|Home | About | Journals | Submit | Contact Us | Français|
The exchangeability of apolipoprotein (apo) E between lipoprotein particles such as very low density lipoprotein (VLDL) and high density lipoprotein (HDL) is critical for lipoprotein metabolism but, despite its importance, the kinetics and mechanism of apoE-lipoprotein interaction are not known. We have used surface plasmon resonance (SPR) to monitor in real time the reversible binding of apoE to human VLDL and HDL3; biotinylated lipoproteins were immobilized on a streptavidin-coated SPR sensor chip and solutions containing various human apoE molecules at different concentrations were flowed across the surface. Analysis of the resultant sensorgrams indicated that the apoE3/lipoprotein interaction is a two-step process. After an initial interaction, the second slower step involves opening of the N-terminal helix bundle domain of the apoE molecule. Destabilization of this domain leads to more rapid interfacial rearrangement as is seen when comparing the lipoprotein binding of apoE4 relative to that of apoE3. The resultant differences in interfacial packing seem to underlie the differing abilities of apoE4 and apoE3 to bind to VLDL and HDL3. The measured dissociation constants for apoE binding to these lipoprotein particles are in the micromolar range. C-terminal truncations of apoE to remove the lipid binding region spanning residues 250–299 reduces binding to both types of lipoprotein but the effect is less with HDL3; this suggests that protein-protein interactions are important for apoE binding to this lipoprotein while apoE-lipid interactions are more significant for VLDL binding. The two-step mechanism of lipoprotein binding exhibited by apoE is likely to apply to other members of the exchangeable apolipoprotein family.
There is great interest in understanding the structure-function relationships of apolipoprotein (apo) E because of its pronounced anti-atherogenic properties (1). ApoE protects against the development of atherosclerosis, in part, by binding to lipoprotein particles and mediating their clearance from the plasma compartment via interaction with cell surface receptors, notably those of the low density lipoprotein (LDL) receptor family (2;3). ApoE also promotes the cellular uptake of lipoprotein particles by binding to cell surface heparan sulfate proteoglycans (4). The endocytosis of apoE-containing lipoprotein particles reduces plasma cholesterol levels and, thereby, the risk for developing cardiovascular disease (1).
Human apoE is a 299 residue single polypeptide chain that folds into two tertiary structure domains; the N-terminal region (residues 1–191) forms a helix bundle while the C-terminal region is folded separately into some unknown structure (3;5;6). ApoE binds readily to lipoprotein particles such as very low density lipoprotein (VLDL) and high density lipoprotein (HDL) and forms both exchangeable and non-exchangeable pools (7;8). However, the mechanisms by which apoE associates with and dissociates from lipoprotein particles are not known. Human apoE exhibits polymorphism and the two common isoforms apoE3 and apoE4 (which differ by the mutation C112R) partition differently between HDL and VLDL particles (9); this effect occurs because of variations in the interactions between the N- and C-terminal domains so that apoE4 binds preferentially to VLDL whereas apoE3 prefers HDL (6). This differential lipoprotein binding leads to variations in lipoprotein profile such that apoE4 is associated with higher plasma cholesterol levels and increased risk of cardiovascular disease relative to individuals with apoE3 (10;11). The molecular basis for these isoform effects on lipoprotein binding is not well understood.
To address some of the unresolved issues about apoE-lipoprotein interaction, we have used surface plasmon resonance (SPR) to study the reversible binding of apoE to VLDL and HDL particles in real time. SPR has been employed successfully to measure the binding of apoE to proteoglycans (12–15) and members of the LDL receptor family (16), and of enzymes to lipoprotein particles (17;18). The present results show that apoE binds by a two-step mechanism to HDL and VLDL.
HDL3 and VLDL were purified by sequential density ultracentrifugation (19;20) from a pool of fresh human plasma obtained by combining several single units from normolipidemic individuals. Full-length human apoE3, apoE4 and their 22kDa (residues 1–191), 12kDa (residues 192–299) and 10kDa (residues 222–299) fragments were expressed and purified as described previously (21–23). The C-terminal truncation variants (Δ251–299, Δ261–299 and Δ273–299) of apoE3 and apoE4 were made as described before (24;25). The apoE preparations were at least 95% pure as assessed by SDS-PAGE. In all experiments, the apoE sample was freshly dialyzed from a 6M GdnHCl and 1% β-mercaptoethanol (or 5mM DTT) solution into a buffer solution before use.
HDL3 and VLDL were dialyzed into phosphate-buffered saline (pH 7.4) prior to biotinylation. The EZ-link sulfo-NHS-LC-biotinylation kit from Pierce Chemical Co. (Rockford, IL) was used for attaching biotin molecules through a 2.24 nm spacer arm to lysine residues on the surface of HDL particles. HDL3 and VLDL, each at 1.0 mg protein/ml, were mixed with a freshly made 10 mM sulfo-NHS-LC-biotin solution at a 10 fold molar excess of biotin. The lipoproteins were incubated under nitrogen at 4°C overnight before dialysis against Tris-buffered saline (TBS, pH 7.4) to remove unreacted sulfo-NHS-LC-biotin. The degree of biotinylation of the particles was determined using conditions recommended by Pierce. Briefly, solutions containing biotinylated lipoproteins were added to a mixture of HABA reagent [2-(4′-hydroxyphenyl) azobenzoic acid] and immunopure avidin (Pierce Chemical Co.). Because of its higher affinity for avidin, biotin, from the biotinylated lipoproteins, displaced avidin-bound HABA. Therefore, the absorbance at 500 nm of the HABA-avidin complex was reduced. The change in absorbance was used to calculate the level of biotin incorporated into the lipoprotein particles. This procedure yielded an average degree of labeling of 1 biotin molecule per lipoprotein particle.
Studies of the binding of apolipoproteins (association and dissociation) to HDL3 and VLDL were performed with a Biacore 3000 SPR instrument (Biacore, Uppsala, Sweden) using SA sensor chips (Biacore, Uppsala, Sweden). This chip is designed to bind biotinylated ligands through a high-affinity capture process. Prior to immobilization of HDL3 or VLDL on the sensor chip, the streptavidin surface was conditioned with three consecutive 1-minute injections of 1 M NaCl in 50 mM NaOH (50μl/minute). The biotinylated HDL3 or VLDL was then immobilized onto the surface through the quasi-covalent biotin-streptavidin interaction by exposing the surface to the biotinylated lipoprotein solutions in running buffer (50mM Tris-buffered saline (TBS) pH 7.4) until 2500–3000 and 5000–7000 response units (RU) of biotinylated HDL3 or VLDL, respectively, were bound to the surface. This was achieved by a 10μl injection of biotinylated HDL3 or VLDL (1.0 mg protein/ml) at a flow rate of 2μl/minute at room temperature. After 5 minutes, the chip was washed with degassed TBS to remove unattached lipoprotein. A 50μg/ml human apoE3 solution was flowed over the chip at 20μl/min for 2 minutes to block any remaining hydrophobic surface areas and reduce subsequent apoE binding to non-lipoprotein sites; the chip was then washed with TBS until the SPR signal reached a steady background value. The surface of the immobilized HDL3 or VLDL was then exposed to a 4 minutes injection of apoE dissolved in degassed TBS at a flow rate of 20μl/min to monitor association and then TBS alone was passed over the sensor surface to monitor apoE dissociation from the immobilized lipoprotein particles. For these experiments, two flow cells were monitored simultaneously with flow cells 1 and 2 containing immobilized biotinylated VLDL and HDL3, respectively. A sensor chip lacking immobilized lipoprotein could not be used as a reference cell because apoE bound more to this surface than to a lipoprotein-coated chip. The apolipoproteins were dialyzed from 6 M GdnHCl containing 5mM DTT into TBS, filtered (Ultrafree – MC centrifugal filter devices, 0.1 μm filter unit, Millipore, Bedford, MA) and degassed before serial dilutions (2.5 to 50 μg/ml) were made just prior to injection. The sensor chip was washed two times with 20μl TBS between each injection of apolipoprotein. The chips were used for 2 days in repetitive experiments. Regeneration of the sensor chip surface was not possible since the lipoproteins were directly immobilized via biotin-streptavidin interaction. The apoE sensorgrams were independent of flow rate in the range 10–40 μl/min indicating that the apoE binding at 20μl/min was not limited by mass transport (diffusion) effects. Steady state binding isotherms and Kd values of the binding to HDL3 and VLDL were obtained by generating sensorgrams at different apoE concentrations. The sensorgrams were analyzed with the BIA evaluation software version 4.1 (Biacore, Uppsala, Sweden). The response curves of various apolipoprotein (analyte) concentrations were fitted to the two-state binding model described by the following equation (26;27).
The equilibrium constants of each binding step are K1 = ka1/kd1 and K2 = ka2/kd2 and the overall equilibrium binding constant is calculated as Ka = K1 (1 + K2) and Kd = 1/Ka. In this model, the analyte (A) binds to the ligand (HDL3 or VLDL) (B) to form an initial complex (AB) and then undergoes subsequent binding or conformational change to form a more stable complex (ABx). A further check of the two-state binding mechanism was obtained by variation of the contact time for association between the apoE and the lipoprotein particle. For a two-state reaction, an increase in the contact time between the analyte and the ligand decreases the dissociation rate since more of the stable ABx complex is formed. For the apolipoproteins, binding responses in the steady-state region of the sensorgrams (Reg) were also plotted against apolipoprotein concentration (C) to determine the overall equilibrium binding affinity. The data were subjected to nonlinear regression fitting (Prism 4, GraphPad Inc.) according to the following equation;
Rmax is the maximum binding response and Kd is the dissociation constant. This SPR approach for measuring Kd is validated by the fact that monitoring the binding of apoE3 and apoE4 to VLDL by ultracentrifugation yielded similar Kd values (23).
Fig. 1 shows the immobilization of biotinylated HDL3 and VLDL on a streptavidin-coated sensor chip when a solution of the lipoprotein is flowed across the surface. It is apparent that there is rapid binding to give a steady state level. Flowing buffer solution alone across the chip washes away a small amount of weakly bound lipoprotein but the remainder is essentially irreversibly bound by the biotin-streptavidin interaction. In multiple experiments, the amounts of HDL3 and VLDL immobilized by this procedure were 2633 ± 508 and 5893 ± 1247 RU (mean ± SD), respectively. It should be noted that these RU levels for bound HDL3 and VLDL cannot be simply interpreted in terms of relative masses of immobilized lipoprotein because of different refractive index properties due to different lipid/protein stoichiometries. The immobilization of HDL3 or VLDL in this fashion permits evaluation of the binding of apoE which is difficult to monitor in solution because of problems, particularly in the case of HDL3, in readily separating the lipoproteins and unbound apoE. Another advantage of the SPR method is that labeling the apoE molecules is not required and the association/dissociation reactions can be monitored in real time. A potential problem in the SPR experiment is steric restriction of apoE binding by the proximity of the sensor chip surface, but, because the HDL3 and VLDL particles are attached to the surface by a 2.24nm spacer arm, protein molecules can access the lipoprotein surface. For example, LCAT (17) and CETP (28) have been shown to bind to HDL in similar SPR experiments.
The sensorgram obtained when a 50μg/ml apoE3 solution was flowed across the sensor chip with immobilized VLDL is shown in Fig. 2A. The data are corrected for bulk refractive index effects and fitted using a two-state binding model (equation 1). The kinetic data were not fitted well by a 1:1 Langmuir binding model as reflected by larger values of the goodness of fit parameter (chi2) (data not shown). It follows that the binding of apoE3 to VLDL particles involves either a sequential two-step process or some conformational change (14;26;27). The binding of apoE3 to HDL3 is also best described by the two-state binding model (data not shown). The deconvoluted curves in Fig. 2A show the initial binding (AB) and subsequent binding or conformational changes (ABx). Because of uncertainties about the valences of the lipoprotein particles (the HDL3 and VLDL particles are heterogeneous and do not contain a discrete number of binding sites), it is difficult to interpret the kinetics of association in detail. However, inspection of the deconvoluted curves in Fig. 2A reveals that at the end of the 4 minute association phase of the experiment, the size of the AB pool of bound apoE3 is approximately twice that of the ABx pool. It is also apparent that in the dissociation phase (after 4 minutes), apoE3 molecules in the AB pool dissociate more readily than those in the ABx pool. Analyzing several sensorgrams, the value for the dissociation rate constant kd1 for the AB pool is in the range of 2–9 × 10−2s−1 while the value of kd2 for the ABx pool is about fifty times smaller. These kd values correspond to halftimes of dissociation of about 14s and 10 min for the AB and ABx pools, respectively. The rate of dissociation of apoE3 from the VLDL surface is dependent upon the length of time the apoE3 is in contact with the particle. As shown in Fig. 2B, progressively shortening the injection (contact) time from 240s to 60 and 30s enhances apoE3 dissociation from the VLDL surface. The fitting data indicate that the fraction of apoE3 in the AB pool approximately doubles to 0.25 as the contact time is decreased to 30s. The value of kd1 for the AB pool is not affected by the change in contact time whereas kd2 increases by about 40% as the contact time is decreased to 30s. These results are consistent with the second step in the two-step mechanism of apoE3 binding to VLDL involving a conformational change in the bound apoE molecules.
Fig. 3 shows a typical series of sensorgrams for binding of apoE at different concentrations to immobilized lipoprotein. Fitting of sensorgrams of this type with the two-state binding model yields a maximal binding value (Rmax) for each concentration of apoE3. Fig. 4A shows a steady-state binding isotherm for WT apoE3 to VLDL plotted using such Rmax values and fitted to a one-site binding model (equation 2). The data in Fig. 4A yield a Kd value of 20 ± 5μg apoE3/ml (0.6 ± 0.1μM, mean ± SEM) for the binding of human apoE3 to VLDL3. This Kd value indicates that the apoE3/VLDL interaction is of moderate affinity and is in agreement with the value of 23 ± 4μg/ml reported previously by us using an ultracentrifuge method (23). Fig. 4A compares the isotherms obtained for WT human apoE3 and its separate N- (residues 1–191) and C-terminal (residues 192–299 and 222–299) domains binding to VLDL; the Kd and Bmax values derived from these isotherms are listed in Table 1. It is apparent that removal of the C-terminal domain to form 22kDa apoE3 gives rise to a weakly hyperbolic binding isotherm so that the errors in the Kd and Bmax determinations are large (Table 1). Neither the amount of apoE3 that binds (Bmax) nor the binding affinity (Kd) are altered significantly by removal of the C-terminal domain although there is a trend towards a higher Kd value for 22kDa apoE3, especially when considered on a molar basis (Table 1). In contrast, the isolated apoE 10 and 12 kDa C-terminal fragments exhibit clearly saturable binding to VLDL (Fig. 4A); the Kd values are lower than that for WT apoE3 (Table 1) indicating higher binding affinity. Interestingly, the presence of the so-called hinge region (residues 192–221) (3) in 12 kDa apoE does not change the binding from that found with the isolated C-terminal 10 kDa apoE domain. Inspection of Fig. 4B and the Kd and Bmax values listed in Table 1 indicates that the pattern of lipoprotein binding of the apoE3 domains is similar with VLDL and HDL3.
The tertiary structure domains in the apoE molecule are influenced by polymorphism. ApoE3 is the most prevalent isoform and has normal functionality whereas the C112R mutation to yield apoE4 induces abnormal behavior (2). Introduction of the arginine residue at position 112 destabilizes the N-terminal helix bundle domain (29;30) and alters the interaction between the N- and C-terminal domains (3;6). The altered domain-domain interactions are known to underlie the differing preferences of apoE3 and apoE4 for binding to HDL and VLDL (9;23;25;31;32). To examine what effect, if any, the apoE3/apoE4 polymorphism has on the mechanism and kinetics of lipoprotein interaction, the binding of these isoforms to VLDL and HDL3 was compared by SPR. Isotherms of the type shown in Fig. 4 were generated and the resultant Kd and Bmax values are listed in Table 1. In the case of VLDL, the Kd values are the same indicating that the binding affinity for apoE3 is essentially unaltered by the C112R mutation. However, the Bmax for apoE4 is approximately twice that for apoE3, consistent with prior observations that apoE4 partitions more to VLDL when added to plasma (9). In the case of HDL3, the situation is reversed with the Bmax for apoE3 being the highest (Table 1). As mentioned earlier, the apoE3-HDL3 binding isotherm is quasi-linear which leads to large uncertainties in the derived Kd values and, because of this, a significant difference in binding affinity could not be detected. Similar considerations apply to the binding data for 22kDa apoE3 and 22kDa apoE4 because the binding to both VLDL and HDL3 was not readily saturable and the binding isotherms were quasi-linear. However, it is apparent from the Bmax values in Table 1 that the isolated 22kDa N-terminal domains of apoE3 and apoE4 do not discriminate between HDL3 and VLDL like the intact apoE3 and apoE4 molecules do.
The sensorgram in Fig. 2A shows that when apoE3 binds to VLDL the kinetics of binding are such that at the end of the association phase (240s), the AB pool is greater than the ABx pool. The sensorgram and deconvolution curves in Fig. 5A demonstrate that the kinetics are different for apoE4 binding to VLDL; in this case, the ABx pool is greater than the AB pool at 240s. This more rapid shift to the ABx pool for apoE4 compared to apoE3 is associated with higher values for the rate constants, kd1 and ka2 (equation 1). Similar differences in kinetics are seen when apoE3 and apoE4 bind to HDL3 (data not shown). The data in Fig. 5B demonstrating that the AB and ABx pools for 22kDa apoE4 binding to VLDL at 240s are of similar size indicates that the presence of both structural domains is required for the relatively rapid transition to the ABx state seen with apoE4. Consistent with the concept, the ABx pool observed when 12kDa apoE binds to VLDL is smaller than the AB pool at 240s (Fig. 5C). The 10kDa apoE C-terminal domain shows similar behavior (data not shown).
The data in Fig. 4 and Table 1 showing that the C-terminal 10kDa and 12kDa domains are capable of higher affinity, saturable, binding to VLDL and HDL3 are consistent with prior reports (3;31;33) of the importance of the region spanning residues 250–299 for lipid binding. To further explore the contributions of this part of the apoE molecule to binding to VLDL and HDL3, we employed apoE3 and apoE4 variants with residues 273–299, 261–299 and 251–299 deleted. The physical properties of these engineered apoE3 and apoE4 molecules have been characterized previously (24;25). The isotherms in Fig. 6A show that C-terminal truncation reduces the level of apoE4 binding to VLDL; in particular the Bmax values for the 1–260 and 1–250 variants are significantly reduced. The results also show that removing residues 273–299 has a relatively small effect and that residues 261–272 make a marked contribution to binding (cf. (24;31)). These C-terminal truncations in apoE3 have similar effects on VLDL binding (Fig. 6B); the hyperbolic nature of the binding isotherms is reduced for the C-terminal truncation variants and saturation is not evident in the concentration range covered. The data in Fig. 6C, D show that the C-terminal truncations cause smaller reductions in the binding of apoE3 and apoE4 to HDL3 compared to the situation with VLDL (Fig. 6A, B). Since residues in the region 250–299 promote lipid interactions, their small contribution to the binding of apoE3 and apoE4 to HDL3 implies that protein-lipid interactions are not a major factor in this case.
This SPR study gives the first direct measurements of real-time apolipoprotein association with and dissociation from lipoprotein particles. The fact that the sensorgrams for apoE interaction with VLDL and HDL3 are best fitted to a two-state model indicates that some sequential binding or conformational events occur. Given that apoE can exist in two lipid-bound states on spherical lipid particles (22;34) with the N-terminal four-helix bundle in either open or closed conformations (3;35;36), it is likely that this conformational change underlies the two-step kinetics seen when apoE binds to VLDL or HDL3. A model incorporating this concept is presented in Fig. 7. On this basis, the second ABx phase of binding for apoE3 or apoE4 corresponds, at least in part, to the rate of helix bundle unfolding. The fact that the isolated 12kDa C-terminal domain of apoE also exhibits a small ABx phase (Fig. 5C) indicates that other conformational changes besides helix bundle opening occur when apoE binds to VLDL or HDL3. Also, the C-terminal domain is not essential for initiating binding of apoE to a lipoprotein particle because the 22kDa N-terminal fragments of apoE3 and apoE4 can bind to some extent (Table 1).
The isotherms depicted in Fig. 4 together with the binding constants listed in Table 1 show clearly that the isolated 10kDa and 12kDa apoE C-terminal domains can bind in a saturable, high affinity, fashion to VLDL and HDL3. Consistent with the extreme C-terminal part of the apoE molecule being important for lipid binding, deletion of residues in the range 250–299 reduces lipoprotein association (24;25;31;33;37). The finding (Fig. 6) that such truncations cause bigger reductions in binding of apoE3 and apoE4 to VLDL than to HDL3 implies that lipid interactions play a bigger role in the former case. This effect might be expected because, compared to HDL3, a larger fraction of the VLDL particle surface contains exposed phospholipid molecules. In the case of HDL3 a large fraction of the particle surface is covered by the resident apolipoproteins so that apoE binding is more likely to involve protein-protein interactions.
Prior comparisons of apoE3 and apoE4 have attributed the differences in preference for binding to HDL3 and VLDL to differences in binding affinity rather than binding capacity (22;23;32). The relatively low affinity and lack of saturability of binding of apoE3 and apoE4 to VLDL and HDL3 indicated by the current SPR experiments makes detection of differences in binding affinity difficult. However, the binding parameters in Table 1 suggest that the variations in apoE isoform binding to VLDL and HDL3 are, at least in part, a consequence of differences in ability to be accommodated on the lipoprotein particle surface. Since the second ABx phase for apoE4 binding to VLDL is greater than that for apoE3 (cf. Fig. 2A and and5A),5A), it follows from the model in Fig. 7 that the helix bundle in apoE4 opens more readily. This effect might be expected because the N-terminal domain of apoE4 is less stable than that of apoE3 (29;30). On this basis, apoE4 would require more interfacial area than apoE3 (where the helix bundle would be more likely to remain closed and out of contact with the surface – see step 1 in Fig. 7) and consequently favor binding to the larger VLDL particle. The smaller HDL3 particle would more readily accommodate apoE3 molecules with only the C-terminal domain occupying interfacial area.
The fact that apoE binding to lipoprotein particles involves a two-step mechanism has physiological implications in that it explains the existence of exchangeable and non-exchangeable pools of apoE on the surface of lipoprotein particles. Thus, apoE molecules that are absorbed with the helix bundle domain closed (step 1 in Fig. 7) can exchange readily whereas those apoE molecules absorbed with the helix bundle open (step 2) do not exchange easily. Since the low density lipoprotein receptor (LDLR) binding site in apoE is not recognized when the helix-bundle is closed but only when it is open (3), it is expected that only molecules that achieve step 2 binding will act as effective ligands for the LDLR. This idea explains the fact that apoE in VLDL particles exists in two different conformations, one accessible and one inaccessible to the LDLR (38). Thus, apoE molecules bound with the helix bundle open are critical for apoE to achieve its major function of mediating the clearance of remnant lipoprotein particles from the circulation. Since apoE4 binds more than apoE3 to VLDL (Table 1) and it tends to bind in the helix bundle open conformation because the helix bundle is relatively unstable, it is expected that apoE4-VLDL will bind better than apoE3-VLDL to the LDLR. This difference in LDLR binding ability has been observed experimentally (39).
In summary, the application of SPR has provided new insights into the mechanism of apoE/lipoprotein interaction and into the contributions of the apoE tertiary structure domains to binding. ApoE binds to VLDL and HDL3 particles by a two-step mechanism with the rate being influenced by the stability and ease of unfolding of the N-terminal helix bundle domain. The mechanism and kinetics of lipoprotein interaction described here for apoE are likely to be relevant to the binding of other exchangeable apolipoproteins that have two-domain tertiary structures.
*This research was supported by NIH Grant HL56083