The role of a conserved acidic residue in calcium-dependent protein folding for a low density lipoprotein (LDL)-A module: implications in structure and function for the LDL receptor superfamily.

One common feature of the more than 1,000 complement-type repeats (or low density lipoprotein (LDL)-A modules) found in LDL receptor and the other members of the LDL receptor superfamily is a cluster of five highly conserved acidic residues in the C-terminal region, DXXXDXXDXXDE. However, the role of the third conserved aspartate of these LDL-A modules in protein folding and ligand recognition has not been elucidated. In this report, using a model LDL-A module and several experimental approaches, we demonstrate that this acidic residue, like the other four conserved acidic residues, is involved in calcium-dependent protein folding. These results suggest an alternative calcium coordination conformation for the LDL-A modules. The proposed model provides a plausible explanation for the conservation of this acidic residue among the LDL-A modules. Furthermore, the model can explain why mutations of this residue in human LDL receptor cause familial hypercholesterolemia.

The LDL 1 receptor superfamily, with human LDL receptor as its prototype, consists of a large number of proteins such as the LDL receptor-related protein (LRP), gp330, and the very low density lipoprotein receptor (1). One common feature of these proteins is that they all contain at least one modular domain (called complement-type repeat, or LDL-A module) of ϳ40 residues in length, including six invariable cysteines and the C-terminal highly conserved acidic residue motif (DXXX-DXXDXXDE). All together there are more than 1,000 known LDL-A modules found in a variety of proteins that are involved in diverse biological processes. In human LDL receptor, seven such imperfect repeats of LDL-A modules at the N terminus of the protein form the ligand binding domain, responsible for binding to its ligands, apoB and apoE (2)(3)(4)(5). Naturally occur-ring point mutations in any of the conserved acidic residues of the individual LDL-A modules of human LDL receptor can cause familial hypercholesterolemia (FH), a genetic disease that ultimately leads to coronary heart disease and atherosclerosis (6). One proposed mechanism for these conserved acidic residues of LDL receptor in ligand binding is to interact with the basic residues of its ligands via ionic interactions (2,(7)(8)(9).
Structural analysis of individual LDL-A modules, and recently the entire ectodomain of human LDL receptor, by x-ray crystallography revealed another mechanism by which four conserved acidic acids exert their role in protein conformation and function of LDL-A modules and thus LDL receptors (10 -12). Among the five conserved acidic residues, DXXXDXX-DXXDE, the side chains of four of them (the first, second, fourth, and fifth acidic residues, in italics), with the carbonyl oxygen groups of two non-conserved residues, are involved in calcium coordination. Thus mutations of these residues in the LDL receptor can result in folding defects of LDL-A modules and thus the overall structure of the LDL receptor, which indirectly lead to an impaired ligand-binding phenotype and eventually heart disease. However, the available biochemical and structural information on LDL-A modules and the ectodomain of the LDL receptor cannot explain why the third acidic residue is also highly conserved and why substitutions of this residue in different LDL-A modules of LDL receptor cause FH.
Tva is the cellular receptor for subgroup A of Rous sarcoma virus (RSV-A), and it is related to the LDL receptor superfamily because it contains a single LDL-A module of 40 amino acids in length within its extracellular region (13). Interaction between Tva and the RSV-A glycoprotein EnvA mediates viral entry. Tva specifically binds to the surface subunit (SU) of EnvA with high affinity, and this high affinity binding is important not only for viral attachment to the host cells but also important for receptor-triggered conformational changes on EnvA, which are essential for EnvA-mediated fusion between the viral and host membranes (14 -17). Extensive molecular and biochemical analysis of Tva/EnvA interaction has demonstrated that the viral interaction domain of Tva is solely determined by the LDL-A module (18 -22).
In this study, we examined the role of the third conserved acidic residue in protein folding and ligand binding using a model LDL-A module. This system is based on our previous work that demonstrated that the Tva LDL-A module can be functionally substituted by the human LDL receptor repeat 4 (hLDL-A4) with minor modifications (19). A unique and important aspect of this system is that several functional assays can be used to dissect the roles of individual residues in both protein folding and ligand recognition, in addition to the in vitro biochemical and structural analysis of individual LDL-A modules. This study reveals that the third conserved acidic residue, like other four acidic residues, is involved in calcium coordination and protein folding, providing a molecular explanation why this residue is conserved in the majority of hundreds of LDL-A modules, and why mutations of this residue in LDL receptor can lead to FH. Furthermore, binding kinetics analysis between the model LDL-A module and its ligand demonstrates that the second and third conserved residues of this LDL-A module are not directly involved in ligand binding. These results, although with potential caveats of a model system, should have broad implications in LDL-A module/ ligand interactions.

EXPERIMENTAL PROCEDURES
DNA Methodology-A chimeric construct (referred to as Chimera in this study) was generated by replacing the C1-C3 region of TL4G-A19L/D23H (19) with the corresponding region of the Tva LDL-A module by an overlapping PCR protocol. This construct was inserted into the LDL-A module of Tva using the myc-Tva backbone (19). The conserved acidic residues of Chimera were substituted with alanines either individually or in combinations (see Fig. 2A). All the constructs were confirmed by DNA sequencing. These constructs were used for examining protein expression and viral receptor function in human embryonic kidney 293T cells by transient transfection as described previously (18).
Protein Expression and Viral Infection Assay-Human embryonic kidney 293T cells were transiently transfected with DNAs of Chimera and its mutants by the CaPO 4 method as described previously (18). The transfected cells were lysed with Triton Lysis Buffer 48 h post-transfection, and protein expression of these constructs was examined by Western blotting following SDS-PAGE, using monoclonal antibody 9E10, which specifically recognizes the Myc tag portion of these proteins. Furthermore, to examine whether these proteins were expressed on the cell surface, transiently transfected 293T cells were subjected to flow cytometry. Briefly, 24 h post-transfection, cells were lifted, and 1 ϫ 10 6 cells were washed three times with FACS buffer (1ϫ phosphatebuffered saline, 1% fetal bovine serum, and 0.1% sodium azide). Cells were resuspended in 100 l of FACS buffer containing 1 l of mouse monoclonal anti-Myc antibody 9E10 and incubated at 4°C for 1 h. The cells were then washed and resuspended in 100 l of FACS buffer containing 1 l of goat anti-mouse antibody conjugated with fluorescein isothiocyanate and incubated at 4°C for 1 h. Finally, cells were washed and resuspended in 500 l of FACS buffer containing 1 l of propidium iodide. The samples were collected and analyzed using a BD Biosciences FACSCalibur flow cytometer and Cellquest software.
To determine the viral receptor function of Chimera and its mutants, transiently transfected 293T cells were challenged with RCAS(A)AP, the recombinant RSV-A viruses, following a previous protocol (18). The alkaline phosphatase-positive 293T cells were detected under a microscope after staining, enumerated, and presented as number of alkaline phosphatase-positive cells/per ml viral stock used.
Expression and Characterization of the in Vitro Folding Properties of the LDL-A Proteins-The coding regions of the LDL-A modules of Chimera and its mutants were PCR-amplified using the DNA templates of these constructs in pcDNA-3. The amplified DNA fragments were digested with restriction endonucleases and cloned into pGEX-4T-1 (Amersham Biosciences), and the identity of each construct was confirmed by DNA sequencing. The LDL-A modules of Chimera and its mutants were expressed in Escherichia coli strain BL21 as glutathione S-transferase fusion proteins and purified by the GSH affinity column. After thrombin cleavage, the glutathione S-transferase portion was removed by the GSH affinity column, and the protein samples were further purified by reverse phase high performance liquid chromatography (HPLC) following a previous protocol established by us (23).
The purified proteins were refolded in the absence or presence of calcium following a published protocol (23), and the folding properties of the LDL-A proteins were examined by reverse phase HPLC and twodimensional NMR spectroscopy. Its Ca 2ϩ binding affinity was measured by isothermal titration calorimeter (Microcal MSC), following a previous protocol with minor modifications. The purified LDL-A proteins were refolded in the absence or in the presence of different concentrations of CaCl 2 (23). Following the refolding reactions, the samples were analyzed by reverse phase HPLC on a Vydac C 18 column operated at a flow rate of 3.00 ml/min, using a linear gradient of 0.1% trifluoroacetic acid and 90% acetonitrile.
The 15 N-labeled LDL-A proteins were used to acquire the [ 1 H-15 N]-HSQC spectra. Following the refolding step as described above, each protein sample was eluted by reverse phase HPLC, and individual peaks were collected and prepared for acquisition of two-dimensional MNR spectra as described previously (23). Briefly, the NMR data were collected on a Bruker DRX600 spectrometer equipped with a pulse-field gradient accessory and operating at 600.13 MHz for 1 H and were processed and analyzed using Triad 6.3. The central frequencies were 4.70 and 118 ppm for 1 H and 15 N, respectively.
Determination of the Binding Kinetics by IAsys-A resonant mirror biosensor (IAsys Autoϩ, Affinity Sensors) with a carboxylate cuvette was used to determine the kinetic constants (k a and k d ) and binding affinities (K D ) of the LDL-A proteins with the SUA-rIgG, which was transiently expressed in 293T cells and purified from the supernatant (17,21). All the experiments were performed following a published protocol with modifications (17). Briefly, the LDL-A proteins were first refolded with calcium and then purified by reverse phase HPLC. The fractions of well folded LDL-A proteins, determined by two-dimensional NMR, were prepared and immobilized to the [1-ethyl-3-(3-dimethylaminopropyl)carbodiimide HCl/N hydroxysuccinimide]-activated carboxylate cuvettes following the manufacturer's protocol. The un-reacted N-hydroxysuccinimide esters and uncovered surface area were blocked by 2 mg/ml ␤-casein. The tightly associated but non-covalently bound protein was removed with 10 mM HCl. Binding of the SUA-rIgG protein to the immobilized LDL-A proteins was performed in phosphate-buffered saline with additional 100 mM NaCl (to minimize the nonspecific binding) with 250 M CaCl 2 . The binding kinetics were examined by monitoring the association phase for 5 min and followed by monitoring the dissociation phase for 4 min. After each cycle, the cuvettes were regenerated with 100 mM HCl for 2-5 min. The binding cycle was repeated by using 6 -8 different concentrations of the SUA-rIgG protein dissolved in phosphate-buffered saline, 100 mM NaCl with 250 M CaCl 2 . The experimental data were processed using FASTfit software (Affinity Sensor). The association constant (k a ) was calculated from the gradient of the plot of apparent rate constant (k on ) versus the SUA-rIgG protein concentration, which was obtained by a linear fit. The dissociation constant (k d ) was determined from the intercept of the k on versus [ligate] plot. When data were poorly fitted to a single exponential phase, a double exponential phase was used to process the association and dissociation data.
Molecular Modeling of the Tva LDL-A Module-The newly proposed calcium-dependent conformation of the Tva LDL-A module, shown in Fig. 5B, was calculated using DYANA (24), a torsion angle dynamics annealing simulation program, following the same procedures as performed previously for the published structure of the Tva LDL-A module (25), shown in Fig. 5A.

Five Acidic Residues Are Highly Conserved at the C Terminus of the LDL-A Modules
Sequence analysis of the LDL-A modules from different members of the LDL receptor superfamily indicates that five acidic residues near the C terminus, DXXXDXXDXXDE, are highly conserved among the LDL-A modules, and they will be referred to as D1, D2, D3, D4, and E5, respectively, in this report. Examination of 132 LDL-A modules from human LDL receptor, very low density lipoprotein receptor, LRP, complements, perlecan, sortilin-related receptor, corin, enterokinase, membrane serine proteases 1 and 2, and Tva, the RSV-A receptor, revealed that D4 and E5 are present in all of the modules and thus are invariable. In contrast, residue variations are found at the three remaining positions. Substitution of the aspartic acid by an asparagine accounts for all the variations (25 of the 132 LDL-A modules) at the D1 position. However, although the total number of residue variations at the D2 and D3 positions is less than that at D1 (16 of 132 for D2, 9 of 132 for D3, respectively), other residues (His, Ser, Gln, Glu, Tyr, Thr, and Gly), in addition to asparagine, can be found at these positions. Fig. 1 shows the sequence alignment of 11 LDL-A modules. Because these positions are highly conserved among so many LDL-A modules, it is reasonable to hypothesize that all of these acidic residues play important roles in LDL-A folding and/or ligand binding.
An important role for four of the five conserved acidic residues has been elucidated elegantly by the x-ray crystal structure of human LDL-A5 (10). Residues Asp-25 (D1), Asp-29 (D2), Asp-35 (D4), and Glu-36 (E5) of LDL-A5 were shown to be involved in calcium coordination via their side chains, thus providing an explanation of why these acidic residues are conserved in LDL-A modules. In contrast, the role of D3 in LDL-A folding and function is not clear. Conservation of this residue in most LDL-A modules strongly suggests an important role. Furthermore, D3 mutations in human LDL receptor cause FH (6). These data support the notion that D3 plays an important role in protein folding and/or function of LDL-A modules.

Substitutions of Both the Second and Third Conserved Acidic Residues in the LDL-A Module Display a Defective Phenotype in Mediating RSV-A Entry
It has been demonstrated previously that the LDL-A module of Tva could be functionally replaced by the fourth LDL-A module (hLDL-A4) of human LDL receptor with a few substitutions to mediate RSV-A entry (19). In addition, the N-terminal region of the Tva LDL-A module (between cysteines 1 and 3) is also required for optimal receptor activity (26). Thus, a chimeric construct, referred to as Chimera in this study, was generated by appending the C1-C3 region of the Tva LDL-A module with the C3-C6 region of the modified hLDL-A4 module ( Fig. 2A), using myc-Tva backbone plasmid. This chimeric protein is indistinguishable from the wt Tva in mediating efficient RSV-A viral entry (see Fig. 2B) and, importantly, retains all of the aforementioned conserved acidic residues.
Each of the five conserved acidic residues of Chimera, Asp-26 (D1), Asp-30 (D2), Asp-33 (D3), Asp-36 (D4), Glu-37 (E5), and Glu-32, which is not conserved among the LDL-A modules, was individually substituted by alanine ( Fig. 2A). The effect of these mutations on RSV-A entry was determined by challenging the 293T cells transiently expressing these constructs with RCAS(A)-AP viral vector, a recombinant RSV-A virus that contains an alkaline phosphatase reporter gene (27). As shown in Fig. 2B, Chimera displays an indistinguishable phenotype from the wt myc-Tva in mediating RSV-A entry at three different plasmid concentrations (1, 5, and 20 g, respectively) in transfection. As expected, substitution mutant of Glu-32, a nonconserved acidic residue, did not display a defective viral receptor phenotype (Fig. 2B). In contrast, mutants D26A, D36A, and E37A were 100 -1000-fold less efficient than wt Tva or Chimera in mediating RSV-A entry, depending on the amounts of plasmid DNA used in transfection. These results are consistent with the notion that these residues are involved in calcium coordination; thus it is expected that substitution of these residues would cause protein misfolding, leading to the defected phenotype. However, surprisingly, substitution of Asp-30 (D30A), another conserved acidic residue implicated in calcium coordination, did not have a detectably adverse effect on the viral receptor function of Chimera (Fig. 2B), unlike that of the other three calcium coordination acidic residues. Similarly, substitution of the third conserved acidic residue (D33A) did not adversely affect the ability of Chimera to mediate efficient viral entry.
Three double substitution mutants (D30A/E32A, D30A/ D33A, and E32A/D33A) were generated to further investigate the potential role of the second and third conserved acidic residues in RSV-A entry ( Fig. 2A). When 293T cells transiently expressing these mutant proteins were challenged with the recombinant RSV-A viruses, only mutant D30A/D33A displayed greatly impaired viral receptor function (ϳ3-4 logs lower than wt Tva). In contrast, the remaining two mutants (D30A/E32A and E32A/D33A) could still mediate efficient RSV-A entry at high DNA concentrations, and only displayed slightly lower receptor function at the lowest DNA concentration (1 g) in transfection (Fig. 2B). These results appear to suggest that the role of the second and third conserved acidic residues is redundant in mediating RSV-A entry. However, as demonstrated below, substitution of each of these residues adversely affects calcium-dependent protein folding. This discrepancy between the viral receptor function and protein folding of this Chimera will be explained under ''Discussion.'' Total protein expression of Chimera and its mutants in 293T cells was examined by Western blotting, and it was found that all of the constructs were well expressed (data not shown). Furthermore, FACS analysis of surface expression of these proteins showed that they were all surface-expressed (data not shown), indicating that the defect of some Chimera mutants in mediating RSV-A entry is not due to problems of surface expression.

Substitutions of Conserved Acidic Residues Result in Protein Folding Defects, HPLC Profiles
To characterize further the roles of these conserved acidic residues in LDL-A folding, the LDL-A modules of Chimera and its mutants were expressed as glutathione S-transferase fusion proteins in Escherichia coli and purified by affinity chromatography as described under ''Experimental Procedures.'' Following thrombin cleavage and further purification, in vitro folding properties of the LDL-A modules of Chimera and its mutants were first examined by reverse phase HPLC after refolding in the presence or absence of calcium, following a protocol published previously.
The Chimera protein was eluted as multiple peaks when CaCl 2 was omitted in the refolding. In contrast, the same protein was eluted as a single sharp peak (labeled as peak 2) as the CaCl 2 concentrations were increased from 2 to 10 or 50 mM in the refolding step (Fig. 3). These results are consistent with the notion that the correct folding of Chimera, like that of Tva and the other LDL-A modules characterized previously by others and us, is calcium-dependent. Similarly, substitution mutant of a non-conserved acidic residue (E32A) of Chimera gave an almost identical HPLC profile as Chimera, suggesting that Glu-32 is unlikely involved in calcium coordination.
In contrast, the HPLC profiles of the substitution mutants of the five conserved acidic residues in Chimera showed that none of these mutant proteins was eluted as a single sharp peak even as the CaCl 2 concentration was increased to 50 mM in the refolding step; instead, multiple peaks were eluted (Fig. 3). These results suggest that any individual substitution of these residues can adversely affect calcium-dependent protein folding. Because Asp-26, Asp-30, Asp-36, and Glu-37 have been implicated in calcium coordination, it is reasonable to assume that substitutions of these residues by alanines disrupted the native calcium cage of Chimera and thus resulted in protein misfolding. However, the elution profiles of mutants D30A and D33A are different from that of mutants D26A, D36A, and Glu-37. The proportion of peak fraction 4 (D30A) increased as more calcium was added in the refolding step (from 2 to 50 mM), whereas the percentage of other peaks such as peak fraction 5 became less prominent with higher concentrations of calcium. Similarly, the percentage of peak fraction 7 of D33A became more predominant than the other peaks (e.g. peak fraction 8) with higher concentrations of calcium. In contrast, peak fractions of 14 (D26A), 16 (D36A), and 18 (E37A) did not become so prominent even with higher calcium concentrations (Fig. 3). More important, as it will become clear in the two-dimensional NMR spectra (see below), the folding defect of D30A and D33A could be partially compensated by higher concentrations of calcium in the refolding step, and more than half of the protein samples (peak fractions 4 and 7, respectively) could be well folded in the presence of 50 mM CaCl 2 . However, we could not detect any well folded fractions of D26A, D36A, E37A, or D30A/ D33A mutant proteins under the same conditions.

Substitutions of Conserved Acidic Residues Result in Protein Folding Defects, Two-dimensional NMR Spectra
Two-dimensional [ 1 H-15 N]-HSQC NMR spectroscopy was used previously to demonstrate that Ca 2ϩ is not only required for correct folding but also for maintaining the structural integrity of the Tva LDL-A module (23). Here we used the same technique to carefully examine the NMR spectra of 18 individual peak fractions indicated in Fig. 3 (numbered 1-18). The   FIG. 2. A, mutants of a chimeric LDL-A module. The six acidic residues at the C-terminal region of the module are substituted by alanines either individually or in combination. On top is the sequence alignment of human LDL receptor repeat 4 (LDL-A4) and the quail Tva LDL-A module. The acidic residues of Chimera targeted for mutagenesis are highlighted by shadows. B, effect of the mutations of the acidic residues of Chimera on viral entry. The ability of each construct to mediate RSV-A infection was assayed by using a recombinant viral vector, RCAS(A)-AP. Human 293T cells were transfected with different amounts of DNAs of the wt Tva (myc-Tva), Chimera, and its mutants. The transfected cells were challenged by the recombinant virus, and the ability of the receptor function of these constructs to mediate viral entry is expressed as the number of positive alkaline phosphatase-staining cells/ml of viral stock used (IU/ml).
NMR spectra, as shown in Fig. 4, can be roughly classified into three groups: 1) Chimera and E32A, 2) D30A and D33A, and 3) the remaining four mutants, D26A, D36A, E37A, and D30A/ D33A. The major features of these classes are described below.
Chimera and E32A-When the proteins were folded in the absence of calcium, the peak fractions gave clustered [ 1 H-15 N]-HSQC NMR spectra (Fig. 4, A1 and E11), indicating that these peaks were not well dispersed. In contrast, when the proteins were folded in the presence of calcium, the peak fractions gave well dispersed spectra (A2 and E12), suggesting that these peaks were well structured. These results demonstrate that calcium is required for proper folding for Chimera and E32A proteins, as expected.
D30A and D33A-Like Chimera and E32A, when the proteins were folded in the absence of calcium, the peak fractions gave clustered NMR spectra (Fig. 4, B3 and C6). When the proteins were folded in the presence of calcium, peak fractions 4 (D30A) and 7 (D33A) gave well dispersed spectra (B4 and C7), suggesting that the proteins were well structured. However, peaks 5 (D30A) and 8 (D33A) were not well dispersed (B5 and C8), suggesting that these fractions were not well folded. Therefore, we can conclude that under these conditions, only a fraction of D30A and D33A proteins could be correctly folded.
D26A, D36A, E37A, and D30A/D33A-The peak fractions of these mutant proteins gave clustered NMR spectra regardless that the proteins were folded in the presence or absence of calcium (Fig. 4, D9, D10, F14, F14, G15, G16, H17, and H18). These results suggest that substitutions of these acidic residues disrupted overall protein folding due to the defect of calcium coordination.

Substitutions of Conserved Acidic Residues Result in Protein Folding Defects, Calcium Binding Affinities
Isothermal titration calorimetry was used to directly measure calcium binding affinities of Chimera and its mutant proteins following a published protocol (23). Calcium titrations were performed on a Microcal MSC isothermal titration calorimeter. Chimera and E32A gave K D values of 72.7 Ϯ 7.69 and 46.2 Ϯ 7.43 mM, respectively. However, under the same conditions, none of the other Chimera mutant proteins displayed any detectable calcium binding activity probably due to the sensitivity limit of this assay (data not shown). Nevertheless, these results clearly demonstrated that substitution of any of the five conserved acidic residues in Chimera adversely affect calcium binding, consistent with the notion that all five conserved acidic residues are involved in calcium coordination.

The Second and Third Conserved Acidic Residues Are Not Involved in Ligand Binding
One interesting question is whether the conserved acidic residues of LDL-A modules are involved in ligand binding in addition to their role in calcium coordination and thus protein folding. This is not an easy question to answer because any substitution of the conserved acidic residues, as demonstrated above, can result in calcium-dependent folding defects, and therefore it is experimentally difficult to determine whether these acidic residues are specifically involved in ligand binding without structural analysis of LDL-A-ligand complexes. Nevertheless, the results described above provide a unique opportunity for us to investigate whether some of the acidic residues of Chimera are directly involved in EnvA binding.
Analysis of Chimera mutant proteins by HPLC and twodimensional NMR demonstrated that a fraction of D30A and D33A mutant proteins were well folded in the presence of calcium. However, D26A, D36A, and E37A mutant proteins were not well folded under the same refolding conditions. Thus, it was possible for us to purify the well folded D30A and D33A proteins by HPLC (fraction peaks 4 and 7 of Fig. 3) and use these samples to measure the binding affinities to SUA-rIgG, as done by us previously (17) for Tva mutant proteins using IAsys. The peak fractions of Chimera, E32A, D30A, and D33A (fraction peaks 2, 12, 4, and 7 of Fig. 3) were collected and prepared for SUA-binding kinetic analysis using IAsys. The immobilized receptor peak samples were incubated with the SUA-rIgG protein, and the binding kinetics were determined in the presence of calcium as described under ''Experimental Procedures.'' Chimera and E32A proteins gave K D values of 20.9 and 15.9 nM, respectively, with similar values of k a and k d ( Table I). E32A mutant displayed same binding kinetics as Chimera, indicating that Glu-32 is not involved in calcium coordination or ligand binding. Importantly, the well folded fractions of D30A and D33A proteins bound to SUA-rIgG with very similar kinetics as Chimera and E32A. D30A and D33A gave K D values of 22.7 and 27.9 nM, respectively. These results indicate that Asp-30 and Asp-33 of Chimera are not directly involved in SUA binding. However, we were unable to determine whether the other three conserved acidic residues, Asp-26, Asp-36, and Glu-37, are directly involved in SUA binding because it was not possible to purify well folded mutant proteins of mutants D26A, D36A, or E37A (see Figs. 3 and 4). DISCUSSION In this study, we revealed the role of the third conserved acidic residue of a model LDL-A module in calcium-dependent protein folding and ligand recognition. This finding should have very broad implications in protein folding and function for all members of the LDL receptor superfamily. Furthermore, this finding provides a plausible explanation for a class of FH mutants that carry the substitutions of the third conserved acidic residue in the human LDL receptor. A unique aspect of this study is that we were able to use several functional assays to examine the role of each of the five conserved acidic residues of this module in mediating ligand recognition (EnvA binding) and viral infection (RSV-A entry). Because four of the five conserved acidic residues in the LDL-A module have been implicated in calcium coordination, we hypothesized that mutations of these residues would disrupt the proper folding and conformation of the LDL-A module, and the folding defect could be easily detected by the functional assays. Indeed, substitution of each of the three conserved acidic residues, Asp-26 (D1), Asp-36 (D4), or Glu-37 (E5), impaired the ability of the LDL-A module to efficiently mediate RSV-A entry (see Fig. 2). However, substitution of Asp-30 (D2) or Asp-33 (D3) did not appear to have any adverse effect on the LDL-A module in mediating RSV-A entry (see Fig. 2). Further analysis indicated that substitution of both Asp-30 and Asp-33 (D3) was required to abrogate the ability of the LDL-A module to mediate RSV-A entry, suggesting that Asp-30 and Asp-33 of the LDL-A module play a somewhat different role in protein folding from the other three conserved acidic residues.
The in vitro protein folding analysis revealed that both Asp-30 and Asp-33 are also involved in calcium-dependent protein folding, but they displayed unique properties in the HPLC elution profiles and the two-dimensional NMR spectra, generally consistent with the results of the viral entry assay. However, in stark contrast to the in vitro folding properties of the substitution mutants of the other three conserved acidic residues, a fraction of the D30A and D33A mutant proteins could be correctly refolded in the presence of calcium, and the percentage of the corrected folded portion appeared to increase with higher concentrations of calcium (see Figs. 3 and 4). These results indicated that the folding defect of D30A and D33A mutant proteins could be partially compensated with high concentrations of calcium in the refolding steps. The same observation has been observed by others who demonstrated that the folding defect of a point mutation in human LDL receptor repeat 5 (D203G), which corresponds to the D3 position of the conserved acidic residue, could be compensated by higher concentrations of calcium (28), supporting the data in the current study. Together, these results strongly suggest that both D2 and D3, like the other three conserved acidic residues, are involved in calcium-dependent protein folding.
It should be pointed out that some discrepancies have been observed regarding the protein folding and function of the LDL-A module. The most dramatic difference is that a single mutation of either Asp-30 or Asp-33 did not show a noticeable effect on the receptor function of the LDL-A module, whereas both mutant proteins displayed a folding defect in vitro. However, this discrepancy can easily be reconciled by the following explanation. Tva expresses at a very low level in avian cells, but RSV-A can efficiently infect these cells, indicating that high expression of the receptor is not a prerequisite for efficient viral infection. In contrast, in the current study, the mutant proteins were highly expressed in mammalian cells by transient transfection. Overexpression of certain mutant proteins (like D30A or D33A) may mask the defect phenotype of these mutations, because a portion of the mutant proteins could be correctly folded in vivo and supported efficient RSV-A entry (20). As revealed by in vitro folding analysis, a fraction of these mutant proteins could be indeed correctly folded with higher calcium concentrations, consistent with our previous assumption.
We believe that the current study has important implications in elucidating the mechanism of protein folding and function of most, if not all, LDL-A modules. Because residues D2 and D3 displayed highly similar properties by both in vivo and in vitro analyses in this study, we can conclude that the third conserved acidic residue (D3), like the second one (D2), is also involved in calcium-dependent protein folding. We propose that each LDL-A module can adopt two calcium-dependent conformations. In one conformation, the side chains of D1, D2, D4, and E5 are involved in calcium binding as demonstrated by the x-ray structures of several individual LDL-A modules, whereas in another conformation, the side chains of D1, D3 (instead of D2), D4, and E5 coordinate calcium binding. In both conformations, the carbonyl oxygen groups of the two additional residues are also involved in calcium coordination. Here the Tva LDL-A module is used to illustrate the proposed two calcium-depend-ent conformations (Fig. 5). Analysis using a molecular modeling technique (DYANA, see Ref. 24) indicates that both conformations are thermodynamically equally favorable, supporting the proposed models (data not shown).
Although the involvement of D3 in calcium-dependent conformation of the LDL-A modules has not been directly demonstrated by x-ray structural analysis, probably due to the crystallization conditions such as pH and calcium concentrations, the proposed models give a plausible explanation why D3 is highly conserved in LDL-A modules, and why substitutions of this residue in several LDL-A modules of LDL receptor cause FH. Five FH point mutations at the D3 position in three LDL-A modules of human LDL receptor have been reported (6). These mutations include D154N (LDL-A4), D203N and D203G (LDL-A5), and D283N and D283E (LDL-A7). We speculate that these mutations, like mutations in D2, cause partial defects in calciumdependent folding of LDL receptor and thus impair its ability to interact with LDL particles, leading to FH in patients carrying these substitutions.
A possible role of the conserved acidic residues of the LDL-A modules in direct ligand interaction has not been well defined. It has been long proposed that apolipoprotein binding to the LDL receptor involves ionic interactions between the basic residues of the apolipoproteins and the conserved acidic residues of LDL receptor. This hypothesis is based on several observations. As discussed above, naturally occurring mutations (or substitutions by mutagenesis) of the conserved acidic residues in human LDL receptor can impair the receptor function (4,6). The x-ray structure of the LDL receptor binding domain of apoE revealed a cluster of positively charged residues (9), and mutations of these residues impaired the ability of apoE to bind the LDL receptor (29,30). It has been suggested that similar ionic interactions play an important role in ligand binding of LRP (31). However, the x-ray structures of several LDL-A modules revealed that the conserved acidic residues are involved in calcium coordination (10 -12), thus appearing not available to directly interact with the basic residues of the ligands. Several possible mechanisms have been suggested by Brown et al. (32) to explain this paradox. However, there is no report to demonstrate if the conserved acidic residues of an LDL-A module can directly participate in ligand binding by ionic interactions. Technically, it is difficult to determine a direct role of the conserved acidic residues of the LDL-A modules of LDL receptor in ligand binding, because these residues are involved in protein folding, and thus substitutions of these residues would disrupt the correct folding of the protein and indirectly influence ligand interaction. This issue has been partially addressed in the current study using a model LDL-A module. We have shown previously that the certain basic residues of EnvA, like the basic residues of apoE in binding to the LDL receptor, are required for receptor (Tva) binding (33). In this study, we have demonstrated that a portion of D30A and D33A mutant proteins could be correctly folded in the presence of calcium in vitro, and the correctly folded portions could be separated from the other fractions and purified by reverse phase HPLC. Using IAsys, an optical biosensor, we have demonstrated that the correctly folded D30A and D33A mutant proteins bound to SUA with similar dissociation constants as Chimera (see Table I). These results clearly showed that neither residue Asp-30 nor Asp-33 is directly involved in ligand interaction, although both of them are involved in calcium coordination and thus protein folding of the module. However, we were unable to determine whether the other three conserved acidic residues (Asp-26, Asp-36, and Glu-37) are directly involved in ligand binding because it was not possible to obtain the correctly folded mutant proteins of these residues under the same experimental conditions. Therefore, the experiments in this study, for the first time, allowed us to exclude a role of Asp-30 and Asp-33 of the Chimera module in direct involvement of ligand interaction. Although these results should be interpreted with caution because they were derived from a model LDL-A module, these results may have important implications in LDL receptor/ligand interactions in general.