the contribution to host surveillance conferred by the two GAG-binding regions of CFH alters depending on the tissue context should permit organ-specific therapeutic strategies to readdress the balance of immune regulation in AMD and other diseases of complement dysregulation.

We thank Dr. Isaac Zambrano at the Manchester Royal Eye Hospital Eye Bank for access to human eye tissues and Drs. Günter Klima and Walter Rabl (Innsbruck Medical University) for providing human kidney material. We also thank Viranga Tilakaratna (University of Manchester) for preparing the biotinylated HA, Dr. Sheona Drummond (University of Manchester) for helpful discussions regarding fluorescent microscopy, and Dr. Robert Sim (University of Oxford) for providing the OX23 anti-CFH Ab.

The authors have no financial conflicts of interest.