FH (Factor H) with 20 SCR (short complement regulator) domains is a major serum regulator of complement, and genetic defects in this are associated with inflammatory diseases. Heparan sulfate is a cell-surface glycosaminoglycan composed of sulfated S-domains and unsulfated NA-domains. To elucidate the molecular mechanism of binding of FH to glycosaminoglycans, we performed ultracentrifugation, X-ray scattering and surface plasmon resonance with FH and glycosaminoglycan fragments. Ultracentrifugation showed that FH formed up to 63% of well-defined oligomers with purified heparin fragments (equivalent to S-domains), and indicated a dissociation constant Kd of approximately 0.5 μM. Unchanged FH structures that are bivalently cross-linked at SCR-7 and SCR-20 with heparin explained the sedimentation coefficients of the FH–heparin oligomers. The X-ray radius of gyration, RG, of FH in the presence of heparin fragments 18–36 monosaccharide units long increased significantly from 10.4 to 11.7 nm, and the maximum lengths of FH increased from 35 to 40 nm, confirming that large compact oligomers had formed. Surface plasmon resonance of immobilized heparin with full-length FH gave Kd values of 1–3 μM, and similar but weaker Kd values of 4–20 μM for the SCR-6/8 and SCR-16/20 fragments, confirming co-operativity between the two binding sites. The use of minimally-sulfated heparan sulfate fragments that correspond largely to NA-domains showed much weaker binding, proving the importance of S-domains for this interaction. This bivalent and co-operative model of FH binding to heparan sulfate provides novel insights on the immune function of FH at host cell surfaces.
The alternative pathway, one of the three activation pathways of the complement system of innate immunity, is a natural defence system for the removal of foreign pathogens in vertebrate tissues [1–4]. The activation of C3, the central component of complement system present at 1 mg/ml in serum, to C3b through the cleavage of the small anaphylatoxin C3a from C3 initiates the alternative pathway. Complement FH (Factor H), a 154 kDa plasma glycoprotein present at 0.235–0.81 mg/ml in serum, is the major regulator of the alternative pathway and enables host and non-host surfaces to be distinguished by inhibiting C3b activity at host cells. FH performs this regulatory function by acting on C3b in three ways: (i) acting as a cofactor in the Factor I-mediated proteolytic inactivation of C3b into iC3b [5,6]; (ii) accelerating the spontaneous decay of the C3 convertase of the alternative pathway ; and (iii) blocking Factor B from binding to C3b to form the C3 convertase (C3bBb) .
FH consists of 20 SCR (short complement regulator) domains, each of which contains aproximately 61 residues. The SCR domains constitute the most abundant domain type in complement, and are also known as short consensus repeats or complement control protein domains . FH has three binding sites for C3b, where SCR-1/4 binds to intact C3b, SCR-6/10 binds to the C3c region of C3b and SCR-16/20 binds to the C3d region of C3b [9,10]. FH inhibits surface-bound C3b activity on host cell surfaces by recognizing polyanionic structures such as sialic acid, HS (heparan sulfate) and dermatan sulfate, all of which are found on host cell surfaces [11,12]. HS is formed from short sulfated S-domain regions that are centrally composed of IdoA2S-GlcNS6S disaccharides and longer unsulfated NA-domain regions composed of GlcA-GlcNAc disaccharides . FH has binding sites for heparin within each of SCR-7 and SCR-20. A third binding site for heparin was originally located at SCR-13, but this was reassigned to SCR-9, then the existence of this SCR-9 site was questioned [14–17]. FH was the first protein to be genetically linked to AMD (age-related macular degeneration), a common disease leading to blindness in the elderly. A biological indicator for the development of AMD are drusen, extracellular deposits found within Bruch's membrane . Drusen contain oxidized lipids, aggregated proteins including FH, and glycosaminoglycans [19,20]. A Y402H polymorphism at FH SCR-7 has been associated with many cases of AMD [21–24]. FH is also genetically linked with aHUS (atypical haemolytic uraemic syndrome), which leads to renal failure in young children and adults, with many aHUS-associated mutations occurring in SCR-19/20 .
Solution structural studies of full-length FH and heparin and HS fragments of sizes up to 36 saccharide rings (dp36, where dp stands for degree of polymerization, or the number of saccharide rings) have resulted in molecular structures for all three at medium resolution [26–28]. Although crystal structures of small FH SCR fragments, and heparin and HS fragments bound to other proteins, are known , the large size and glycosylation of full-length FH have precluded its crystallization to date, as well as its complexes with heparin or HS. Recent crystal structures for C3b and C3d in association with its ligands including FH do not consider heparin (except for SCR-6/8 in association with a sucrose octasulfate disaccharide), and dissociation constant (Kd) values for these FH–C3b/C3d interactions are generally known [26,30,31]. In solution, FH has a partially folded-back SCR domain structure (Figure 1a), and exists as 85% monomer, together with 15% of dimer, trimer and higher oligomers that are formed from reversible and irreversible FH self-associations in physiological concentration ranges . In the present study, we make use of six heparin fragments dp6–dp36, that correspond structurally to S-domains of HS, and show extended and semi-rigid structures (Figure 1b) . We have also prepared HS dp6–dp24 structures, corresponding to the NA-domains of HS with remnants of the partially sulfated transitional domains between NA- and S-domains. Their structures are longer and more bent than those for heparin (Figure 1c) . The molecular basis of FH–HS binding is not known. However, it is clear that structures as large as heparin dp36 are unable to bind simultaneously to both heparin sites within one FH molecule unless FH becomes significantly more compact to bring these two sites close together (Figure 1). The Kd values for the FH–HS interaction are also not well known. Previously FH tetramers were reported in the presence of dextran sulfate or unfractionated heparin by AUC (analytical ultracentrifugation) .
Molecular models for FH, heparin dp36 and HS dp16
In the present study, to clarify the molecular basis of the FH–HS interaction, we performed binding studies of each of the S- and NA-domains of HS to FH by AUC, X-ray scattering and SPR (surface plasmon resonance). For the S-domains, large structurally well-defined FH–heparin oligomers were formed for only the larger heparin fragments, and modelling showed that these are explained in terms of bivalent complexes . No FH–heparin oligomers were seen with the shorter heparin fragments. Bivalency was confirmed using the SCR-6/8 and SCR-16/20 fragments of FH. The reduced interaction of the minimally sulfated NA-domains of HS with FH showed the importance of S-domains in HS for FH binding. We show that FH does not rearrange itself into a more compact conformation after heparin binding, and that heparin binds independently with micromolar Kd values to two major FH sites, but more strongly when binding bivalently. Our models imply a co-operative molecular mechanism for FH binding to HS-coated surfaces, where double binding promotes stronger interactions. The implications of bivalency and co-operativity for the regulatory function of FH are discussed.
Purification of FH and the heparin and HS fragments
Native FH was purified from pooled outdated human plasma using monoclonal affinity chromatography with an MRC-OX23 Sepharose column . Bound FH was eluted with 3 M MgCl2 (pH 6.9), dialysed into Hepes buffer [10 mM Hepes and 137 mM NaCl (pH 7.4)] in the presence of 0.5 mM EDTA to remove Mg2+ and then passed through a HiTrap Protein G HP column to remove contaminant IgG. Non-specific aggregates and human serum albumin were removed using Superose 6 gel filtration. The FH concentration was determined from an absorption coefficient (1%, 1 cm, 280 nm) of 16.2 . FH was dialysed into Hepes buffer for scattering and ultracentrifugation experiments, and its integrity was routinely checked by SDS/PAGE before and after experiments. SCR-6/8 was expressed in Escherichia coli and purified following previous procedures (A. Miller and S.J. Perkins, unpublished work) . SCR-16/20 was expressed as a His6-tagged product in Pichia pastoris and purified using nickel-affinity and size-exclusion chromatography .
The dp6–dp36 fragments of heparin were prepared from unfractionated bovine lung heparin using heparinase I digests as described previously [26,36]. The monodispersity and purity of these six dp6–dp36 fragments was shown by the single peaks seen by AUC c(s) size distribution analyses (see below) . Heparin concentrations were measured by removing the volatile buffer (2% ammonium bicarbonate) by rotary evaporation three times at 50°C following the addition of deionized water each time, then weighing the dried product after evaporation by freeze-drying. The dp6–dp24 fragments of HS were prepared in a similar manner, this time starting from unfractionated HS, where exhaustive heparinase I digestion was used to minimize the content of fully sulfated sequences .
AUC sedimentation velocity data for FH mixtures with heparin and HS
Sedimentation velocity data were obtained using two Beckman XL-I analytical ultracentrifuges (Beckman Coulter) equipped with both absorbance and interference optics . Experiments were carried out in an eight-hole An-Ti50 rotor or four-hole An-Ti60 rotor with standard double-sector cells with sapphire windows, with column heights of 12 mm at 20°C using interference optics. Sedimentation velocity data were collected with rotor speeds of 40000, 50000 and 60000 rev./min using interference optics. Experiments with FH and six 1:1 mixtures with heparin dp6–dp36 were performed at 1.08 mg/ml FH in 10 mM Hepes and 137 mM NaCl (pH 7.4). Experiments with FH mixtures were also performed with heparin dp6 and dp12 in molar ratios that ranged from 1:0.5 to 1:8 at 1.0–1.12 mg/ml FH. Experiments with FH and four 1:1 mixtures with HS (dp6, dp12, dp18 and dp24) were also performed in this Hepes buffer. The buffer density of 1.00480 g/ml was measured at 20°C using an Anton-Paar DMA5000 density meter. Partial specific volumes were calculated for FH and its 1:1 heparin mixtures to be 0.717 ml/g for FH, 0.714 ml/g for FH and dp6, 0.711 ml/g for FH and dp12, 0.708 ml/g for FH and dp18, 0.705 ml/g for FH and dp24, 0.702 ml/g for FH and dp30, and 0.699 ml/g for FH and dp36 .
The continuous c(s) analysis method was used to determine the sedimentation coefficients s20,w using SEDFIT software (version 9.4) [39,40]. In the c(s) analysis, the sedimentation boundaries are fitted using the Lamm equation, the algorithm for which assumes that all species have the same frictional ratio f/fo in each fit. The final SEDFIT analyses used a fixed resolution of 200 and a fixed frictional ratio f/fo, of 1.78 for FH and all the heparin and HS mixtures. The c(s) fits were optimized by floating the meniscus and cell bottom when required, and the baseline and cell bottom fixed until the overall root mean square deviations and visual appearance of the fits were satisfactory. The percentage of FH oligomers was derived using the c(s) integration function.
X-ray scattering data for FH mixtures with heparin and HS
X-ray scattering data for mixtures of FH with the six heparin fragments dp6–dp36 were acquired during a single-beam session at the ESRF (European Synchrotron Radiation Facility) at Grenoble, France. This session followed trial experiments in two earlier sessions that employed 0.4–1.0 mg/ml FH and fewer heparin fragments. The ID02 beamline was operated in four-bunch mode with a ring energy of 6.0 GeV . Storage ring currents ranged from 40 to 43 mA. Data were acquired using an improved fibre optical-coupled high-sensitivity and dynamic-range CCD (charge-coupled device) detector (FReLoN) with a smaller beamstop. The sample-to-detector distance was 2 m. Data were collected for FH in mixtures with a 1:1 molar ratio of heparin in 10 mM Hepes and 137 mM NaCl (pH 7.4). The FH concentration was 0.47 mg/ml. Samples were measured in a flow cell which moved the sample continuously during beam exposure in ten time frames with different exposure times between 0.1 and 0.5 s to avoid radiation damage effects. Exposure times were optimized using on-line checks during acquisitions to verify the absence of radiation damage, after which the frames were averaged. Similar X-ray scattering for native FH at 0.45 mg/ml in 1:1 ratios with eight HS fragments (dp6, dp8, dp10, dp12, dp14, dp16, dp18 and dp24) was performed in a further beam session, this time using storage ring currents of 69–76 mA.
Guinier analyses give the radius of gyration, RG, which characterizes the degree of structural elongation in solution if the internal inhomogeneity of scattering within the macromolecules has no effect. Guinier plots at low Q values (where Q = 4πsinθ/λ; 2θ is the scattering angle; λ is the wavelength) gives the RG and the forward scattering at zero angle I(0)  (eqn 1):
This expression is usually valid in a Q·RG range up to 1.5. If the structure is elongated (i.e. rod-shaped), the radius of gyration of the cross-sectional structure, RXS, and the mean cross-sectional intensity at zero angle [I(Q)·Q]Q→0 parameters are obtained from (eqn 2):
The RG and RXS analyses were performed using an interactive PERL script program SCTPL7 (J.T. Eaton and S.J. Perkins, unpublished work) on Silicon Graphics OCTANE workstations. Indirect Fourier transformation of the full scattering curve I(Q) in reciprocal space gives the distance distribution function P(r) in real space. This yields the maximum dimension of the macromolecule L and its most commonly occurring distance vector M in real space (eqn 3):
The transformation was carried out using GNOM software . For this, the X-ray curves contained 192–215 intensity data points in the Q range between 0.08 nm−1 extending up to between 1.3 and 2.2 nm−1 for the FH mixtures with heparin dp6–dp36. The X-ray curves for the FH mixtures with HS contained 211–360 data points in the Q range between 0.08 and 1.7 nm−1.
SPR data of FH binding to heparin and HS
Heparin–biotin and HS–biotin conjugates were prepared using a previously described method . In summary, 3.4 mg of unfractionated heparin, heparin dp32, heparin dp24, unfractionated HS, HS dp18 and HS dp24 were each dissolved separately in 340 μl of 50 mM sodium bicarbonate (pH 8.3). For biotinylation, these heparin and HS solutions were incubated with sulfo-NHS-LC-biotin [succinimidyl-6-(biotinamido)hexanoate] (2.4 mg in 10 μl of dimethylformamide) for 2 h at 4°C. Unreacted biotin was removed by dialysis of the reaction mixture against 50 mM sodium bicarbonate (pH 8.3) buffer overnight using a membrane filter (3.5 kDa cutoff). The biotinylated heparin and HS samples were freeze-dried and stored at −20°C. In order to remove non-specifically-bound contaminants, the streptavidin sensor chip was manually pretreated with two injections (10 μl each) of 50 mM NaOH in 1 M NaCl. This was followed by two injections (10 μl each) of biotinylated heparin, biotinylated HS and their biotinylated fragments [0.5–2 μg/ml in 10 mM Hepes (pH 7.4), 150 mM NaCl, 3 mM EDTA and 0.005% surfactant P20] in flow cell 2. Unfractionated heparin and HS (parent material) were each immobilized to approximately 41 and 46 RUs (resonance units) respectively. Heparin dp32 and dp24 were each immobilized to 13 and 10 RUs respectively, whereas HS dp18 and HS dp24 were each immobilized to 10 RUs. Because binding of FH SCR-1/20 was not observed with 10 RU-immobilized HS dp24, more HS dp24 was immobilized to reach 30 and 300 RUs.
For SPR experiments, each of full-length FH, SCR-6/8 and SCR-16/20 were injected over the heparin-immobilized sensor chip in the same Hepes buffer used for scattering and ultracentrifugation, but with 0.005% P20 surfactant added. The flow-rate was 30 μl/min. Six FH concentrations of 0.1, 0.5, 1, 2, 4 and 6 μM were used with immobilized heparin dp32. SCR-16/20 was injected at eight concentrations of 1, 2, 4, 8, 12, 16, 24 and 30 μM over the heparin dp32-immobilized sensor chip. On the same heparin dp32-immobilized sensor chip, SCR-6/8 was passed at six concentrations of 4, 8, 12, 16, 20 and 24 μM. Seven FH concentrations of 0.5, 1, 2, 3, 4, 6 and 8 μM were used with immobilized heparin dp24. Five FH concentrations of 0.1, 0.2, 0.5, 1 and 3 μM were used with immobilized unfractionated heparin. Six FH concentrations of 1, 1.5, 2, 2.5, 3 and 4 μM were used with immobilized unfractionated HS. For the binding analyses of FH with HS dp18 and HS dp24, only one concentration of 8 μM was used. The binding response (measured in RUs) was recorded over a 150 s association phase and 600 s dissociation phase.
FH–heparin dp36 oligomer modelling
The most recent structural model for full-length FH is that of the best-fit X-ray solution scattering structure (PDB code 3GAV) . This structure was obtained by fitting 20 individual SCR models joined by 19 variable linkers to the scattering curve of pooled FH. Models to account for the s20,w values of the FH oligomers formed with heparin dp36 were created using this best-fit FH model (PDB code 3GAV) together with that for heparin dp36 (PDB code 3IRL) . Using Discovery Studio molecular graphics software (version 2.1) (Accelrys), low-resolution FH–heparin models were created manually by positioning one end of the dp36 structure in close proximity to FH SCR-7 or SCR-20 in the first FH structure and the other end of the dp36 structure in close proximity to FH SCR-7 or SCR-20 in the next FH structure. No information is available on individual residue contacts in SCR-7 or SCR-20 because the HS-binding sites are not known. This procedure was used to create ranges of FH–heparin dp36 structures based on FH dimer, trimer, tetramer and pentamer associations. To check for self-consistency of these models, two other best-fit FH structures with different SCR-7 and SCR-20 separations were used (R. Nan, A.I. Okemefuna and S.J. Perkins, unpublished work). One came from fitting five multi-SCR fragments to the same FH scattering curve; the other fitted the 20 SCR models to the scattering curve of homozygous wild-type FH. The sedimentation coefficients s020,w for all of these models were calculated directly from molecular structures using the HYDROPRO shell modelling program . The default value of 0.31 nm for the atomic element radius for all atoms was used to represent the hydration shell surrounding FH and heparin.
Sedimentation velocity data for FH with heparin and HS
Biophysical studies were performed with full-length native FH purified from the same pooled plasma stock used previously . This FH gave a homogenous peak in gel-filtration chromatography and a single clean band on SDS/PAGE. AUC experiments were carried out with FH concentrations of 1.08 mg/ml (7.0 μM), whereas scattering experiments were carried out at 0.47 mg/ml (3.0 μM), both of which are comparable with physiological FH concentrations in plasma between 0.235 and 0.81 mg/ml . Given the ionic nature of the FH–glycosaminoglycan interactions, all experiments were performed in buffers with 137 mM NaCl in order to correspond to near-physiological conditions in plasma .
AUC reports on macromolecular sizes and structures in solution by following the sedimentation behaviour under a high centrifugal force . AUC is able to resolve different FH–heparin complexes as individual peaks in sedimentation coefficient distribution analyses c(s). Sedimentation velocity data were collected for 7.0 μM FH with heparin dp6–dp36 at 1:1 molar ratios. This higher FH concentration made more visible the c(s) peaks from the FH oligomers (see below).
Excellent boundary fits were obtained (Figures 2a–2c). More rapid sedimentation of FH with dp36 was visible compared with that with dp18 and dp6. The c(s) analyses also showed that FH sedimented differently with different heparin fragments (Figure 2d). Because free heparin dp6–dp36 sediments at 1.2–1.9 S , the FH peaks between 5 and 22 S showed no overlap with unbound heparin, and provided unequivocal views of the FH–heparin complexes. For all of the FH–heparin dp6–dp36 mixtures, the c(s) plots showed that monomeric FH (peak 1) sedimented at an s20,w value of 5.66±0.09 S, which is unchanged from that of 5.48±0.02 S for free FH at 0.87 mg/ml . In order to validate the c(s) analysis, mass calculations with peak 1 showed that this corresponded to a molecular mass of 132±4 kDa, this being similar to one of 134±4 kDa for free FH at 0.87 mg/ml. These values agreed well with previous values of 5.65±0.05 S and 142±2 kDa for free FH and the sequence-calculated mass of 154 kDa . The appearance of the free FH peak 1 in all of the heparin mixtures showed that the binding between FH and heparin is incomplete, i.e. both unbound and bound FH are present in the mixtures. The similarity in the s20,w values for peak 1 with and without heparin (Table 1) showed that no major conformational change occurred in FH when heparin dp6–dp36 was bound to it.
Ultracentrifugation size-distribution analyses
c( s) of FH titrated with six dp6–dp36 heparin fragments
|Experimental .||RG (nm) .||RXS-1 (nm) .||RXS-2 (nm) .||Length (nm) .||s20,w (S) (monomer) .||s20,w (S) (dimer) .||s20,w (S) (trimer) .||s20,w (S) (tetramer) .||s20,w (S) (pentamer) .|
|Experimental .||RG (nm) .||RXS-1 (nm) .||RXS-2 (nm) .||Length (nm) .||s20,w (S) (monomer) .||s20,w (S) (dimer) .||s20,w (S) (trimer) .||s20,w (S) (tetramer) .||s20,w (S) (pentamer) .|
R. Nan, A.I. Okemefuna and S.J. Perkins, unpublished work.
Heparin induced FH oligomer formation. Free FH forms approximately 15% of oligomers at different pH and salt concentrations . These oligomers are visible as resolved peaks 2–9 in the c(s) plots (Figure 2d). With heparin dp6 and dp12, peaks 2–9 were unchanged. For dp6, the first six s20,w values for peaks 2–7 ranged from 8.0 to 19.6 S at 50000 rev./min (Figure 2d and Table 1). For dp12, similar s20,w values were seen. The mean proportion of the oligomers at 40000, 50000 and 60000 rev./min were 9±3% and 9±2% for dp6 and dp12 respectively, which are both slightly less than the 12±3% observed for free FH (Figure 3a). Thus heparin dp6 and dp12 may slightly inhibit free FH oligomer formation. In distinction, the FH mixtures with dp18–dp36 showed notable increases in oligomer formation from the increased peak intensities seen between 7 and 22 S (Figure 2d). The more complex peaks seen in the c(s) analyses for dp18–dp36 meant that the peak numbering beyond the pentamer may not correspond to a given FH oligomer number. Integrations showed that the oligomers increased from 9 to 12% up to as much as 63% for dp36 (Figure 3a). For heparin dp30–dp36, the exceptional size of peak 8 at s20,w values of 18.6 and 18.0 S respectively accounted for 49% and 28% of these FH oligomers (Figure 2d). If FH and heparin dp30/dp36 are assumed to form a 1:1 complex, the 61–63% of oligomer formation corresponds to an estimated Kd value of 0.5 μM for the FH–heparin interaction. Even the proportion of dimer increased with the increase in heparin size. For free FH and with dp6 and dp12, the dimer proportion was similar at 4–5%. With the larger heparin fragments, the proportion of dimer increased to as much as 14% with dp36 (Figure 3a).
Summary of the
c( s) plots of FH–heparin mixtures
Concurrently with the increases in FH oligomers, the s20,w values of the oligomers shifted to lower S values. The FH dimer peaks in the dp18–dp36 complexes shifted by approximately 0.7 S towards lower S values, compared with those seen with dp6 and dp12 (Figure 2d and Table 1). Similar shifts towards lower S values were also observed for the trimer, tetramer and pentamer peaks (Table 1). These changes indicated that a new type of oligomer structure had formed with a more extended conformation than for the free FH oligomers.
Additional AUC experiments were performed to extend these findings. To show that both dp6 and dp12 bind to FH, sedimentation velocity experiments were performed for increasing molar ratios of dp6 up to 1:8. The c(s) peak for free unbound dp6 at 1 S only appeared at the highest molar ratios, showing that dp6 interacted with FH (Supplementary Figure S1a at http://www.BiochemJ.org/bj/444/bj4440417add.htm). No change in the proportion of FH oligomers was seen (Figure 3b). Similar sedimentation velocity experiments for the FH–dp12 molar ratios showed that the proportion of FH oligomers increased from 12% without dp12 up to 28% at the highest dp12 molar ratios, showing that dp12 bound to FH (Figure 3c and Supplementary Figure S1b). Velocity experiments for the same FH–dp36 mixture were performed at three different rotor speeds (Supplementary Figure S1c). Although peak broadenings or narrowings were observed, the absence of shifts in the S values of the individual c(s) peaks indicated that the oligomers were structurally well-defined with slow association rates on the time scale of sedimentation.
To complete these experiments for the NA-domains of HS, FH was mixed with four purified minimally sulfated HS fragments (dp6, dp12, dp18 and dp24) using two rotor speeds of 40000 and 50000 rev./min (Figure 2e). In all four mixtures, peak 1 for monomeric FH gave s20,w values of 5.52±0.04 S that was unchanged from the s20,w value of 5.48±0.02 S for free FH at 0.87 mg/ml and an s20,w value of 5.67±0.07 S for the FH–heparin mixtures. For HS dp6 and dp12, the individual peaks 2–7 were again well resolved at s20,w values between 8.0 and 14.9 S (Figure 2e). For HS dp18 and dp24, the s20,w values for the dimer were reduced (Figure 2e), showing that FH–HS complex formation had occurred, although the changes were less than those seen with heparin. The proportion of FH oligomers with HS dp6 and HS dp12 were 8% and 11% respectively, increasing to 21% and 14% with HS dp18 and dp24 (Figure 3a). In conclusion, an interaction between FH and minimally sulfated HS was detected, even though this was weaker than that seen with heparin. Interestingly, when FH was mixed with the unfractionated HS parent material containing both S- and NA-domains, a strong interaction between FH and HS was seen, including the observation of three large new peaks 7, 8 and 9 in the c(s) analysis (Figure 2e). This showed that the S-domains of HS interacted more strongly with FH than the NA-domains.
X-ray scattering data for FH with heparin and HS
Solution scattering is a diffraction technique that reports on the overall structure of biological macromolecules in solution . Scattering is complementary to AUC because of its greater sensitivity to oligomer formation. The conformational and self-association properties of FH were studied in the presence of the six heparin fragments dp6–dp36. FH (3.0 μM) was used with 1:1 molar ratios of each heparin fragment. Here, the reduced FH concentrations of 0.47 mg/ml reduced the proportion of the larger FH oligomers. Because heparin dp6–dp36 (2–11 kDa) is small compared with FH (154 kDa), the Guinier analyses primarily monitor structural and associative changes in FH only. Linear RG analyses for FH and its heparin mixtures were obtained in satisfactory QRG ranges as required (Supplementary Figure S2a at http://www.BiochemJ.org/bj/444/bj4440417add.htm). The mean RG values for the dp6 and dp12 mixtures were 8.4 and 7.3 nm respectively, being slightly smaller than that of 8.9±0.3 nm for free FH (Table 1). The Guinier I(0)/c value is proportional to molecular mass. Because I(0)/c also decreased slightly for the dp6 and dp12 mixtures (Figure 4b), this also suggested that dp6 and dp12 partly blocked the FH self-association sites, in agreement with the AUC results. In contrast, the RG values increased significantly up to 11.7 nm when heparin dp18–dp36 was added (Table 1, Figure 4a and Supplementary Figure S2a). The I(0)/c values also increased significantly with dp18–dp36, being doubled for dp36 (Figure 4b). This showed that strong FH oligomer formation was induced by dp36.
Dependence of the Guinier parameters of the 1:1 FH–heparin and 1:1 FH–HS mixtures with oligosaccharide size
The two cross-sectional RXS-1 and RXS-2 Guinier analyses (Supplementary Figure S2) at larger Q values monitor the proximity relationships between non-neighbouring (RXS-1) and neighbouring (RXS-2) SCR domains . The RXS-1 value of free FH was 2.93 nm at 0.47 mg/ml (Table 1), in good agreement with the previous value of 2.7±0.1 nm for free FH at 1 mg/ml . The RXS-1 values were almost unchanged at 2.80–3.08 nm when dp6 and dp12 were added (Table 1). The RXS-1 values increased significantly to as much as 5.11 nm when dp18–dp36 was added (Figure 4c). Thus the increased RXS-1 values (Table 1) showed that new proximity relationships between non-adjacent SCR domains were formed in the FH–heparin complexes. In distinction, the RXS-2 value of free FH was 1.82 nm (Table 1), in good agreement with the previous RXS-2 value of 1.79±0.03 nm , and this was not affected on the addition of dp6–dp36 (Figure 4d and Table 1). Thus the extended arrangement of neighbouring SCR domains in FH remained similar in the presence of heparin.
The distance distribution function P(r) is the summation of all the distances r between atoms within the macromolecule (Figure 5). The individual RG and I(0)/c values calculated from the P(r) analyses agreed well with those from the Guinier analyses, showing consistency (Figures 4a and 4b). The maximum length L is measured when the P(r) curve reaches zero at a large r value. In the absence of heparin, the L value of FH at 0.47 mg/ml was 34 nm, in good agreement with the previously reported L value . The maximum M corresponds to the most frequently occurring distance. For free FH, the main peak M1 and a sub-peak M2 were observed at r values of 5.8 and 11.6 nm respectively. When dp6 and dp12 were added, L, M1 and M2 were unchanged at 32–35 nm, 6.0–6.7 nm and 11.6–11.7 nm respectively (Figure 5a). For dp18–dp36, L increased with heparin size (Table 1), M1 disappeared and M2 became more prominent. The increased L values showed an increase in FH oligomer formation. The unnormalized P(r) curves showed larger intensity increases with dp24–dp36 (Figure 5b). In summary, the P(r) curves confirmed the formation of FH oligomers with larger dimensions when heparin dp24–dp36 was added.
The distance distribution function
P( r) analyses for the 1:1 FH mixtures with heparin dp6–dp36 and HS dp6–dp24
Scattering showed that the use of minimally sulfated HS dp6–dp24 (which corresponded to the NA-domains of HS) resulted in much reduced oligomerization of FH. Linear Guinier plots were again observed with 1:1 mixtures of FH and HS (fits not shown). The mean RG values for FH with eight HS fragments (dp6–dp24) increased from 7.77 nm for dp6 to 8.27 nm for dp24. These RG increases were reduced compared with those with heparin (Figure 4a). Small increases in the Guinier I(0)/c values were seen that were comparable with those seen for heparin (Figure 4b). The mean RXS-1 values seen for FH with these eight HS fragments ranged between 2.50 and 2.87 nm, which were not much increased compared with the RXS-1 value of 2.71 nm for free FH (Figure 4c). No significant change in the RXS-2 values was similarly observed (Figure 4d).
The P(r) curves for FH with HS dp6–dp24 showed that some FH oligomer formation had taken place. The maximum length L was 30 nm for FH and increased slightly to 32 nm for HS dp24 (Figures 5c and 5d). The maximum at M1 was unchanged at r values of 4.8, 4.8, 5.1, 4.8 and 4.2 nm. Unlike FH with heparin, the maximum at M2 remained almost unaffected in the five samples. Nonetheless the intensity increases seen in the unnormalized P(r) curves showed that HS promoted some FH oligomer formation (Figure 5d).
SPR studies of the FH–HS interaction
SPR monitors the interaction between a binding partner in solution (analyte) and an immobilized partner (ligand) attached to the surface of a sensor chip . The present study of the FH interactions with heparin and HS-coated sensor chips is analogous to the FH interactions with HS-coated cell surfaces. Initial SPR experiments utilised immobilized biotinylated heparin dp32 on streptavidin-coated sensor chips. The use of six FH concentrations resulted in a binding curve showing a Kd value of 2.7 μM (Figure 6a, inset). The same experiment with dp24 in place of dp32 resulted in a weaker Kd value of 5.4 μM (Figure 6d, inset). The use of polydisperse unfractionated heparin in place of dp32 and dp24 gave a stronger Kd value of 0.5 μM (Figure 6e, inset). In all three cases, even though the sensorgrams dipped slightly at 200 s (attributable to the slight dissociation of self-associated FH bound to immobilized heparin), the sensorgrams show that the on- and off-rates were rapid. Heparin dp32 (10 kDa) and dp24 (7.5 kDa) were each immobilized to 13 RUs and 10 RUs respectively. From this, the maximum binding response (Rmax) for the binding of full-length FH (154 kDa) is predicted to be 200 RUs and 205 RUs respectively. Since the observed intensities were less than 200 RUs in Figures 6(a) and 6(d), the binding data indicated a 1:1 interaction between FH and the heparin-coated sensor chip. It was concluded that SPR revealed low micromolar Kd values for the FH–heparin association, in good agreement with the Kd estimate of 0.5 μM determined independently from AUC.
SPR analyses of the interaction of FH and its SCR-6/8 and SCR-16/20 fragments with immobilized heparin and HS
A bivalent model of heparin binding to FH would require that the FH fragments SCR-6/8 and SCR-16/20 would each bind to the dp32-coated sensor chips. Both were indeed observed to bind to dp32. SCR-16/20 (39.7 kDa) showed rapid on- and off-rates (Figure 6b). The observed increase of up to 85 RUs was greater than the predicted Rmax of 52 RUs. This discrepancy is best explained if more than one SCR-16/20 molecule is bound to a single immobilized heparin dp32 molecule. The Kd value was 20 μM, which is weaker than that for full-length FH binding. SCR-6/8 (21.1 kDa) also showed rapid on- and off-rates (Figure 6c). The observed increase of up to 27 RUs was slightly above the Rmax of 25 RUs predicted for this interaction. More than one SCR-6/8 molecule is inferred to bind to immobilized dp32. This second Kd value was 4.3 μM, which is also weaker than that for full-length FH binding. These Kd values show that the bivalent binding of full-length FH to heparin is stronger than binding at either SCR-6/8 or SCR-16/20, and therefore this binding is co-operative.
In order to compare the binding of the S-domains and NA-domains of HS, SPR experiments were performed with the unfractionated HS parent material, and with minimally sulfated HS dp24 and dp18. FH binding with a Kd value of 2.3 μM to unfractionated HS was observed (Figure 6f, inset). This is comparable with the Kd value of 0.5 μM for FH binding to unfractionated heparin (Figure 6e, inset), showing that the presence of both S- and NA-domains in HS has a similar affinity to that of the S-domains in heparin alone. When minimally sulfated HS dp24 and dp18 were tested, no FH binding was seen to either of the two HS-coated sensor chips (Figures 6g and 6h), even when 300 RUs of HS dp24 was bound to the sensor chip. These results confirmed the much weaker binding of FH to the NA-domains of HS observed above by AUC and scattering (Figures 2e and 4).
Modelling of FH–heparin dp36 complexes
Molecular modelling was used to generate structures for bivalent FH–heparin complexes in order to validate the experimental AUC data . Previous s20,w modelling had explained the FH peaks 2–7 in terms of self-associated dimeric to heptameric FH structures . This modelling was now extended to fit the multiple peaks for the FH–heparin complexes using the most recent published solution structure for FH (PDB code 3GAV) and the heparin dp36 model (PDB code 3IRL) [26,27]. The SCR-7 and SCR-20 domains in this FH model were separated by 26 nm (Figure 1). The SCR-7 and SCR-20 domains in FH were readily joined with heparin dp36 molecules to create ring-like models for dimeric, trimeric, tetrameric and pentameric oligomers (Figure 7a). Similar models could be made using heparin dp18 (results not shown). It was not possible to fit the AUC models to the X-ray scattering curves because the scattering data correspond to a polydisperse mixture of macromolecular species that could not be resolved.
FH interaction with a polyanionic host cell surface deduced from the present study
The FH–heparin models gave good agreements between the experimental and predicted s020,w values. The predicted s020,w value for an antiparallel dimer model created from two dp36 and two FH molecules was 7.5 S. Given that this modelling is accurate to ±0.21 S , the fair agreement with the observed values of 7.0–7.2 S for dp18–dp36 was obtained (Table 1). In distinction, a parallel dimer model gave a more compact shape and a larger predicted s020,w value of 8.1 S [Figure 7a(i) and 7a(ii)]. The predicted s020,w values for the trimeric, tetrameric and pentameric ring-like complexes were 8.8, 9.5 and 10.6 S [Figure 7a(iii)–(v)]. These agreed well with the observed values of 8.2–8.6, 9.1–9.5 and 10.4–10.8 S for the dp18–dp36 complexes (Table 1). These predicted s020,w values were different from the experimental s20,w peaks for free FH (Table 1). The modelling was therefore able to explain the shape changes in FH after binding to heparin. Other tetrameric and pentameric complexes with more compact crossed-over FH structures gave much increased predicted s020,w values of 11.3 and 12.4 S (Table 1), and these structures were not favoured (results not shown).
Because the relative separation of 26 nm between SCR-7 and SCR-20 in this low-resolution FH model (PDB code 3GAV) cannot be considered to be definitive, two other scattering best-fit FH models were evaluated (R. Nan, A.I. Okemefuna and S.J. Perkins, unpublished work). The second was based on fitting multi-SCR fragments to the scattering data; the third used the scattering curve of homozygous FH for fits. In these other models, SCR-7 and SCR-20 showed reduced separations of 10 and 13 nm. The second and third models gave predicted s020,w values that were 1 S higher compared with the 2009 3GAV model (Table 1). Thus the closer separations of 10 and 13 nm between SCR-7 and SCR-20 in alternative FH models resulted in models that were too compact to explain the observed c(s) peaks for the FH–heparin complexes.
Our present study of the HS interactions in solution and on sensor chips with FH suggests a bivalent and co-operative mechanism for FH binding to host cell surfaces (Figure 7). This has not previously been established. Unfractionated HS is composed of distinct regions of sulfated S-domains and unsulfated NA-domains ; in the present study these two regions were experimentally represented by our heparin and HS fragments respectively. Our FH-binding studies with anionic oligosaccharides provided several novel molecular insights into FH–heparin and FH–HS interactions at near-physiological conditions. As two independent methods, AUC and X-ray scattering both consistently showed from the large intensity changes that long purified heparin fragments promote large FH oligomers, whereas the shorter ones bind to FH but do not promote oligomers. AUC also showed from the lack of change in the FH monomer c(s) peaks that no major structural re-arrangement occurred in FH after heparin binding. SPR confirmed the low micromolar Kd value for the FH–heparin affinity, and weaker Kd values for two individual FH–heparin sites. These results on FH oligomer formation can only be explained by bivalency (i.e. the presence of at least two independent heparin-binding sites on FH). This was verified by AUC molecular modelling of the FH–heparin complexes. The SPR results also showed that the two sites exhibited co-operative heparin binding. Our comparison of FH binding to unfractionated HS, sulfated heparin fragments and minimally sulfated HS fragments shows that FH binds preferentially to the S-domains of HS. FH thus behaves in the same way as numerous other proteins with which HS exerts its biological activities through the S-domains . Polyanionic molecules such as HS and sialic acids on the host cell surface enhance the regulatory effectiveness of FH by 10-fold through its inhibition of complement activation [47–49].
The affinity of the FH–HS interaction has now been quantified by AUC and SPR for comparisons with other FH–ligand affinities, and this is supported by the scattering data. FH occurs in plasma at concentrations of 0.235–0.81 mg/ml (1.5–5.3 μM) . Free FH self-associates to form 15% oligomers in a range of salt concentrations and pH with a monomer–dimer Kd value of 28 μM . FH self-association sites are located within the SCR-6/8 and SCR-16/20 domains, both of which also bind to HS [26,29,35]. Heparin dp6 and dp12 consistently caused only small changes in the AUC and scattering data which indicated that dp6 and dp12 inhibit FH self-association (Figures 3 and 4). These FH–heparin Kd values are accordingly comparable in their magnitudes with the 7 μM FH concentration used for AUC and scattering. Such a FH–heparin Kd value is less than (and therefore consistent) with the FH monomer–dimer Kd value of 28 μM. For heparin dp30 and dp36, the AUC data indicated stronger binding to FH with an estimated Kd value of 0.5 μM (Figure 3a). This was confirmed by SPR that revealed a Kd value of 0.5 μM for unfractionated heparin (Figure 6). Our Kd values differ from that of 9.2 nM for FH binding to unfractionated heparin in 50 mM sodium phosphate and 100 mM NaCl buffer by SPR, and that of 9 μM for FH SCR-19/20 binding to dp4 in 20 mM acetate and 200 mM NaCl buffer (pH 4), by NMR [44,50]. These differences are attributable to the different and less physiological conditions used by others. The major contribution to FH–HS binding is from the S-domains (heparin) of HS, this being confirmed by AUC experiments with HS NA-domains that showed much weakened binding. This was confirmed by SPR studies of FH with the unfractionated HS material containing both S- and NA-domains, giving a Kd value of 2.3 μM that is comparable with the Kd value of 0.5 μM for FH binding to unfractionated heparin. Continuous S-domains are therefore not necessary for the interaction with FH; two S-domains separated by the NA-domain, known as the SAS motif , is sufficient. Additional SPR experiments using heparin dp32 with each of SCR-6/8 and SCR-16/20 showed that binding was observed with both fragments. For both fragments, the Kd values were weaker than that for full-length FH, showing that the two separate sites act co-operatively when together in full-length FH. Interestingly, our low micromolar Kd values for the FH–heparin interaction are similar to those of 2–4 μM reported for the other major FH–C3b and FH–CRP (C-reactive protein) interactions . Therefore we conclude that all of the major physiological FH–ligand interactions exhibit moderate micromolar binding affinities.
Molecular modelling of the FH–heparin complexes compared with the experimental AUC results for the resolved oligomers validated the AUC data showing bivalency (Figure 7a). Heparin dp18–dp36 resulted in significant amounts of large FH–heparin oligomers, observed both by AUC and scattering (Figures 3 and 4). Ring-like models of FH–heparin accounted well for the positions of the c(s) peaks, and explained why these were shifted to lower S values compared with those for free FH. They also explained the intensity increases seen by scattering. These models require the existence of at least two separate heparin-binding sites within SCR-7 and SCR-20. Notably, SCR-7 has the second-highest basic charge density in FH (pI = 6.1–6.3), surpassed only by SCR-20, these basic charges being optimal for interactions with anionic S-domains. It is envisaged that heparin dp18–dp36 are long enough to cross-link SCR-7 and SCR-20 sites on different FH molecules to form daisy-chained rings (Figure 7a and Table 1), and this assembly is likely to be assisted by co-operativity between the two heparin sites. Small dp6–dp12 fragments were too small to cross-link FH. These oligomers are consistent with those seen by others using AUC with both dextran sulfate and unfractionated heparin (12–14 kDa; approximately dp38–dp46) . That 2009 study was based on sedimentation equilibrium analyses, in which the presence of dimer and tetramer was inferred indirectly using sedimentation equilibrium fits. Our sedimentation velocity c(s) analyses directly identified multiple FH–heparin oligomers from the individual peaks for dimer to octamer. The FH–heparin oligomers are well-ordered and structurally distinct from the multiple FH oligomers induced through the cross-linking of weak zinc-binding sites on the surface of FH by >10 μM zinc . No well-defined FH–zinc oligomers were observed.
The bivalent FH–HS structures provide new insight into the FH domain structure, and the way in which FH interacts with host cell surfaces (Figure 7). Foremost, the results of the present study show that the inter-SCR flexibility within FH is less than is generally assumed. Even though two heparin-binding sites exist in FH, heparin dp18–dp36 binding did not cause a detectable conformational change in FH to pull these two sites closer together. Such a conformational change would reduce the separation of the SCR-7 and SCR-20 domains from 26 nm (Figure 1) to 6 or 12 nm (the lengths of dp18 or dp36 respectively). Increased s20,w values for the FH monomer peak were never observed (Figure 2 and Table 1). This deduction concurs with salt and pH studies of FH which showed small changes with salt or pH, but little major variation . It appears that the heavily glycosylated and smaller SCR-12/15 domains at the centre of FH maintain the structural independence of the N-terminal (SCR-7) and C-terminal (SCR-20) ends of FH. The lack of high inter-SCR flexibility accounts for the observation of well-defined c(s) peaks for the FH–heparin complexes (Figure 2 and Supplementary Figure S1). A relatively inflexible FH domain structure means that the two heparin-binding sites in FH will function independently of each other, and are able to act co-operatively, both as observed. In summary, the ring-like models of the FH–dp36 oligomers (Figure 7a) are readily transformed into a view of a FH monomer binding to two different S-domains of HS molecules on a host cell surface (Figure 7b).
A bivalent HS-binding mechanism to FH has functional implications. For reason of co-operativity, FH will preferentially bind to surfaces showing the right spatial density of anionic oligosaccharides to bind simultaneously to both SCR-7 and SCR-20. In theory, two separate weak binding events with micromolar affinities will become a strong interaction with a picomolar affinity if both weak sites bind simultaneously . Bivalency will position FH SCR-1/4 next to cell-surface-bound C3b for its regulation at the same time as SCR-7 and SCR-20 are tethered (Figure 7a). Reduced HS binding at either SCR-7 or SCR-20 will have a disproportionate effect on FH regulatory function. In FH-associated disease, such as AMD or aHUS, the FH–heparin interaction may be affected by polymorphisms or mutations. Disease-risk polymorphisms will exert their effect over a period of decades, whereas disease-causing mutations will show a much earlier effect in life. This difference is reflected in the occurrence of AMD in the aged population, this being responsible for over 50% of blindness in the elderly in the Western world , whereas aHUS is a common cause of renal failure in young children. Two possible mechanisms for FH-associated disease involve either FH aggregation to form pathogenic deposits or the loss of FH regulatory control of inflammation . Although further experiments will be needed to clarify these, the present study provides insight into both possible mechanisms. (i) In terms of aggregation leading to deposit formation, FH occurs in drusen (subretinal pigment epithelial deposits) that are a hall-mark of AMD . Glycosaminoglycans have also been identified within Bruch's membrane, although their size and structure in drusen is not yet known . The present study shows that the availability of S-domains in glycosaminoglycans with size dp18 or more would lead to FH aggregates, by analogy with the same process proposed for FH–zinc aggregates . These may form if host cell surfaces are damaged to release anionic oligosaccharides during inflammatory attack. (ii) In terms of reduced inflammatory regulation, a Y402H polymorphism in SCR-7 is a risk factor for AMD [21–24]. Some reports state that FH His402 binds more weakly to heparin than FH Tyr402, whereas others report no difference [55,56]. In addition the weaker binding of CRP to FH His402 compared with FH Tyr402 may also contribute to disease . As illustrated (Figure 7b), the weaker binding of FH His402 to HS S-domains would compromise both the FH interaction with C3b and the bivalent binding of FH. (iii) In the alternative scenario of FH mutations in SCR-19/20 that lead to aHUS (Figure 7c), these occur mostly in young individuals, often triggered by an immune insult to the kidney . aHUS is primarily caused by mutations within SCR-19/20, often those affecting heparin-binding or C3d-binding properties, but curiously much less so within SCR-7. The bivalency of the FH–HS S-domain interaction accounts for this different phenotype in terms of compromising FH binding to host cell surfaces, rather than the FH–C3b interaction.
In conclusion, the present AUC, scattering and SPR studies have identified a bivalent and co-operative FH binding mechanism to HS that clarifies the molecular mechanism of FH binding to host cell surfaces.
Sanaullah Khan designed and performed experiments, analysed data and wrote the paper; Ruodan Nan and Jayesh Gor designed and performed experiments; and Barbara Mulloy and Stephen Perkins designed the study, analysed data and wrote the paper.
We thank Dr Anuj Shukla and Dr T. Narayanan at the European Synchrotron Radiation Facility for outstanding instrumental support, and Dr Ami Miller for useful discussions.
This work was supported by the Higher Education Commission of Pakistan; the Medical Research Council [grant number G0801724]; and the Mercer Fund of the Fight For Sight Charity.