Complement Factor H Binds to Denatured Rather than to Native Pentameric C-reactive Protein*

Binding of the complement regulatory protein, factor H, to C-reactive protein has been reported and implicated as the biological basis for association of the H402 polymorphic variant of factor H with macular degeneration. Published studies utilize solid-phase or fluid-phase binding assays to show that the factor H Y402 variant binds C-reactive protein more strongly than H402. Diminished binding of H402 variant to C-reactive protein in retinal drusen is posited to permit increased complement activation, driving inflammation and pathology. We used well validated native human C-reactive protein and pure factor H Y402H variants to test interactions. When factor H variants were incubated with C-reactive protein in the fluid phase at physiological concentrations, no association occurred. When C-reactive protein was immobilized on plastic, either non-specifically by adsorption in the presence of Ca2+ to maintain its native fold and pentameric subunit assembly or by specific Ca2+-dependent binding to immobilized natural ligands, no specific binding of either factor H variant from the fluid phase was observed. In contrast, both factor H variants reproducibly bound to C-reactive protein immobilized in the absence of Ca2+, conditions that destabilize the native fold and pentameric assembly. Both factor H variants strongly bound C-reactive protein that was denatured by heat treatment before immobilization, confirming interaction with denatured but not native C-reactive protein. We conclude that the reported binding of factor H to C-reactive protein results from denaturation of the C-reactive protein during immobilization. Differential binding to C-reactive protein, thus, does not explain association of the Y402H polymorphism with macular degeneration.

Binding of the complement regulatory protein, factor H, to C-reactive protein has been reported and implicated as the biological basis for association of the H402 polymorphic variant of factor H with macular degeneration. Published studies utilize solid-phase or fluid-phase binding assays to show that the factor H Y402 variant binds C-reactive protein more strongly than H402. Diminished binding of H402 variant to C-reactive protein in retinal drusen is posited to permit increased complement activation, driving inflammation and pathology. We used well validated native human C-reactive protein and pure factor H Y402H variants to test interactions. When factor H variants were incubated with C-reactive protein in the fluid phase at physiological concentrations, no association occurred. When C-reactive protein was immobilized on plastic, either non-specifically by adsorption in the presence of Ca 2؉ to maintain its native fold and pentameric subunit assembly or by specific Ca 2؉ -dependent binding to immobilized natural ligands, no specific binding of either factor H variant from the fluid phase was observed. In contrast, both factor H variants reproducibly bound to C-reactive protein immobilized in the absence of Ca 2؉ , conditions that destabilize the native fold and pentameric assembly. Both factor H variants strongly bound C-reactive protein that was denatured by heat treatment before immobilization, confirming interaction with denatured but not native C-reactive protein.
We conclude that the reported binding of factor H to C-reactive protein results from denaturation of the C-reactive protein during immobilization. Differential binding to C-reactive protein, thus, does not explain association of the Y402H polymorphism with macular degeneration.
Complement (C) 2 is the principle effector in the innate host defense against infection and also contributes to tissue damaging inflammatory reactions, processing and clearance of immune complexes and regulation of the immune response (1). Activation is controlled by a group of C regulatory proteins that includes factor H (fH), the plasma protein responsible for regulating the alternative pathway, an amplification loop that is pivotal in physiological and pathological activation of C (2). fH is a single-chain serum glycoprotein of 150 kDa consisting of a string of 20 homologous units, each of about 60 amino acids, called short consensus repeats (SCR). Deficiency of fH causes unregulated consumption of C with secondary deficiency of C3 (3). Although fH is a plasma protein, many of its activities occur on the cell membrane. Indeed, mutations in the carboxyl terminus of fH that affect membrane binding capacity cause defective C regulation at the membrane and are associated with the renal disease, atypical hemolytic uremic syndrome (4).
Within the fH molecule there are multiple binding sites for C3b, heparin, and several other ligands (5). An interaction of fH with the plasma pentraxin molecule, C-reactive protein (CRP), the classical major acute phase reactant, was first reported almost 25 years ago. Mold et al. (6) showed that CRP coating of pneumococci had divergent effects, increasing activation of the classical pathway while inhibiting the alternative pathway; this latter effect was fH-dependent and associated with a doubling of fH binding to the CRP-coated pneumococcus. The same group later provided direct evidence of fH binding to plasticimmobilized CRP, although when the assay format was reversed and fluid phase CRP was offered to immobilized fH, less than 1% of added CRP bound (7). In a surface plasmon resonance study, when CRP was covalently immobilized on the chip surface through amine coupling in the absence of Ca 2ϩ , calcium-dependent binding of the recombinant fH fragments comprising both SCR 7 and SCRs 8 -11 was seen (8).
The seventh SCR of fH is a "hotspot" for ligand interactions with reported binding to heparin, CRP, and streptococcal M-protein (9). When a fH polymorphism, resulting in a single amino acid change (Y402H) in SCR7 was implicated as a powerful risk factor for age-related macular degeneration (AMD) (10 -12), there was immediately intense interest in exploring differences in ligand binding between these polymorphic variants that might explain the disease association. Several groups have lately reported that the fH Y402 and H402 variants show differential binding affinity for CRP (13)(14)(15)(16)(17). Each used similar methods with CRP directly immobilized on a surface (ELISA plate or surface plasmon resonance chip) and either intact fH variants or recombinant fragments in the fluid phase. All concluded that both fH variants bound CRP and that the H402 variant bound less well than the Y402 variant. C proteins are present in drusen, the hallmark of the retinal pathology of AMD, but there is controversy about CRP, some reporting that it is absent (18), whereas others find it in the retinal layers and drusen deposits and more abundantly in H402 homozygotes than Y402 homozygotes (19). These observations have led to the suggestion that binding of fH to CRP in the retina limits C activation and the H402 variant because it binds less well and permits more C activation, thus, predisposing to disease. This has potentially important therapeutic implications, especially with the recent development of the first specific inhibitor of CRP ligand binding and pathogenic function in vivo (20).
As a first step toward evaluation of this suggestion, we have investigated the interaction between CRP and fH under conditions in which CRP retains its native fold and homopentameric assembly of protomers and which are, thus, most likely to be of physiological and pathophysiological relevance. Here we report that there is no interaction between fH and intact native CRP in the fluid phase or with CRP immobilized directly to plastic in the presence of Ca 2ϩ or immobilized via its specific Ca 2ϩ -dependent binding to its natural ligands. fH binds CRP only after the CRP has been denatured, either before immobilization or by coating onto plastic in the absence of Ca 2ϩ . We conclude that the reported binding of fH to CRP is likely caused by denaturation of CRP, which has long been known to result from its adsorption to plastic surfaces in the non-physiological absence of Ca 2ϩ (21,22) or by other structural modifications of CRP resulting from its covalent immobilization. Alternative mechanisms for the strong association of the fH H402 variant with AMD should be sought. Human C3 was either purchased from Calbiochem or purified in-house (23). C3b was generated by limited digestion with trypsin as previously described (24). The monoclonal anti-human C3b antibody, C3-30, has been reported previously (25).

Materials-Chemicals
Purification of Endogenous fH Y402 and H402 Variants from Human Plasma-Healthy volunteers homozygous for either the Y402 or H402 polymorphic variants of fH were identified by PCR analysis. Venous blood was taken into EDTA, leukocytes were harvested, genomic DNA was extracted using proteinase K/phenol/chloroform, and ethanol was precipitated. A fragment of 458 bp containing exon 9 of the CFH gene was PCRamplified from genomic DNA using specific primers derived from the 5Ј and 3Ј intronic sequences (forward, 5Ј-CCT TTG TTA GTA ACT TTA GTT CGT C-3Ј, and reverse, 5Ј-GGT CCA TTG GTA AAA CAA GG-3Ј) and the Platinum Blue PCR SuperMix polymerase kit (Invitrogen). Product was verified by electrophoresis on a 1.6% agarose gel and purified using QIAquick PCR purification kit (Qiagen, Crawley, UK). Direct sequencing of PCR products was performed using one of the amplification primers on an ABI Prism 3130xl Genetic Analyser (Applied Biosystems, Warrington, UK). fH was purified from plasma of individuals homozygous for either the Y402 or H402 variants by a sequential three-step fast protein liquid chromatography method at 4°C on Ä KTA Prime chromatography station (GE Healthcare). First, filtered plasma (100 ml) was applied to a 5-ml HiTrap column (GE Healthcare) to which 30 mg of mouse anti-human fH mAb 35H9 (generated in house) was coupled. Bound protein was eluted at low pH, and fractions containing fH (identified by ELISA) were pooled, dialyzed against phosphate-buffered saline (PBS), and applied to a 5-ml HiTrap-Heparin column (GE Healthcare) equilibrated with PBS. Bound proteins were eluted with 1 M NaCl in phosphate-buffered saline. Fractions containing fH were pooled and polished by gel filtration either on Superdex-200 preparative grade matrix in a XK16/70 column or a Superose 6 10/300 column (GE Healthcare). Purity was confirmed by SDS-PAGE, and preparations of fH-H402 and fH-Y402 were obtained without any detectable aggregates or contaminants, specifically free of fH-like-1 and fH-related proteins.
Purification and Denaturation of CRP-Intact, fully functional native CRP was isolated at Ͼ99.9% purity from malignant effusion fluids by DEAE anion exchange followed by calciumdependent binding to immobilized phosphoethanolamine and elution with free phosphocholine (26). The final product was separated from phosphocholine by the addition of EDTA to 10 mM final concentration followed by buffer exchange into 10 mM Tris, 140 mM NaCl, pH 8.0, and then the addition of CaCl 2 to 2 mM final concentration before concentration to 4 -6 g/liter. Pure CRP was stored frozen at Ϫ80°C. The CRP preparations were thoroughly characterized by overloaded SDS 8 -18% gradient PAGE, analytical gel filtration, ultracentrifugation, electrospray ionization mass spectrometry, crystallization, and solution of the x-ray three-dimensional structure (27). Concentration was measured in both Roche MIRA and Dade-Behring BNII latex enhanced immunoassays for CRP standardized on the World Health Organization International Reference Standard for CRP Immunoassay (85/506) (28). Binding to phosphocholine was measured in solution by isothermal titration calorimetry, binding by C1q was demonstrated using ELISA and surface plasmon resonance assays, and classical pathway C activation with pneumococcal C-polysaccharide as ligand was monitored by C3 binding and cleavage (29). These exhaustive tests confirmed that the material was pure, not aggregated, and retained full native structure, immunochemical integrity, and functional properties. CRP at 1 g/liter in TBS was completely and irreversibly denatured either by heating at 70°C for 1 h (30) or by incubation in 8 M urea, 10 mM EDTA for 1 h (21,31). After heating, denaturation was confirmed by SDS-PAGE using reduced amounts of SDS and without boiling the sample, as described (30). Aggregated material was removed by centrifugation, the remaining solution was filtered (0.2 m), and protein concentration (A 280 nm ) measured. Although there was ongoing aggregation of the residual soluble protein, it could be used to coat plates. After urea-chelation denaturation of CRP spiked with a tracer of 125 I-labeled CRP (26), the CRP was dialyzed into TBS and analyzed by size exclusion chromatography on Superdex 200 in the Åkta Explorer system (GE HealthCare), in comparison with the native starting material. The binding of native and denatured CRP to phosphocholine immobilized on Sepharose beads was also tested. Although more than 90% of radiolabeled CRP offered to phosphocholine-Sepharose absolutely showed calcium dependent binding, about 60% of denatured CRP became bound regardless of the presence or absence of calcium. Trypsin sensitivity of native and urea-chelation denatured CRP was compared in the presence (TBS) and absence (10 mM phosphate, 15 mM NaCl, pH 7.4) of Ca 2ϩ by incubation at 1 mg/ml with trypsin (20 g/ml) at 37°C. Samples of the digestion mixture were taken at intervals and analyzed by 8 -18% SDS-PAGE. Binding of C1q was sought by coating native or denatured CRP (10 g/ml in TBS) on 96-well Microtiter plates (Nunc MaxiSorp, Invitrogen), blocking with bovine serum albumin, then incubating with various dilutions of biotinylated C1q (made in-house). Wells were washed, and bound C1q was detected using avidin-peroxidase (Bio-Rad; 1:1000).
Production and Purification of Polyclonal Immunoglobulins-Monospecific polyclonal rabbit antisera against CRP and fH were raised by immunization with the respective antigens. Anti-fH antibodies were isolated by affinity chromatography on fH (5 mg) immobilized on a Hi-Trap N-hydroxysuccinimide column (GE Healthcare) and labeled with horseradish peroxidase using the EZ-Link Plus Activated Peroxidase kit (Pierce).
Interaction of fH and CRP in the Fluid Phase-Equal volumes of isolated fH (either Y402 or H402) and CRP, each in VBS 2ϩ at 1 g/liter, were mixed, incubated at 37°C for 2 h, and then analyzed by gel filtration chromatography on Superdex 200 (HR 10/30; GE Healthcare) equilibrated and eluted with VBS 2ϩ . The individual proteins were similarly analyzed for comparison. Absorbance at 280 nm of the eluted fractions was monitored.
Individual proteins or protein mixtures (each at 0.2 g/liter in VBS 2ϩ ; preincubated 2 h at 37°C as above) were also sedimented at different velocities (5,000, 7,000, 9,000, and 16,000 rpm) with absorbance monitored at 230, 250, and 280 nm, depending on the experiment, using short columns (80 l) on a Proteomelab XL-A analytical ultracentrifuge equipped with UV-VIS detection and 12 mm double-sector flow-through centerpiece (Beckman-Coulter Inc., High Wycombe, UK). The equilibrium data were analyzed globally using the program SEDPHAT to obtain the corresponding molar masses and interaction parameters (complex stoichiometry and affinity constants) (32).
Interaction of fH with CRP Directly or Indirectly Immobilized on a Solid Phase-Microtiter plates (96-well Nunc MaxiSorp, Invitrogen) were directly coated with CRP (10 mg/liter in TBS, HEPES-buffered saline, VBS 2ϩ , or ELISA buffer) at 37°C for 1 or 16 h. In some experiments coating was at 4°C and for up to FIGURE 1. Fluid-phase interaction between native CRP and fH by gel filtration. A, native pentameric CRP (pCRP) at 1 g/liter was preincubated at 37°C for either 1 h (shown) or 2 h (not shown; identical pattern) with either fH-Y402 (solid line) or fH-H402 (dashed line), each at 1 g/liter, then applied to an analytical gel filtration column. Each of the chromatograms shows only the individual native proteins; no higher molecular weight complexes were detected. The elution volumes from the mixture of Y402 and native CRP were 9.63 and 12.91 ml. For the mixture of H402 and pCRP the elution volumes were 9.48 and 12.88 ml. B, as controls, fH-Y402 (solid line) or fH-H402 (dashed line), each at 1 g/liter, were applied to the same analytical gel filtration column. The elution volumes were 9.70 and 9.63 ml, respectively, for fH Y402 and H402 proteins run separately. C, pCRP at 1 g/liter was applied to the same analytical gel filtration column. The elution volume was 12.84 ml. In all profiles, the y axis plots absorbance (milliabsorbance units (mAU)) at 280 nm, the x axis plots elution volume in ml. NOVEMBER 7, 2008 • VOLUME 283 • NUMBER 45 96 h. In others, heat-denatured or urea-denatured CRP was substituted for native CRP. Wells were then blocked by incubation with 1% w/v bovine serum albumin in VBS 2ϩ for 1 h at 37°C. Immobilization of CRP, native or denatured, under the various conditions was confirmed by incubating representative wells with rabbit anti-CRP antiserum (1:5000 in VBS 2ϩ ), and binding was detected using peroxidase-conjugated goat antirabbit IgG (Bio-Rad; 1:10000 in VBS 2ϩ) . In other wells serial dilutions in VBS 2ϩ of native fH Y402 and H402 variants were added and incubated. Binding of fH was detected by incubating with peroxidase-labeled affinity-purified rabbit anti-fH antibodies (1:5000 in VBS 2ϩ ). In some experiments CRP-coated wells were incubated with various dilutions of biotin-labeled C1q; binding of C1q was detected using avidin-peroxidase. Incubations were performed at 37°C for 1 h unless otherwise indicated. Between each step the wells were washed extensively with VBS 2ϩ containing 0.1% v/v Tween 20 (VBS 2ϩ -T). Wells treated only with coating buffer served as negative controls. Bound peroxidase was detected with OPD chromogenic substrate (Dako, Cambridgeshire, UK), and absorbance was read at 492 nm. All the experiments testing for binding of fH to CRP were repeated at least twice with triplicate samples. In some experiments, recombinant bacterially expressed CRP (Calbiochem) was substituted for natural human CRP. N-Hydroxysuccinimide-activated ELISA plates (Corning Costar, Bucks, UK) were derivatized by covalent attachment of 6-aminohexanoic acid as a molecular spacer from the plastic surface. Each well was incubated for 2 h at room temperature with 200 l of 6-aminohexanoic acid at 50 g/liter in 0.1 M NaHCO 3 , pH 9.6, and then washed with the same buffer before blocking residual active groups with 0.2 M Tris pH 8.0 (300 l/well) for 30 min. After further washing with water, different plates were then coupled with either phosphoethanolamine or p-aminophenylphosphocholine. These ligands were each dissolved at 10 g/liter in H 2 O, the pH was adjusted to 4.5 with HCl, and 150 l was added to each well followed by 50 l per well of a fresh 0.4 M solution of N-(3-dimethylaminopropyl)-NЈ-ethylcarbodiimide-HCl in H 2 O. Plates were then incubated on a shaker for 90 min at room temperature with frequent monitoring and adjustment of pH to 4.5 with 0.1 M NaOH or 0.1 M HCl. After a further 2 h, the reaction mixture was removed, and 0.2 M Tris, pH 6.8, was added for 30 min to block the remaining active groups. Plates were then washed in H 2 O, tightly sealed, and stored at 4°C until use.

Factor H Does Not Bind Native CRP
Optimal binding of CRP to immobilized phosphoethanolamine and phosphocholine was first determined by incubating dilutions of CRP in VBS 2ϩ in coated wells at 4°C, room temperature, or 37°C, then washing with VBS 2ϩ before detection with rabbit anti-CRP and peroxidase labeled anti-rabbit IgG as described above. Denatured CRP does not bind specifically to either phosphoethanolamine or phosphocholine (27) and was not tested in this assay. Plates optimally coated with CRP were washed in VBS 2ϩ , and dilutions of fH in VBS 2ϩ -bovine serum albumin was added to the wells and incubated for between 1 and 24 h at 37°C. Binding of fH was detected using anti-fH antibody as described above.
Binding of fluid phase native CRP by immobilized fH was tested using MaxiSorp plates that had been coated with native fH Y402 or H402, each at 10 mg/liter in ELISA buffer for 1 h at 37°C. After blocking with 1% bovine serum albumin in VBS 2ϩ , serial dilutions of CRP in VBS 2ϩ were added and incubated at 37°C for 1 h. Plates were washed in VBS 2ϩ -T, and bound CRP was detected by incubation with rabbit polyclonal anti-CRP antiserum (1:5000) followed by peroxidase-conjugated anti-rabbit IgG and developed as described above. As a positive control for this experiment, binding of fluid phase C3b by fH immobilized as described above was tested. Serial dilutions of C3b in VBS 2ϩ were added and incubated at 37°C for 1 h. Plates were washed in VBS 2ϩ -T, and bound C3b was detected by incubation with mouse monoclonal anti-C3b antibody (C3-30, 10 mg/liter) followed by peroxidase-conjugated anti-mouse IgG and developed as described above.

RESULTS
fH Does Not Bind CRP in the Fluid Phase-When CRP and fH, each at 1 g/liter, were incubated together in Ca 2ϩ -containing buffer then fractionated by size exclusion chromatography in the presence of Ca 2ϩ , no complex formation was evident. The elution profile in the fractionated mixture was identical to the sum of the profiles for the proteins fractionated in isolation (Fig. 1, A-C). No species of molecular weights greater than the FIGURE 3. Binding of fH variants to CRP immobilized on plastic. A, CRP was immobilized on plastic in Ca 2ϩ -containing buffers, VBS 2ϩ , TBS, or HEPES-buffered saline (HBS) for either 1 or 16 h at 37°C. Various doses of fH variants were added, and binding was detected using anti-fH antibody. Results shown are for fH-Y402, but identical results were obtained with fH-H402. No specific binding of either form of fH was detected at any dose and in any of the buffers tested. Each point is the mean of triplicate wells Ϯ S.D. B, CRP was immobilized on plastic in Ca 2ϩ -free ELISA buffer for either 1 or 16 h at 37°C. Various doses of fH variants were added, and binding was detected using anti-fH antibody. Both the fH-H402 and fH-Y402 variants bound efficiently to CRP immobilized in the absence of Ca 2ϩ , and binding was slightly increased when CRP had been coated for 24 h compared with 1 h. Each point is the mean of triplicate wells ϮS.D. C, CRP was denatured either by heat treatment or incubation with 8 M urea and then immobilized on plastic. Various doses of fH variants were added, and specific binding was demonstrated by using anti-fH antibody. fH bound specifically in a dose-dependent manner to CRP denatured by heat or urea; binding by each fH variant was similar. Each point is the mean of triplicate wells Ϯ S.D. uCRP, urea-denatured CRP. D, fH Y402H variants immobilized on the ELISA plate failed to bind native CRP (solid line) from the fluid phase even at high concentrations, whereas each fH variant bound to urea-denatured CRP in a dose-dependent manner (dashed line) (E, Y402; Ⅺ, H402). Each point is the mean of triplicate wells Ϯ S.D. E, fH Y402H variants immobilized on the ELISA plate efficiently bind C3b from the fluid phase (E, Y402; Ⅺ, H402). Each point is the mean of triplicate wells Ϯ S.D. F, native pCRP was captured onto plates by its specific calcium dependent binding to phosphoethanolamine with which the surfaces had been covalently coated. Various doses of fH variants were added, and binding was detected using anti-fH antibody. Neither fH variant showed any specific binding to pCRP (F, Y402; f, H402) above background (, no pCRP, data are shown for Y402; H402 was identical in control). Similar results were obtained when pCRP was captured on immobilized phosphocholine (not shown). Each point is the mean of triplicate wells ϮS.D. NOVEMBER 7, 2008 • VOLUME 283 • NUMBER 45

JOURNAL OF BIOLOGICAL CHEMISTRY 30455
individual free proteins were detected. Of note, there was a small but consistent difference in the elution of the fH H402Y variants, the Y402 variant eluting later (Fig. 1B).
When CRP was incubated with fH in the presence of Ca 2ϩ as above and the mixture was analyzed by analytical ultracentrifugation, there was again no evidence of complex formation.  2A shows the sedimentation profile at 9000 rpm and 280 nm for CRP and the Y402 and H402 variants of fH run alone and the respective mixtures of fH and CRP. The proteins were each individually gel filtered immediately before the coincubation step to ensure that only monomer was analyzed (Fig. 2, B and  C). The sedimentation equilibrium distributions of the individual proteins taken at several rotor speeds and wavelengths are best described by a model for single monomer species with molar masses of 124 Ϯ 7, 167 Ϯ 8, and 155 Ϯ 9 kDa for CRP and the Y402 and H402 variants of fH, respectively, values that, within the experimental uncertainty, are compatible with the expected size of the corresponding non-aggregated native proteins (CRP, 115 kDa; fH, 165 kDa). When CRP and either fH variant were mixed before analysis, the sedimentation equilibrium gradients were globally well fitted by a model in which the two proteins do not interact and sediment independently as monomers, with no evidence of higher molecular weight species suggestive of complexes between fH and CRP. As a control to confirm that sedimentation equilibrium analysis would detect interactions of fH with a protein ligand of comparable mass, interaction between fH variants and C3b was tested. Fig. 2D shows the sedimentation equilibrium gradients at 9000 rpm and 280 nm of the H402 and Y402 variants of fH and C3b alone and the respective mixtures. C3b was gel-filtered immediately before the coincubation step to ensure that only monomer was analyzed (Fig. 2E). Individual gradients are well fitted by a model for single species with average molar masses that correspond to the expected values of the protein monomers. As expected, the sedimentation distribution of the C3b/fH mixtures is steeper, indicating a molecular interaction. The inset in Fig. 2D illustrates the gradients for the individual species and also demonstrates the gradients that would result if the two proteins were combined but did not interact (dotted line) or combined and did interact (dashed line). It can be seen from this inset that the gradient resulting from the mixture of fH and C3b is intermediate between these two theoretical curves. Further analysis reveals that the gradient represents a mixture of the two individual monomers and a 1:1 complex of C3b and fH with a mass fraction of complex of ϳ0.4; this is compatible with an apparent dissociation constant of the binary complex in the M range.
fH Does Not Bind Native CRP Immobilized on Plastic-In view of the well known stabilization of CRP by Ca 2ϩ (21,22), the protein must always be stored and used in Ca 2ϩ -containing buffers. Immobilization onto plastic and all other steps in testing for binding of fH were, therefore, initially carried out in the presence of Ca 2ϩ . Binding of CRP to the plate was first confirmed by subsequent binding of anti-CRP antibodies and was similar for all buffers tested (see Fig. 4D and results not shown). There was no specific binding of either fH-Y402 or fH-H402 when they were incubated with CRP immobilized in any buffer in the presence of Ca 2ϩ (Fig. 3A, results for fH-Y402 shown). Different times and temperatures of CRP coating were tested, but no significant binding of fH was observed even when coating was extended to 96 h at 37°C. Because most published studies had used bacterial recombinant CRP rather than natural human CRP, the experiments were repeated with recombinant CRP from a commercial supplier immobilized in the presence of Ca 2ϩ . No binding of fH was detected (negative results not shown), eliminating the possibility that the different results were due to the source of CRP. We next coated plates with natural human CRP using the non-physiological alkaline and Ca 2ϩ -free buffers described in previous reports (8,(15)(16)(17). We confirmed using anti-CRP that a similar amount of CRP became immobilized as in the presence of Ca 2ϩ (Fig. 4D); however, each variant of fH now bound efficiently to the solid phase CRP, and this binding increased with increased duration of the coating step in Ca 2ϩ -free buffer (Fig. 3B). There was no significant difference in binding of the fH variants.
Incubation of CRP in the absence of Ca 2ϩ and specifically coating onto plastic (21,22) inevitably causes denaturation, and our findings strongly suggested that fH bound to denatured CRP. To confirm this we first denatured CRP by heat or urea treatment. Denaturation was confirmed in both cases by demonstrating aggressive spontaneous aggregation and precipitation in physiological buffers of the normally highly soluble CRP, fragmentation of the native pentameric molecular assembly into individual protomers (Fig. 4A), complete loss of specific calcium dependent binding to phosphocholine, acquisition of extreme sensitivity to trypsin digestion (Fig. 4B), and loss of capacity to bind C1q (Fig. 4C). The denatured CRP generated by either heat or urea was efficiently adsorbed to the plastic surface (Fig. 4D), and subsequent incubation with fH demonstrated specific binding that was dose-dependent and similar for each fH variant (Fig. 3C).
We confirmed the absence of any significant interaction between fH and native CRP by immobilizing the fH Y402H variants onto plates and testing for binding of native or denatured CRP from solution. No binding of native CRP was observed even at high CRP concentration, whereas urea-denatured CRP was bound by immobilized fH, with the Y402 variant consistently binding more (Fig. 3D). No binding of heat-dena-  3 ml), consistent with free protomers, M r ϳ 23,000. The y axis plots radioactivity (cpm), and the x axis plots elution volume in ml. B, pCRP and uCRP at 1 g/liter were incubated at 37°C with trypsin at 20 mg/liter and sampled at the times shown for analysis by SDS-PAGE. Left panel, in TBS, which contains calcium, there was no digestion of native CRP even after 6 h (lane 1, CRP alone; lanes 2-4, CRP ϩ trypsin at 0, 2, and 6 h). uCRP was rapidly cleaved, and no intact protomer was observed after 30 min (lane 5, denatured CRP alone; lanes 6 -10, denatured CRP ϩ trypsin at 0 min, 30 min, and 2, 4, and 6 h). Right panel, in hypotonic phosphate saline without calcium there was some very minor cleavage of native CRP (lane 1, native CRP alone; lanes 2-5, CRP ϩ trypsin at 0 min, 30 min, 2 h, and 6 h). The degradation of denatured CRP was even more extreme than when calcium was present (lane 6, denatured CRP alone; lanes 7-11, 0 min, 30 min, 2 h, 4 h, and 6 h). C, CRP, either native pentameric (pCRP, open circles) or heat-denatured (hCRP, closed circles) was immobilized on plastic in Ca 2ϩ -containing buffers for 1 h at 37°C. Plates were blocked, and various doses of purified, biotinylated C1q was added. After a further hour at 37°C, plates were washed, and bound C1q was detected using avidin-peroxidase. pCRP bound C1q in a dose-dependent manner, whereas heat-denatured CRP failed to bind C1q. D, CRP, either native (pCRP, open circles), heat-denatured (hCRP, closed squares), or urea-denatured (uCRP, open squares) was immobilized on plastic in Ca 2ϩ -containing buffers for 1 h at 37°C. Plates were blocked, and various dilutions of anti-CRP antibody were added. After a further hour at 37°C, plates were washed, and bound anti-CRP was detected using anti-rabbit IgG-peroxidase. tured CRP by immobilized fH was detected because of rapid irreversible precipitation of the denatured CRP during incubation. C3b was readily bound by immobilized fH in a dose-dependent manner, confirming the capacity of immobilized fH to bind ligands (Fig. 3E).
fH Does Not Bind Native CRP Captured on Natural Ligands-Physiological immobilization of CRP in vivo takes place through binding of CRP to its natural ligands. We replicated this by allowing native CRP to undergo physiological Ca 2ϩ dependent binding to phosphoethanolamine or phosphocholine covalently attached to plastic. Binding was confirmed using anti-CRP. When fluid phase fH variants were incubated on CRP immobilized in this way there was no specific binding (Fig. 3F).

DISCUSSION
The demonstration of a strong association between a common polymorphism in the alternative pathway regulator fH and AMD in multiple cohorts has placed the C system center stage in the search for the cause of this common and disabling condition (10 -12, 33). Attention has focused on defining the functional differences between the fH-Y402H variants that might influence C regulation and predispose to disease (13)(14)(15)(16). fH binds multiple ligands, and differences between the variants in ligand binding properties have been sought. SCR7, the location of the polymorphism, is a ligand binding hotspot in fH that has been implicated in binding to CRP, heparin, and the streptococcal M protein (9).
Retinal deposition of CRP has been reported in AMD, particularly in association with drusen deposits (19,34). CRP bound to its natural ligands binds C1q and is a potent activator of the classical C pathway (35); CRP in drusen may, therefore, drive C activation and inflammation. Recent reports that the AMD-linked fH-H402 variant binds to CRP with lower affinity than fH-Y402 (13)(14)(15)36) have attracted enormous interest because they suggest a mechanism for the association. Binding of fH to CRP in the retina could, by regulating the alternative pathway amplification loop, restrict CRP-triggered local C activation. In individuals homozygous for fH-H402, reduced binding of fH-H402 compared with fH-Y402 would result in less efficient control of alternative pathway amplification and, thus, permit potentially pathogenic C-mediated inflammation and tissue damage. The hypothesis implies that manipulation of C activation and/or inhibition of ligand binding and C activation by CRP might be of therapeutic benefit in AMD. We, therefore, investigated binding of the fH variants to CRP to test effects on C activation and the possible therapeutic role of inhibition of CRP binding (20).
Each of the five identical protomers in the native pentameric CRP molecule has a shallow ligand binding pocket within which are coordinated the two Ca 2ϩ ions essential for all physiological ligand binding by CRP (27). It has been known for 25 years that these Ca 2ϩ ions stabilize the native pentameric assembly of CRP and that, in the absence of Ca 2ϩ , denaturation of CRP by urea or even heating at 63°C for just 5 min generates free protomers that express a neo-epitope that is absent from the native molecule even when it is immobilized by binding to a solid phase natural ligand or captured by anti-CRP antibodies (21,22). The neo-epitope of denatured CRP has also long been known to be generated by immobilization of CRP on plastic surfaces (22,37). Denatured CRP does not show the normal Ca 2ϩ -dependent ligand binding of native CRP and rapidly irreversibly aggregates at physiological ionic strength. Furthermore, denatured CRP, whether aggregated and precipitated or kept in solution/suspension at sub-physiological ionic strength, is highly susceptible to proteolysis, in marked contrast to native CRP, which is extremely protease resistant in the presence of Ca 2ϩ (38), as we show again here with trypsin (Fig. 4B). Ligation of Ca 2ϩ by native CRP maintains the organization of the loops of the binding pocket, prevents their proteolytic cleavage, and ensures both the native fold of the protomer (27,39) and the intact pentameric structure.
We first tested the interaction of fH variants and CRP in the fluid phase in Ca 2ϩ -containing physiological buffers. No specific interaction was observed when mixtures were analyzed either by gel filtration or ultracentrifugation (Figs. 1 and 2). These findings effectively rule out significant binding of fH to CRP in the fluid phase under physiological conditions. We next tested binding of fH to CRP immobilized from Ca 2ϩ -containing buffers both by direct non-covalent adsorption to plastic and by the physiological Ca 2ϩ -dependent capture of CRP on its natural ligands, phosphocholine or phosphoethanolamine, covalently immobilized on plastic. Although we confirmed by binding of anti-CRP antibodies that CRP efficiently bound in both sets of experiments (Fig. 4D), we could not detect any specific binding of either variant of fH to the solid phase CRP (Fig. 3, A and F). Others have reported specific binding of fH to CRP immobilized either on ELISA plates or Biacore chips (8,9,(13)(14)(15)(16)(17)(18)(19). In some of these reports it is clear that CRP immobilization and/or blocking steps were performed in Ca 2ϩ -free buffers (8,(15)(16)(17); for the others the coating conditions were not given. Denaturation of CRP is likely in these conditions, which in addition to disrupting the pentameric assembly and native fold, also produces non-physiological forms of CRP which display promiscuous nonspecific binding to a wide variety of different ligands (30,31,40). When we coated plates with CRP in the absence of Ca 2ϩ , similar amounts of CRP were immobilized as from the physiological Ca 2ϩ -containing buffers, but substantial and similar binding of both fH variants was then observed and was related to the duration of exposure of the CRP to plastic in the absence of Ca 2ϩ (Fig. 3B). To confirm that the observed binding was to denatured CRP, the native protein was denatured by heat treatment or urea as previously described (30,31). Denaturation was confirmed by demonstrating the appearance of protomer (Fig. 4A), loss of C1q binding capacity (Fig. 4B), and acquisition of trypsin sensitivity (Fig. 4C). When plates were coated with denatured CRP, fH bound in a dose-dependent manner (Fig. 3C).
A recent report by Bíró et al. (31) confirmed the original study of Potempa (22) that CRP immobilized on plastic or on Biacore chips in the absence of Ca 2ϩ underwent structural changes that included the emergence of neoepitopes specific for the denatured protein (31). These authors also showed that fH bound to urea-denatured CRP that had been immobilized but not to CRP captured by its physiological calcium-dependent binding to solid phase phosphocholine, results entirely consistent with our present findings.
It is not clear whether the interaction between fH and denatured CRP involves recognition of the damaged CRP by fH or nonspecific binding of denatured CRP to fH. However, because Ca 2ϩ is always present in the extracellular milieu in vivo, the properties of denatured CRP are unlikely to be of physiological or pathophysiological significance. The native pentameric structure of CRP is very stable in the presence of Ca 2ϩ . Denaturation in vitro requires extreme pH, heat, or harsh chaotropes such as 8 M urea, all in the absence of calcium (21,30). It is unlikely that CRP is exposed to a significant degree to comparable conditions in vivo. If any CRP did become denatured in vivo, it would be readily degraded by proteolysis as we show here (Fig. 4B) and be unlikely to activate C (Fig. 4C).
The possibility remains that denatured CRP is present in drusen and interacts with fH; however, in multiple experiments we did not find a significant difference in binding of the two variants to denatured CRP. We, therefore, conclude that differential binding to CRP, whether native or denatured, does not explain the association of this polymorphism with pathology in AMD. SCR7 binds several other ligands (9,36), and more attention should now be paid to these. M protein, a surface component of group A streptococci, captures fH through SCR7 to resist complement. Several groups have reported that fH Y402 binds M protein better than fH H402, suggesting that expression of H402 might aid bacterial clearance (16,36,41). Using recombinant SCR6 -8 constructs containing either H or Y at position 402, Clark et al. (42) showed an increased binding of the H402 variant to some but not all of a panel of sulfated heparins. Co-crystallization of SCR7 H402 with a sulfated glycosaminoglycan analogue showed that the H402 directly coordinated the ligand (43,44). Whether differential binding of these other ligands influences risk in AMD remains to be determined. One suggestion is that efficient binding of fH-H402 to sulfated glycosaminoglycans in drusen drives pathology by reducing C activation and the resultant clearance of the debris (44).
Although our observations do not solve the conundrum of how the common fH polymorphism predisposes to disease, they robustly question involvement of fH binding to CRP and regulation of C activation thereby. Physiologically and pathophysiologically relevant and appropriate analysis of C activation and regulation in AMD patients and in relevant models is now required to understand whether C activation is beneficial or harmful in the disease process and how changes in fH may affect the balance between activation and regulation to predispose to disease.