Characterization of the role of EGF-A of low density lipoprotein receptor in PCSK9 binding.

Proprotein convertase subtilisin kexin-like 9 (PCSK9) promotes the degradation of low density lipoprotein receptor (LDLR) and plays an important role in regulating plasma LDL-cholesterol levels. We have shown that the epidermal growth factor precursor homology domain A (EGF-A) of the LDLR is critical for PCSK9 binding at the cell surface (pH 7.4). Here, we further characterized the role of EGF-A in binding of PCSK9 to the LDLR. We found that PCSK9 efficiently bound to the LDLR but not to other LDLR family members. Replacement of EGF-A in the very low density lipoprotein receptor (VLDLR) with EGF-A of the LDLR promoted the degradation of the mutant VLDLR induced by PCSK9. Furthermore, we found that PCSK9 bound to recombinant EGF-A in a pH-dependent manner with stronger binding at pH 6.0. We also identified amino acid residues in EGF-A of the LDLR important for PCSK9 binding. Mutations G293H, D299V, L318D, and L318H reduced PCSK9 binding to the LDLR at neutral pH without effect at pH 6.0, while mutations R329P and E332G reduced PCSK9 binding at both pH values. Thus, our findings reveal that EGF-A of the LDLR is critical for PCSK9 binding at the cell surface (neutral pH) and at the acidic endosomal environment (pH 6.0), but different determinants contribute to efficient PCSK9 binding in different pH environments.


Degradation of receptors by PCSK9
The degradation experiment was performed as described previously ( 18 ). Briefl y, the mouse hepatoma cell line, Hepa1C1C7, was cultured in MEM ␣ medium containing 10% (v/v) FBS at 37°C and seeded in 12-well dishes (1.5 × 10 5 cells/well). After 24 h, cells were transiently transfected with pCDNA3.1 or expression plasmids containing cDNAs for wild-type or mutant LDLR or VLDLR using Lipofectamine 2000 according to the manufacturer's protocol. Forty-eight hours after transfection, cells were washed, incubated in 0.5 ml of MEM ␣ medium containing 5% (v/v) newborn calf lipoprotein-poor serum, 10 g/ml cholesterol, 1 g/ml 25-hydroxycholesterol, and various amounts of purifi ed wild-type or mutant PCSK9 (D374Y) for the indicated time. Cells were then washed three times with ice-cold PBS and lysed in 60 l of lysis buffer. Whole-cell lysate protein extracts were then analyzed by SDS-PAGE (8%) and immunoblotted using a monoclonal anti-hLDLR antibody (HL-1).

Purifi cation of GST:EGF-A fusion protein and ligand blotting
Wild-type EGF-A of the LDLR was expressed as recombinant glutathione-S-transferase (GST) fusion proteins using the vector pGEX-4T (GE Healthcare) in Escherichia coli BL-21-DE3 cells (EMD Biosciences, San Diego, CA) and purifi ed as described ( 17 ). Briefl y, the transformed cells were grown at 37°C, induced with 1 mM isopropylthio-␤ -galactoside and then harvested. The cells were lysed using a French pressure cell. The GST:EGF-A fusion protein was purifi ed using Glutathione Sepharose 4 Fast Flow (GE Healthcare) affi nity gel chromatography according to the manufacturer's protocol. The protein was concentrated and further purifi ed using size-exclusion chromatography on a Tricorn Superose 12 10/300 fast-performance liquid chromatography column (GE Healthcare). Fractions containing GST:EGF-A were concentrated using a 3 kDa MW (molecular mass) cut-off Centriplus fi lter. Protein purity was monitored by SDS-PAGE and Coomassie Brilliant Blue R-250 staining (Bio-Rad, Hercules, CA).
are not required for PCSK9-promoted LDLR degradation ( 27 ). We have shown that PCSK9 interacts with the epidermal growth factor precursor homology domain A (EGF-A) of the LDLR at the cell surface and binds to the full-length receptor with a much higher affi nity in the acidic environment of the endosome. Consequently, the receptor transports from the endosome to the lysosome for degradation, rather than being recycled ( 17 ). Consistently, the crystallographic structures of PCSK9 and the EGF-AB of the LDLR complex reveal that the N terminus of EGF-A is associated with the catalytic domain of PCSK9 (28)(29)(30). We also demonstrate that leucine at position 318 in EGF-A of the LDLR is critical for effi cient binding of PCSK9 ( 17 ). The replacement of Leu 318 in the LDLR with Asp, as it is in the very low density lipoprotein receptor (VLDLR), significantly reduces binding of PCSK9 to the LDLR. Here we further characterized the role of EGF-A of the LDLR in PCSK9 binding to the receptor. We found that Gly 293 , Asp 299 , Arg 329 , and Glu 332 in EGF-A of the LDLR contributed to PCSK9 binding at the cell surface. We also found that PCSK9 bound to recombinant EGF-A in a pH-dependent way with a stronger binding at pH 6.0.

Materials
Lipofectamine 2000 and cell culture medium were obtained from Life Technologies. Fetal bovine serum (FBS) was purchased from Sigma. Complete EDTA-free protease inhibitors and X-tremeGENE HP were from Roche. The QuickChange site-directed mutagenesis kit was obtained from Agilent Technologies. All other reagents were obtained from Fisher Scientifi c unless otherwise indicated.
The recombinant wild-type human PCSK9 or mutant PCSK9 D374Y containing a FLAG tag (DYKDDDDK) at the C terminus was purifi ed from HEK-293S cells as described ( 31,32 ). The extracellular domain of the LDLR (LDLR-ECD) (amino acids 1-699) contains a six histidine residue tag at the C terminus and was purifi ed exactly as described ( 33 ). LDLR-ECD and PCSK9 were labeled with IRDye680 and IRDye800, respectively, using IRDye protein labeling kits (LI-COR Biosciences) according to the manufacturer's protocol, and the proteins were visualized using an Odyssey infrared imaging system (LI-COR Biosciences).

Site-directed mutagenesis
A recombinant expression vector containing the full-length LDLR cDNA linked to pCDNA5 was used to generate the mutant forms of the LDLR using the QuikChange TM site-directed mutagenesis kit according to the manufacturer's instructions. The VLDLR and low density lipoprotein receptor-related protein (LRP)6 expression constructs contained one copy of a hemaglutinin epitope (HA) tag (CYPYDVPDTAG) at the C terminus. The oligonucleotides containing the residues to be mutated were synthesized by IDT, Inc. (Coralville, IA). The presence of the desired mutation and the integrity of each construct were verifi ed by DNA sequencing.

Binding of PCSK9 to the LDLR family members
The binding assay was performed as described previously ( 17 ). Briefl y, COS-7 cells were cultured in Dulbecco's modified Eagle's medium (DMEM):Ham's F12 medium (1:1 mixture) containing 10% (v/v) FBS at 37°C and seeded in 12-well dishes (1.5 × 10 5 cells/ conserved among LDLR, VLDLR, apoER2, and LRP1. Leucine is only present in LDLR ( Fig. 1A ). Thus, we expressed these LDLR family members in monkey kidney cells (COS-7 cells) and examined their ability to bind PCSK9. COS-7 cells were used to study PCSK9 binding to the LDLR because PCSK9 can bind to the LDLR but cannot induce the degradation of the receptor in this cell line ( 37 ). In addition, endogenous LDLR in COS-7 cells is low, which reduces background PCSK9 binding. The expression of different LDLR family members was detected by their specifi c antibodies. As shown in Fig. 1B , when purifi ed PCSK9 (2 g/ml) was added to the medium and incubated with the cells for 2 h at 37°C (pH 7.4), the PCSK9 association signal was detected only in cells expressing LDLR ( Fig. 1B , top, lane 2), but not in cells expressing other LDLR family members ( Fig. 1B , lanes 3-7). Next, we investigated if EGF-A of the LDLR was suffi cient to confer PCSK9-promoted receptor degradation. We have reported that VLDLR:EGF-A-LDLR, in which EGF-A of the VLDLR is replaced with EGF-A of the LDLR, binds to PCSK9 very effi ciently. However, binding of PCSK9 to LDLR:EGF-A-VLDLR, in which EGF-A of the LDLR is substituted by EGF-A from the VLDLR, is dramatically reduced ( 17 ). Therefore, the two chimeric proteins were transiently expressed in mouse hepatoma cells (Hepa1C1C7) and incubated with PCSK9 (0.5 g/ml) for 4 h at pH 7.4. The VLDLR that had a HA tag at its C terminus was detected by an anti-HA antibody, which recognizes both the precursor (p) and the mature (m) fully glycosylated form of the receptor ( Fig. 1C ). We observed that addition of PCSK9 resulted in robust degradation of the mature form of hybrid VLDLR:EGF-A-LDLR ( Fig. 1C , lane 6) and the wild-type LDLR ( Fig. 1D , lane 5). Conversely, wild-type VLDLR ( Fig.  1C , lane 4) and the hybrid LDLR:EGF-A-VLDLR ( Fig. 1D , lane 3) could not be degraded by PCSK9. Thus, EGF-A of the LDLR is suffi cient to confer PCSK9-mediated degradation of a receptor, when it is placed in a cell surface protein that normally does not bind PCSK9 effi ciently.
Recently, it has been reported that PCSK9 promotes the degradation of both VLDLR and apoER ( 38 ). To elucidate the differences between our fi ndings and others, we transiently expressed the VLDLR and the LDLR in Hepa1C1C7 cells and incubated the cells with various doses of wild-type PCSK9 or mutant PCSK9 D374Y for 4 h and overnight at pH 7.4, respectively. The levels of PCSK9 in human plasma range from 0.033 to 2.988 g/ml ( 39 ). Thus, the concentrations of PCSK9 we tested ranged from 0.5 to 4 g/ml. We observed that when incubated with the cells for 4 h or overnight, wild-type PCSK9 could not effi ciently induce VLDLR degradation, even at a concentration of 4 g/ml ( Fig. 2A, B , lane 6), but effi ciently promoted LDLR degradation at a concentration of 0.5 g/ml ( Fig. 2A, B , lane 12). Mutant D374Y that binds to the LDLR with a much higher affi nity could induce VLDLR degradation effi ciently at a concentration of 4 g/ml when incubated with the cells for 4 h ( Fig. 2A , lane 10) or overnight ( Fig. 2B , lane 10). These fi ndings suggest that the VLDLR can be degraded by PCSK9, but with much less effi ciency when compared with the LDLR.
The ligand blotting assay was performed as described with modifi cations ( 17 ). Briefl y, purifi ed LDL-ECD, GST, and GST:EGF-A were labeled with IRDye680 using IRDye680 protein labeling kit. The proteins were directly blotted to nitrocellulose membranes. The membranes were then cut into individual strips and blocked with PBS containing 2.5% nonfat milk. After rinsing briefl y in pH buffer [50 mM Tris maleate (pH 7.4-6.0), 75 mM NaCl, 2 mM CaCl 2 , and 2.5% nonfat milk], the strips were incubated at room temperature for 60 min with 200 ng/ml IRDye800-labeled PCSK9 in the pH buffer, followed by three 15 min washes with pH buffer. The signals were detected by a LI-COR Odyssey infrared imaging system.
Binding of PCSK9 to the LDLR at pH 6.0 The experiments were performed as described in ( 35 ) with modifi cations. COS-7 cells were seeded in 12-well dishes (1.5 × 10 5 cells/well). After 24 h, cells were transiently transfected with expression plasmids containing cDNAs for wild-type or mutant LDLR and pCDNA3.1 vector using XtremeGENE HP. Forty-eight hours later, the cells were washed twice with ice-cold pH 6.0 buffer [50 mM Tris maleate buffer, 150 mM NaCl, 2 mM CaCl 2 , and 2.5% nonfat milk (pH 6.0)] and incubated on ice for 30 min in 0.5 ml pH 6.0 buffer. The cells were then incubated with 0.5 ml pH 6.0 buffer containing PCSK9 (0.5 g/ml) for 1 h at 4°C, washed twice with ice-cold pH 6.0 buffer without milk, collected, and then lysed in 60 l of lysis buffer. The whole cell lysates were subjected to SDS-PAGE (8%) and immunoblotting. The LDLR and PCSK9 were detected as described above.

Biotinylation of LDLR
COS-7 cells were transiently transfected with expression plasmids containing cDNAs for wild-type or mutant LDLR using XtremeGENE HP. After 48 h, cell surface proteins were biotinylated exactly as described ( 31 ). The cells were lysed in 150 l of lysis buffer and then subjected to centrifugation at 15,000 rpm for 5 min. A total of 50 l of the cell lysates was saved and 100 l of the lysates was added to 60 l of 50% slurry of Neutravidin agarose (Pierce). The mixture was rotated overnight at 4°C. After centrifugation at 3,000 g for 5 min, the pellets were washed in lysis buffer three times for 10 min at 4°C. The cell surface proteins were eluted from the beads by adding 1× SDS loading buffer [31 mM Tris·HCl (pH 6.8), 1% SDS, 12.5% glycerol, and 0.0025% bromophenol] and incubated for 5 min at 85°C. Proteins were then analyzed by SDS-PAGE and immunoblotting.

Statistics
All statistical analyses were carried out by GraphPad Prism version 4.0 (GraphPad Software). Student's t -test was used to determine the signifi cant differences between groups. Signifi cance is defi ned as P < 0.05. Results are presented as mean ± SD.

Binding of PCSK9 to LDLR family members
We have reported that PCSK9 effi ciently binds to the LDLR but not to the VLDLR ( 17 ). The residues Asn 295 , Glu 296 , Asp 310 , Tyr 315 , and Leu 318 in EGF-A play an important role in PCSK9 binding. ( 17 ). Asn 295 , Glu 296 , Asp 310 , and Tyr 315 are the calcium binding sequences in EGF-A ( 36 ). Sequence alignment of EGF-A in LDLR family members shows that asparagine is completely conserved in all members we analyzed; glutamate and aspartate are

Effects of mutations in EGF-A on PCSK9 binding
Replacement of Leu 318 with Arg enhanced PCSK9 binding, while mutation L318D reduced PCSK9 binding, suggesting that the charge on the amino acid side chain at position 318 in EGF-A of the LDLR affects PCSK9 binding to the

Effect of mutations of Leu 318 in EGF-A of the LDLR on PCSK9 binding
We have previously reported that replacement of Leu 318 in EGF-A of the LDLR with Asp, as it is in the VLDLR, dramatically reduces PCSK9 binding ( 17 ). Sequence alignment of EGF-A of the LDLR shows that the Leu residue is highly conserved among different species except for rabbit, which has a His residue at that position ( Fig. 3A ). Next, we investigated the specifi c requirement of Leu 318 in EGF-A of the LDLR for effi cient PCSK9 binding. Leu 318 was changed to fi ve other amino acids including a negatively charged residue Asp, as it is in the VLDLR; a Thr residue, as it is in apoER2; a positively charged residue Arg, as it is in LRP1; a neutral amino acid residue Ala, as it is in LRP4 and LRP6 ( Fig. 1A ); a His residue, as it is in the rabbit LDLR ( Fig. 3A ); and a more structurally similar residue Val. Each mutant or wild-type LDLR cDNA was introduced into COS-7 cells. Purifi ed PCSK9 was added to the medium (pH 7.4). The concentrations of PCSK9 we used in the experiments were 0.5 g/ml because the median PCSK9 levels in normal human plasma are 0.487 g/ml ( 39 ). As shown in Fig. 3B , the antibody used to detect the LDLR recognizes both the precursor (p) and the mature The whole lysates were then analyzed by SDS-PAGE (8%) and immunoblotted. VLDLR was detected by a polyclonal anti-HA antibody. m, mature fully glycosylated form of the VLDLR; p, precursor of the VLDLR. LDLR was detected by a monoclonal anti-hLDLR (HL-1). Calnexin was detected by a polyclonal antibody and used as a loading control. Similar results were obtained from at least one more independent experiment . on the amino acid side chains. However, Pro 320 and Arg 329 are mutated to Arg and Pro, respectively, in FH patients. Thus, we also examined the effect of these two FH mutations on PCSK9 binding. PCSK9 (0.5 g/ml) was added to COS-7 cells transiently expressing wild-type or mutant LDLR and incubated with the cells for 2 h at 37°C (pH 7.4). Overexpression of wild-type LDLR signifi cantly enhanced PCSK9 binding ( Fig. 4B , lane 1; Fig. 4C , lane 2). Binding of PCSK9 to the cells expressing mutant LDLRs including D299V ( Fig. 4B , lane 3), R329P ( Fig. 4B , lane 7), G293H ( Fig. 4C , lane 3), and E332G ( Fig. 4C , lane 8) was signifi cantly reduced, while binding of PCSK9 to N309K was increased receptor at the cell surface. Thus, we investigated the potential effects of changes in the charge on amino acid side chains at other positions in EGF-A of the LDLR on PCSK9 binding. We focused on amino acid residues in EGF-A that are different between the LDLR and the VLDLR. As shown in Fig.  4A , 15 amino acid residues are different between the LDLR and the VLDLR, among which 11 residues including Leu 318 change the charge on amino acid side chains. We replaced these amino acid residues in the LDLR with their corresponding residues in the VLDLR ( Fig. 4A ). The corresponding residues of Pro 320 and Arg 329 in the VLDLR are Ala and Lys, respectively, which do not change the charge The experiment was performed as described in Materials and Methods. Briefl y, Hepa1C1C7 cells transiently expressing VLDLR or LDLR were incubated with various concentrations of PCSK9 for 4 h (A) or overnight (B) at 37°C. VLDLR was detected by a polyclonal anti-HA antibody. LDLR was detected by a monoclonal anti-hLDLR (HL-1). m, mature fully glycosylated form of the receptors; p, precursor of the receptors. V, cells were transfected with the empty vector pCDNA3.1. Binding of antibody was detected by IRDye800labeled goat anti-mouse IgG and IRDye680-labeled goat anti-rabbit IgG and imaged on a LI-COR Odyssey system. Calnexin was detected by a polyclonal antibody and used as a loading control. Similar results were obtained from at least two more independent experiments. The bottom fi gures in (A) and (B) are representative ones of protein levels. The top fi gures in (A) and (B) are percentage of densitometry of the receptors. The densitometry of VLDLR and LDLR signals was determined by a LI-COR Odyssey system. Relative densitometry was the ratio of densitometry of VLDLR and LDLR in the presence of a different amount of PCSK9 to that of VLDR and LDLR in the absence of PCSK9 (0). The percentage of densitometry of VLDR and LDLR in the absence of PCSK9 (0) was defi ned as 100%. Values are mean ± SD of three or more independent experiments. the lysosome for degradation ( 17 ). Most recently, it has been shown that the C-terminal domain of PCSK9 interacts with the ligand binding repeats of the LDLR under acidic conditions ( 35,40 ), which may contribute to the stronger binding between PCSK9 and the LDLR in the endosome. Here we investigated if PCSK9 bound to recombinant EGF-A in a pH-dependent manner. EGF-A was purifi ed as GST fusion proteins and shown as a single band ( Fig. 5A ). Same amounts of IRDye680-labeled purifi ed LDLR-ECD that contains the extracellular domain of the LDLR (amino acids 1-699), GST, and GST:EGF-A fusion protein were directly blotted to nitrocellulose membranes ( Fig. 5B ). The membranes were cut into individual strips and then incubated with IRDye800-labeled PCSK9 at different pH values. As shown in Fig. 5C , PCSK9 bound to the LDLR-ECD in a pH-dependent manner. The binding was stronger at pH 6.0 ( Fig. 5C , top), consistent with previous reports ( 16,17 ). There was no detectable PCSK9 binding to GST at all pH values tested ( Fig. 5C , middle). The binding pattern of PCSK9 to GST:EGF-A was similar to that of LDLR-ECD; the strongest binding was at pH 6.0 ( Fig. 5C , bottom). To confi rm these fi ndings, we examined binding of labeled PCSK9 (200 ng/ml) to six different concentrations of GST:EGF-A (10, 25, 50, 100, 250, and 500 ng) at pH 7.4 and pH 6.0. The densitometry of PCSK9 binding was determined. Apparent K m and V max were obtained from the curve fi t of the data to a nonlinear regression (one site binding equation). The apparent V max values of PCSK9 binding to GST:EGF-A at pH 7.4 and at pH 6.0 were comparable (22 and 27 arbitrary units/h for pH 7.4 and pH 6.0, respectively) ( Fig. 5D ). The apparent K m values of PCSK9 binding to GST:EGF-A at pH 7.4 and at pH 6.0 were 747 and 307 ng, respectively ( Fig.  5D ). We also analyzed the binding data with the Scatchard plot (inserts in Fig. 5D ). Similar results were obtained. The apparent K m and V max values of PCSK9 binding to GST:EGF-A were 966 ng and 28 arbitrary units/h for pH 7.4 and 298 ng and 30 arbitrary units/h for pH 6.0, respectively. Thus, PCSK9 has increased affi nity toward the purifi ed recombinant GST:EGF-A at pH 6.0.
We next examined the effect of mutations in EGF-A of the LDLR on PCSK9 binding to the receptor in an acidic environment. The binding assay was performed at 4°C and under pH 6.0 to minimize the internalization of the LDLR and to mimic PCSK9 binding in the endosomal environment. This assay has been widely used to study PCSK9-LDLR binding and LDL-LDLR binding ( 35,41,42 ). COS-7 cells transiently expressing wild-type or mutant LDLR were incubated with ice-cold pH buffer (pH 6.0) containing PCSK9 (0.5 g/ml) for 1 h at 4°C. PCSK9 and the LDLR in the whole cell lysates were then determined by immunoblotting. We found that mutations R329P ( Fig. 6A , lane 7) and E332G ( Fig. 6B , lane 8) led to a signifi cant reduction in PCSK9 binding at pH 6.0. Other mutations including D299V ( Fig. 6A , lane 3), G293H ( Fig. 6B , lane 3), and N309K ( Fig. 6B , lane 4) that affected PCSK9 binding at pH 7.4 ( Fig. 4 ) had no signifi cant effect on PCSK9 binding at pH 6.0. Mutation of Leu 318 to His, Asp, or Arg also had no effect on PCSK9 binding at pH 6.0 ( Fig. 6C ), even though these mutations signifi cantly affected PCSK9 ( Fig. 4C , lane 4) , also involve in PCSK9 binding to the LDLR.

Characterization of binding of PCSK9 to purifi ed EGF-A
We have shown that PCSK9 interacts with EGF-A of the LDLR at the cell surface and binds to the full-length receptor with a much higher affi nity in the acidic environment of the endosome. Consequently, the receptor is redirected to Fig. 3. Binding of PCSK9 to wild-type and mutant LDLR. A: Sequence alignment of EGF-A of LDLR among different species. The sequence alignment was performed by ClustalW2. Leu 318 in the LDLR is in bold. B: Binding of PCSK9 to the LDLR. The experiment was performed as described in the legend to Fig. 1B except that COS-7 cells transiently expressing wild-type (WT) or mutant LDLR were incubated with PCSK9 (0.5 g/ml) for 2 h at 37°C and pH 7.4. LDLR and PCSK9 were detected by HL-1 and 15A6, respectively. Antibody binding was detected using HRP-conjugated goat anti-mouse IgG, followed by ECL. m, mature fully glycosylated form of the LDLR; p, precursor of the LDLR. Calnexin was detected by a polyclonal antibody. Binding of antibody was detected by IRDye800-labeled goat anti-rabbit IgG and imaged on a LI-COR Odyssey system. The bottom fi gure in (B) is representative one of protein levels . V, cells were transfected with the empty vector pCD-NA3.1. Similar results were obtained from at least two more independent experiments. The top fi gure in (B) is percentage of relative densitometry of PCSK9 binding signal. The densitometry of PCSK9 and LDLR signals was determined by Image J Analysis Software. The relative densitometry of PCSK9 binding was the ratio of PCSK9 densitometry to LDLR densitometry. Percentage of the relative densitometry of PCSK9 binding was the percentage of the relative densitometry of PCSK9 binding to mutant LDLR to that of PCSK9 binding to wild-type LDLR. The percentage of the relative densitometry of PCSK9 binding to wild-type LDLR was defi ned as 100%. Values are mean ± SD of three or more independent experiments.

The effect of conserved histidine and lysine residues in YWTD on PCSK9-promoted LDLR degradation
Our data showed that mutations D299V, G293H, N309K, and L318R in EGF-A affected PCSK9 binding at pH 7.4 but not at pH 6.0, suggesting that these residues may not contribute signifi cantly to PCSK9 binding in the acidic endosomal environment. Given that PSCK9 binds to the LDLR with a much higher affi nity in the endosome ( 17 ), other parts of the LDLR may contribute to PCSK9 binding at the low pH environment. YWTD has been shown to interact with the prodomain of PCSK9 ( 30 ). We have previously reported that the LDLR lacking YWTD binds PCSK9, binding at pH 7.4 ( Fig. 3 ). Given that mutations R329P and E332G dramatically reduced PCSK9 binding at both pH values, we examined whether the two mutations affected the cell surface expression of the LDLR via labeling the cells with biotin. The cell surface proteins were then precipitated using NeutrAvidin beads and then immunoblotted for the LDLR. As shown in Fig. 6D , the cell surface levels of wild-type and mutant LDLR (pellets) were proportional to the levels of mature forms of the receptors in the whole cell lysates, indicating that the two mutations had no effect on the traffi cking of the LDLR to the plasma membrane.  (A). B, C: Binding of PCSK9 to wild-type (WT) and mutant LDLR was performed as described in Materials and Methods. COS-7 cells transiently expressing wild-type or mutant LDLR were incubated with PCSK9 (0.5 g/ml) for 2 h at 37°C and pH 7.4. LDLR and PCSK9 were detected by HL-1 and 15A6, respectively. Calnexin was detected by a polyclonal antibody. The binding of antibody was detected by IRDye800-labeled goat anti-mouse IgG and IRDye680-labeled goat anti-rabbit IgG and imaged on a LI-COR Odyssey system. The bottom fi gures in (B) and (C) are representative ones of protein levels. m, mature fully glycosylated form of the LDLR; p, precursor of the LDLR; V, cells were transfected with the empty vector pCDNA3.1. The top fi gures in panels (B) and (C) are percentage of relative densitometry of PCSK9 binding signal. The relative densitometry of PCSK9 binding was the ratio of PCSK9 densitometry to LDLR densitometry. Percentage of the relative densitometry of PCSK9 binding was the percentage of the relative densitometry of PCSK9 binding to mutant LDLR to that of PCSK9 binding to wild-type LDLR. The percentage of the relative densitometry of PCSK9 binding to wild-type LDLR was defi ned as 100%. Values are mean ± SD of three independent experiments. investigated whether Lys 560 and Lys 582 participated into PCSK9's action on the LDLR. Lys 560 and Lys 582 were replaced by Met individually (K560M and K582M) and the mutants were transiently expressed in Hepa1C1C7 cells. As shown in Fig. 7B , these mutations had little detectable effect on PCSK9-promoted degradation of the LDLR. Thus, the three conserved histidine residues and the two lysine residues are not required for PCSK9-promoted degradation of the LDLR.

DISCUSSION
The data reported here provide direct evidence for the critical role of EGF-A of the LDLR in PCSK9-mediated degradation of the receptor. First, PCSK9 only effi ciently but cannot be subjected to PCSK9-induced degradation ( 18 ). The crystallographic structure of the LDLR at the low pH suggests that His 190 , His 562 , and His 586 serve as pH sensors to promote closure of the receptor under acidic conditions ( 33 ). Thus, we changed His 190 to Asp, His 562 to Arg, and His 586 to Lys simultaneously to examine the role of these residues in PCSK9's action on the LDLR. The mutant receptor (H190DH562RH586K) was expressed in Hepa1C1C7 cells at a similar level as the wild-type receptor ( Fig. 7A , lanes 3 and 5). Addition of PCSK9 to the medium resulted in a robust reduction in mature wild-type and mutant LDLR ( Fig. 7A , lanes 4 and 6). In addition, it has been reported that two highly conserved lysine residues, Lys 560 and Lys 582 , present in YWTD play an essential role in the release of ligand from the LDLR ( 42 ). Therefore, we also The membrane was cut into individual strips. The strips were blocked and rinsed briefl y in the pH buffer indicated, and then incubated for 1 h at room temperature with 200 ng/ml PCSK9-IRDye800 in the pH buffer. Following washes, the strips were imaged on a LI-COR Odyssey infrared imaging system. PCSK9 binding is shown in (C). Similar results were obtained from one more independent experiment. D: Binding of PCSK9 to purifi ed GST-EGF-A. The experiment was performed as described in the legend to Fig. 5C except that various amounts of GST-EGF-A (0, 10, 25, 50, 100, 250, and 500 ng) were applied directly on a nitrocellulose membrane. The membrane was cut into individual strips. The strips were blocked and rinsed briefl y in pH 7.4 or 6.0 buffer, and then incubated for 1 h at room temperature with 200 ng/ml PCSK9-IRDye800 in the pH buffer. Following washes, the strips were imaged and the densitometry of PCSK9 signals was determined by a LI-COR Odyssey system. Mean values of densitometry were plotted using GraphPad Prism version 4.0 with nonlinear regression (curve fi t, equation; one site binding) and with the Scatchard plots (inserts) PCSK9 binding at pH 6.0 ( ). PCSK9 binding at pH 7.4 ( ᮀ ). Values are mean ± SD of three independent experiments. demonstrated that highly conserved histidine residues (His 190 in LR5, His 562 and His 586 in YWTD) and lysine residues (Lys 560 and Lys 582 in YWTD), which play an important role in the release of ligand LDL from the receptor, were not required for PCSK9's action on the LDLR ( Fig. 7 ).
Previously, we have reported that the association of PCSK9 with COS-M cells is detectable only in cells expressing the LDLR, but not in cells expressing the VLDLR after addition of PCSK9 at a physiological concentration (0.5 g/ml) ( 17 ). Most recently, Shan et al. ( 43 ) and Poirier et al. ( 38 ) reported that PCSK9 could bind to the VLDLR and the apoER. It is possible that binding of PCSK9 to the VLDLR might be too weak to be detected in our previous experiment when we added a physiological concentration of PCSK9 to COS-M cells. Thus, in the current study, we incubated COS-7 cells expressing various LDLR family members with the medium containing 2 g/ml of bound to the LDLR among the LDLR family members we tested ( Fig. 1B ). Second, EGF-A from the LDLR was sufficient to confer VLDLR degradation after addition of PCSK9 at a normal physiological concentration (0.5 g/ ml) ( Fig. 1C ). Third, wild-type PCSK9 could not induce VLDLR degradation even at a concentration of 4 g/ml with overnight incubation ( Fig. 2 ). In addition, we demonstrated that PCSK9 bound to recombinant EGF-A in a pH-dependent manner with greater binding effi ciency at pH 6.0 ( Fig. 5 ). Replacement of amino acid residues Gly 293 , Asp 299 , and Leu 318 in EGF-A of the LDLR with their corresponding residues in the VLDLR signifi cantly reduced PCSK9 binding at pH 7.4 without effects on PCSK9 binding at pH 6.0 ( Figs. 4, 6 ). On the other hand, substitution of Glu 332 , with its corresponding amino acid residues in the VLDLR (E332G) and FH mutation R329P, reduced PCSK9 binding at pH 7.4 and 6.0 ( Figs. 4, 6 ). Finally, we Fig. 6. Effect of mutations in EGF-A of the LDLR on PCSK9 binding at pH 6.0. A-C: Binding of PCSK9 to wild-type and mutant LDLR was performed as described in Materials and Methods. COS-7 cells transiently expressing wild-type or mutant LDLR were incubated with pH 6.0 buffer containing PCSK9 (0.5 g/ml) for 1 h at 4°C. LDLR and PCSK9 were detected by HL-1 and 15A6, respectively. Calnexin was detected by a polyclonal antibody. The binding of antibody was detected by IRDye800-labeled goat anti-mouse IgG and IRDye680-labeled goat anti-rabbit IgG and imaged on a LI-COR Odyssey system. V, cells were transfected with the empty vector pCDNA3.1. Similar results were obtained in at least two more independent experiments. The bottom fi gures in (A-C) are representative ones of protein levels. m, mature fully glycosylated form of the LDLR; p, precursor of the LDLR. The top fi gures in (A-C) are percentage of relative densitometry of PCSK9 binding signal that was determined as described in the legend to Fig. 3 . Percentage of the relative densitometry of PCSK9 binding was the percentage of the relative densitometry of PCSK9 binding to mutant LDLR to that of PCSK9 binding to wild-type LDLR that was set at 100%. Values are mean ± SD of three independent experiments. D: Biotinylation of the LDLR. COS-7 cells transiently expressing wildtype (WT) or mutant LDLR were biotinylated exactly as described. Biotinylated proteins from the cell surface (pellets) and proteins from the whole cell lysate were analyzed by SDS-PAGE (8%) and immunoblotting. LDLR was detected using HL-1 and calnexin was detected with a polyclonal antibody. V, cells were transfected with the empty vector pCDNA3.1. Similar results were obtained in two independent experiments. LDLR ( 30 ). EGF-A of the VLDLR is identical in humans and mice. Most recently, Surdo et al. ( 30 ) reported that the prodomain of PCSK9 contacts with the YWTD domain of the LDLR via van der Waals interactions. There is 98% amino acid identity in YWTD between human and mouse VLDLR. More studies are needed to determine if these different amino acid residues affect PCSK9 binding to the VLDLR. Taken together, these fi ndings indicate that the physiological role of PCSK9 to either VLDLR or apoER is still uncertain. The plasma levels of PCSK9 in people without statin treatment range from 33 to 2988 ng/ml ( 39 ). It has been shown that plasma levels of PCSK9 are increased in patients treated with atorvastatin ( 46,47 ). Thus, it is possible that PCSK9 may promote VLDLR degradation in individuals with high plasma levels of PCSK9 or gainof-function PCSK9 mutants, especially while under statin treatment.
Previously, we reported that replacement of Leu 318 in the LDLR with Asp, as it is in the VLDLR, signifi cantly reduces PCSK9 binding, and mutation of the corresponding Asp in the VLDLR to Leu increases PCSK9 binding ( 17 ). Here, we observed that replacement of Asn 309 in LDLR with Lys, as it is in the VLDLR, increased PCSK9 binding. However, mutation of the Lys residue in the VLDLR to its corresponding residue in the LDLR, Asn, has no effect on PCSK9 binding ( 17 ). Thus it appears that Leu at position 318 of EGF-A of the LDLR plays a critical role in binding of PCSK9 to the receptor. Indeed, we found that substitution of Leu 318 in EGF-A of the LDLR with other residues including Asp, Thr, and Ala, as they are in VLDLR, apoER2, and LRP4/6, respectively, reduced PCSK9 binding. However, replacement of Leu 318 with Arg, as it is in LRP1, enhanced PCSK9 binding ( Fig. 3B ). Sequence alignment of EGF-A of LDLR family membranes reveals that EGF-A in LRP1 contains amino acid residues that are required for binding of PCSK9 (Asn 295 , Glu 296 , and Asp 310 ; Fig. 1A , bold) ( 17 ). However, no detectable binding was observed in COS-7 cells overexpressing LRP1 ( Fig. 1B ), suggesting PCSK9 for 2 h. However, we still observed PCSK9 binding only in cells expressing the LDLR but not in cells expressing other LDLR family members tested, including the VLDLR and the apoER ( Fig. 1B ). The different results may be simply accounted for by the different protocols used in each study. We incubated COS-7 cells for 2 h with the medium containing purifi ed PSCK9 so that we could control the amount of proteins used. Poirier et al. ( 38 ) incubated CHO-7 cells overnight in the conditional medium isolated from HEK293 cells expressing PCSK9. Shan et al. ( 43 ) used an in vitro assay by mixing purifi ed receptors and purifi ed PCSK9 together. To elucidate these different fi ndings, we incubated the cells expressing the VLDLR with various doses of wild-type and mutant PCSK9 D374Y for 4 h or overnight. We observed that only mutant D374Y that has a much higher affi nity for the LDLR, but not wild-type PCSK9, induced VLDLR degradation at a concentration of 4 g/ml ( Fig. 2 ). Similarly, wild-type PCSK9 (0.5 g/ml) could not induce the degradation of mutant LDLR in which EGF-A was replaced by EGF-A of VLDLR ( Fig. 1D ), but promoted the degradation of mutant VLDLR, in which EGF-A was substituted with EGF-A of LDLR ( Fig. 1C ). These results indicate that PCSK9 may promote VLDLR degradation, but with much less efficiency when compared with the LDLR. The VLDLR and apoER are essential during mouse cerebellar development ( 44 ). Most recently, Roubtsova et al. ( 45 ) reported that PCSK9 induces VLDLR degradation in mouse adipose tissue. Pcsk9 Ϫ / Ϫ mice show higher cell surface expression of VLDLR and accumulate more visceral adipose tissue ( 45 ). However, absence of PCSK9 in humans is not associated with any obvious phenotypes except for hypocholesterolemia ( 14,24 ). The underlying mechanism for the different phenotypes observed in mice and in humans is unclear. The overall sequence homology between mouse and human VLDLR is high, with 97% amino acid identity. The interaction between PCSK9 and LDLR mainly happens between the catalytic domain of PCSK9 and EGF-A of the Fig. 7. The effects of mutations of His residues (A) and Lys residues (B) in the LDLR on PCSK9-promoted LDLR degradation. The experiments were carried out as described in the legend to Fig. 2 except that Hep-a1C1C7 cells transiently expressing wild-type (WT) or mutant LDLR were incubated with PCSK9 (2 g/ml) for 4 h at 37°C (pH 7.4). LDLR was detected by a monoclonal anti-hLDLR (HL-1). Calnexin was detected by a polyclonal antibody. V, cells were transfected with the empty vector pCDNA3.1. Antibody binding was detected using HRP-conjugated goat anti-mouse IgG or donkey anti-rabbit IgG, followed by ECL and exposure to fi lms. m, mature fully glycosylated form of the LDLR; p, precursor of the LDLR. Similar results were obtained from at least two more independent experiments . conditions ( 33 ), and play an important role in the release of bound LDL ( 42 ). YWTD of the LDLR also interacts with the prodomain of PCSK9 ( 30 ), and is required for PCSK9induced LDLR degradation ( 18 ). However, unlike the situation of LDL binding and releasing, substitution of His 190 to Asp, His 562 to Arg, and His 586 to Lys simultaneously or mutation of the two highly conserved lysine residues, Lys 560 and Lys 582 , in YWTD to Met had no effect on PCSK9-promoted LDLR degradation, indicating that these residues have no essential roles in PCSK9's action on the receptor.
FH mutation H306Y increases PCSK9 binding ( 48 ). Here, we found that FH mutation P320R had no effect on PCSK9 binding while R329P reduced PCSK9 binding at pH 7.4 and 6.0. The nuclear magnetic resonance structure of EGF-AB reveals that mutation R329P may disrupt the geometry of the region of the calcium binding site in EGF-A ( 36 ). The integrity of the calcium-binding site in EGF-A is important for PCSK9 binding ( 17 ). Thus, R329P may impair PCSK9 binding through disruption of the calcium-binding site in EGF-A of the LDLR. The crystallographic structures of PCSK9-EGF-AB complex reveal that PCSK9 interacts with the N-terminal EGF-A (28)(29)(30). Asp 299 forms a salt bridge to the N-terminal amine of Ser 153 in PCSK9. We found that mutation D299V reduced PCSK9 binding. Zhang et al. ( 49 ) reported that replacement of Asp 299 with Ser (D299S) in recombinant EGF-A has no signifi cant effect on PCSK9 binding. Thus, it is possible that D299V, but not D299S, disrupts the salt bridge to Ser 153 in PCSK9, leading to a reduction in PCSK9 binding. Asn 309 contributes to PCSK9 binding through forming a hydrogen bond to Thr 377 in PCSK9. Replacement of Asn 309 with Lys (N309K) increases PCSK9 binding at pH 7.4, consistent with the previous fi nding that replacement of Asn 299 with a positively charged residue Arg or Lys in recombinant EGF-A improves binding affi nity for PCSK9 ( 49 ). Mutation N309K introduces a positive charge in the side chain that may stabilize the negative charge on the side chain of Asp 374 in PCSK9; meanwhile mutation N309K retains the hydrogen bond to Thr 377 in PCSK9. Thus, mutation N309K enhances PCSK9 binding. The side chain of Leu 318 reaches out and forms a van der Waals interaction with Cys 378 in PCSK9 ( 28 ). Like mutation N309K, mutation of Leu 318 to Arg also introduces a positive charge in the side chain, which may stabilize the negatively charged side chain of Asp 374 in PCSK9, thereby enhancing PCSK9 binding. The side chain of Asp 374 in PCSK9 is stabilized by His 306 in EGF-A via a salt bridge at low pH ( 29 ). Thus, mutations N309K and L318R did not increase PCSK9 binding at pH 6.0 ( Fig. 6 ). Glu 332 is the last amino acid residue in the C terminus of EGF-A. PCSK9 primarily interacts with the Nterminal EGF-A and does not contact with the C terminus of EGF-A. Gly 293 is the fi rst amino acid residue in EGF-A. Gly 293 also does not contact with PCSK9. Thus, it is unlikely that Gly 293 and Glu 332 contribute signifi cantly to binding of PCSK9 to the LDLR via direct interactions. G293H had no effect on PCSK9 binding at pH 6.0 and E332G had no effect on the traffi cking of the LDLR to the cell surface, suggesting that the two mutations did not result in a major perturbation of the structure of the protein. However, we that there may be some other determinants in EGF-A of the LDLR that contribute to effi cient PCSK9 binding. We did observe that in addition to Leu 318 , replacement of Gly 293 , Asp 299 , and Glu 332 in EGF-A of the LDLR with their corresponding amino acid residues in the VLDLR significantly reduced PCSK9 binding at pH 7.4 ( Fig. 4 ). LRP1 has His, Ser, and Cys at the corresponding positions ( Fig.  1A ), which may cause low-affi nity binding of PCSK9.
PCSK9 binds to the LDLR in a pH-dependent manner with a greater affi nity at low pH. Recently, several studies reported that the ligand binding repeats in the LDLR might interact with the C terminus of PCSK9 at low pH and subsequently contribute a higher affi nity to PCSK9 in the acidic endosomal environment. In the present study, we observed that PCSK9 bound to purifi ed EGF-A more strongly at pH 6.0 than at pH 7.4 ( Fig. 5 ). Our binding experiments indicate that binding of PCSK9 to recombinant GST:EGF-A was increased more than 2-fold at pH 6.0, consistent with previous fi ndings that the pH 6.0 binding environment leads to a 3.8-fold increase in the binding affi nity of the recombinant EGF-AB fragment to PCSK9, when compared with pH 7.4 ( 28 ). Taken together, these fi ndings suggest that EGF-A also contributes to the higher affi nity between PCSK9 and the LDLR at the acidic environment of the endosome. Interestingly, mutations G293H and D299V reduced PCSK9 binding at pH 7.4, but had no detectable effect on PCSK9 binding at pH 6.0, and mutation N309K enhanced PCSK9 binding only at pH 7.4 even though EGF-A bound to PCSK9 more strongly at pH 6.0 ( Figs. 4 and 6 ), suggesting that determinants in EGF-A of the LDLR required for effi cient PCSK9 binding are different at different pH environments. The overall structures of the PCSK9-EGF-AB complex at neutral and low pH are highly similar ( 28,29 ). However, conformation rearrangements happen for EGF-A ( 29 ). For example, His 306 in EGF-A of the LDLR forms an intramolecular hydrogen bond with Ser 305 at neutral pH ( 29 ), but forms an intermolecular salt bridge with Asp 374 in PCSK9 at pH 4.8 ( 28 ). Thus, it is possible that the delivery of the PCSK9-LDLR complex from physiological neutral pH to acidic pH in the endosome leads to conformational changes in EGF-A, which may result in different amino acid residues in EGF-A involved in PCSK9 binding at an acidic pH. Alternatively, these residues may still contribute to PCSK9 binding at pH 6.0, but the other parts of the LDLR, such as the ligand binding repeats, also interact with PCSK9 at the acidic endosomal environment ( 35,40 ), which may compensate for the loss of contributions from these residues in EGF-A at pH 6.0. Yamamoto, Lu, and Ryan ( 40 ) proposed a two-step binding model for interaction between PCSK9 and the LDLR. The catalytic domain of PCSK9 interacts with EGF-A of the LDLR at the cell surface. The conformation of the LDLR is changed when the receptor is exposed to the low pH endosomal environment. The ligand binding repeats of the LDLR then interact with the positively charged C terminus of PCSK9, enhancing PCSK9 binding at the acidic endosomal environment. It has been shown that His 190 in LR5 and His 562 and His 586 in YWTD serve as pH sensors to promote closure of the receptor under acidic cannot exclude a possibility that the two mutations may cause subtle conformational changes in EGF-A, reducing PCSK9 binding indirectly.
In summary, we characterized the role of EGF-A of the LDLR in PCSK9 binding and identifi ed several amino acid residues in EGF-A that contribute to PCSK9 binding. Among them, we found that mutations L318R and N309K increased PCSK9 binding. Biochemistry and crystallography studies reveal that EGF-A directly interacts with PCSK9 ( 17,28,29 ). Purifi ed EGF-AB of the LDLR and synthetic EGF-A peptide can inhibit PCSK9-promoted LDLR degradation in HepG2 cells ( 29,43,48 ). Studies in cultured cells and parabiotic mice demonstrate that PCSK9 can promote LDLR degradation extracellularly (16)(17)(18)(19). Therefore, the EGF-A domain that contains only 40 amino acid residues is a very good target for inhibiting PCSK9-mediated LDLR degradation. However, EGF-A binds to PCSK9 with a relatively low affi nity, which makes it less attractive. Thus, the identifi cation of mutations in EGF-A that can enhance PCSK9 binding will help develop a peptide homologous to EGF-A that can bind to PCSK9 with a high affi nity.