Lipoprotein(a) Catabolism Is Regulated by Proprotein Convertase Subtilisin/Kexin Type 9 through the Low Density Lipoprotein Receptor

Background: Plasma lipoprotein(a) (Lp(a)) levels can be reduced through proprotein convertase subtilisin/kexin type 9 (PCSK9) through an unknown mechanism.

Results: Lp(a) catabolism in hepatoma cells and primary fibroblasts is inhibited by PCSK9 via the low density lipoprotein receptor (LDLR).

Conclusion: LDLR mediates the effects of PCSK9 on Lp(a) internalization.

Significance: Our results provide a mechanistic explanation for the effects of PCSK9 inhibitors on plasma Lp(a) levels.

Abstract

Elevated levels of lipoprotein(a) (Lp(a)) have been identified as an independent risk factor for coronary heart disease. Plasma Lp(a) levels are reduced by monoclonal antibodies targeting proprotein convertase subtilisin/kexin type 9 (PCSK9). However, the mechanism of Lp(a) catabolism in vivo and the role of PCSK9 in this process are unknown. We report that Lp(a) internalization by hepatic HepG2 cells and primary human fibroblasts was effectively reduced by PCSK9. Overexpression of the low density lipoprotein (LDL) receptor (LDLR) in HepG2 cells dramatically increased the internalization of Lp(a). Internalization of Lp(a) was markedly reduced following treatment of HepG2 cells with a function-blocking monoclonal antibody against the LDLR or the use of primary human fibroblasts from an individual with familial hypercholesterolemia; in both cases, Lp(a) internalization was not affected by PCSK9. Optimal Lp(a) internalization in both hepatic and primary human fibroblasts was dependent on the LDL rather than the apolipoprotein(a) component of Lp(a). Lp(a) internalization was also dependent on clathrin-coated pits, and Lp(a) was targeted for lysosomal and not proteasomal degradation. Our data provide strong evidence that the LDLR plays a role in Lp(a) catabolism and that this process can be modulated by PCSK9. These results provide a direct mechanism underlying the therapeutic potential of PCSK9 in effectively lowering Lp(a) levels.

Introduction

Lipoprotein(a) (Lp(a))² has been identified as an independent, causal risk factor for cardiovascular disease, including coronary heart disease (1, 2). Lp(a) is similar to low density lipoprotein (LDL) in lipid core composition and the presence of apolipoprotein B-100 (apoB) but also contains a unique glycoprotein, apolipoprotein(a) (apo(a)), that has strong structural homology with the fibrinolytic zymogen plasminogen (3). Apo(a) contains multiple copies of plasminogen-like kringle IV (KIV) sequences, followed by sequences closely resembling plasminogen kringle V (KV) and an inactive protease domain (3, 4). The KIV domain can be further subdivided into ten types (KIV_1–10) differing from each other in amino acid sequence. In Lp(a) particles, apo(a) is disulfide-linked to the apoB component of the LDL-like moiety through a free cysteine residue in KIV₉ (5); formation of Lp(a) requires initial non-covalent interactions between lysine residues in apoB with weak lysine binding sites present in apo(a) KIV₇ and KIV₈ (6). Additionally, apo(a) contains a strong lysine binding site present in KIV₁₀, which is important for its ability to interact with substrates such as fibrin (7).

Apo(a) can contain from 3 to >40 identically repeated KIV₂ domains, which gives rise to the isoform size heterogeneity reported in the population (8). A general inverse relationship between the size of apo(a) and Lp(a) plasma concentration has been well established, with Lp(a) levels varying widely in the population (8). It has been reported that the inverse correlation between size and concentration is primarily controlled by the level of synthesis rather than catabolism (9, 10). Up to 90% of the variation is genetically determined based on variation in LPA, the gene encoding apo(a), including its size heterogeneity (11).

Many of the details of Lp(a) catabolism remain obscure. Various receptors have been suggested to mediate Lp(a) catabolism by the liver, which include the LDL receptor (LDLR) (12–15), very low density lipoprotein receptor (VLDLR) (16), low density lipoprotein receptor-related protein 1 (LRP1) (17), megalin/gp330 (18), scavenger receptor class B type 1 (SR-B1) (19), and plasminogen receptors (12). The role of the LDLR remains highly controversial, however. Hofmann and co-workers (20) reported that Lp(a) clearance was significantly increased in mice overexpressing LDLR. Additionally, several other studies both in vitro and in vivo have shown that the LDLR is capable of mediating Lp(a) binding and uptake (12–15). A recent cross-sectional analysis of 1,960 patients with familial hypercholesterolemia (FH) revealed that Lp(a) levels were significantly higher in patients with a null LDLR allele compared with control subjects (21), a finding that is in agreement with an earlier report on this topic (22). Conversely, Cain et al. (23) reported that whereas plasma clearance of Lp(a) in mice occurs primarily through the liver and is mediated by apo(a), the catabolism of Lp(a) in Ldlr^−/− mice was similar to that in wild-type mice. Similar results were observed in metabolism studies of Lp(a) in human subjects with FH (24). In addition, plasma Lp(a) concentrations are largely insensitive to statins, which act by increasing the abundance of hepatic LDLR (1).

Recent studies have shown that Lp(a) levels in plasma can be reduced up to 30% using a proprotein convertase subtilisin/kexin type 9 (PCSK9)-inhibitory monoclonal antibody (25–30). In patients treated with a PCSK9 monoclonal antibody, the extent of Lp(a) lowering correlated with the extent of LDL lowering in some studies (27, 28) but not others (30); a more robust effect was consistently observed for LDL levels, which decreased up to ∼70% (27, 28).

PCSK9 is an important regulator of hepatic LDLR number and consists of a pro-domain, followed by a catalytic domain, a hinge region, and a carboxyl-terminal cysteine/histidine-rich domain (31–33). PCSK9 is synthesized as an inactive proenzyme that undergoes intramolecular autocatalytic cleavage in the endoplasmic reticulum (32, 33). The cleaved prosegment remains associated with PCSK9, maintaining PCSK9 in a catalytically inactive form, and the complex is transported to the Golgi apparatus and subsequently secreted. PCSK9 acts as an endogenous regulator of LDLR levels and has been implicated in some cases of FH due to the dominant gain-of-function (GOF) mutations identified in the population (34). GOF mutations lead to increased affinity of PCSK9 for the LDLR, which results in a more rapid degradation of the LDLR and thus higher plasma LDL (34). Conversely, loss-of-function mutations in PCSK9 result in dramatically lowered plasma LDL (34). It is not yet known whether PCSK9 mutations influence Lp(a) concentrations. PCSK9 can target the LDLR for degradation as well as the VLDLR, LRP1, and apolipoprotein E receptor 2 (apoER2; LRP8) (35, 36). However, plasma LDL is predominately cleared through the LDLR (37, 38).

In the current study, using a human hepatocellular carcinoma model system, we sought to understand the mechanistic basis of the ability of PCSK9-inhibitory antibodies to lower plasma Lp(a) concentrations, in the context of the ongoing controversy about the role of the LDLR in Lp(a) catabolism.

EXPERIMENTAL PROCEDURES

Cell Culture

Human embryonic kidney (HEK293) cells were maintained in MEM (Life Technologies) containing 5% fetal bovine serum (FBS; Life Technologies) and 1% antibiotic-antimycotic (Life Technologies). Human hepatocellular carcinoma (HepG2) cells were obtained from the American Type Culture Collection (ATCC) and maintained in MEM supplemented with 10% FBS (ATCC) and 1% antibiotic-antimycotic (Life Technologies). Primary FH fibroblasts were obtained from Coriell Institute (catalogue numbers GM01386, GM01355, and GM00701) and maintained in MEM containing 10% FBS (ATCC). Experiments with FH fibroblasts were performed between passages 5 and 20.

Construction, Expression, and Purification of Recombinant Apo(a)

The construction of expression plasmids encoding the various recombinant apo(a) (r-apo(a)) variants utilized in this study (17K, 17KΔLBS₁₀, and 17KΔLBS_7,8), their transfection into HEK293 cells, and the purification of r-apo(a) from conditioned medium were described previously (6). Briefly, the conditioned medium was subjected to lysine-Sepharose affinity chromatography, and r-apo(a) was eluted using the lysine analogue ϵ-aminocaproic acid (ϵ-ACA). Following concentration and buffer exchange, protein concentrations were determined spectrophotometrically. The purity of r-apo(a) was assessed using SDS-PAGE followed by silver staining.

Construction, Expression, and Purification of Recombinant PCSK9

PCSK9, and PCSK9 D374Y expression plasmids in pIRES2-EGFP (Clontech) were described previously (32, 33). The PCSK9 cDNAs were excised from pIRES2-EGFP using AfeI and AgeI restriction endonucleases and ligated into pcDNA4C (Invitrogen), previously digested with EcoRV and AgeI, such that the expressed protein would contain a carboxyl-terminal His₆ tag. HEK293 cells at 1.8 × 10⁶ cells/well of a 6-well plate were seeded and transfected 24 h later with 2 μg of expression plasmid using MegaTran version 1.0 (Origene) with a 3:1 ratio of reagent to DNA as per the manufacturer's instructions. Stable cells were selected with zeocin (150 μg/ml) 48 h post-transfection. Stable cells were seeded into triple flasks with Opti-MEM (Life Technologies), and conditioned medium was collected every 3 days with the addition of phenylmethylsulfonyl fluoride at a final concentration of 1 mm to the harvest. The harvested medium was adjusted to 50 mm phosphate buffer pH 8.0, 0.5 m NaCl, 1 mm β-mercaptoethanol, 5 mm imidazole, and 10% glycerol, applied to a nickel-Sepharose Excel (GE Healthcare) column, washed, and eluted with 15 mm and 400 mm imidazole, respectively. The eluted pool (4 column volumes) was extensively dialyzed against PCSK9 storage buffer (25 mm HEPES, pH 7.9, 150 mm NaCl, 0.1 mm CaCl₂, and 10% glycerol). The dialyzed samples were then concentrated with PEG 20,000 (Sigma) and dialyzed against storage buffer. Concentrations were determined through a bicinchoninic acid assay (BCA assay; Pierce) using BSA as a standard. The purity of PCSK9 was assessed through SDS-PAGE followed by silver staining and stored in aliquots at −70 °C until use.

Labeling of PCSK9

Purified PCSK9 was dialyzed against 0.1 m Na₂CO₃, pH 8.6, 0.2 m NaCl. PCSK9 was then incubated with a 5-fold molar excess of Alexa Fluor 488 carboxylic acid, succinimidyl ester mixed isomers dissolved in dimethyl sulfoxide at 10 mg/ml (Invitrogen). The reaction mixture was rocked for 4 h at 4 °C to ensure complete labeling. The reaction was quenched with the addition of 0.01 volumes of 1 m Tris, pH 8.0, followed by incubation for 30 min at 4 °C. Free dye was removed through extensive dialysis against 25 mm HEPES, pH 7.5, 300 mm NaCl, 50 mm KH₂PO₄, 0.1 mm CaCl₂, and 10% glycerol. PCSK9 was concentrated using an Amicon Ultra-4 centrifugal filter with a 10 kDa membrane cut-off (Millipore). Concentration was determined spectrophotometrically with a dye/protein (mol/mol) ratio of 2.8.

Transient Transfection

HepG2 cells were transfected with clathrin heavy chain siRNA or scrambled control siRNA (Santa Cruz Biotechnology, Inc.) at a concentration of 80 nm as per the manufacturer's protocol. The transfection mixture was incubated on cells for 8 h, followed by the addition of complete medium. Cells were assayed 48–72 h post-transfection. The percentage knockdown was determined using quantitative RT-PCR (see below). HepG2 cells were transiently transfected with v5 (empty vector), LDLR, or LDLRΔCT (36) using MegaTran version 1.0 (Origene) as per the manufacturer's protocol. Briefly, HepG2 cells were seeded at a density of 2 × 10⁵ cells/6-well plate in antibiotic-free medium and transfected 24 h later with 1.3 μg of cDNA with a 3:1 ratio of reagent to DNA. Cells were assayed 72 h post-transfection.

Quantitative RT-PCR

Determination of clathrin heavy chain knockdown efficiency was determined through the iTaq one-step RT-PCR kit with SYBR® Green (Bio-Rad). The following primers were used: clathrin heavy chain forward, 5′-GGC CCA GAT TCT GCC AAT TCG TTT-3′; clathrin heavy chain reverse, 5′-TGA TGG CGC TGT CTG CTG AAA TTG-3′; GAPDH forward, 5′-GGA GCC AAA AGG GTC ATC ATC-3′; GAPDH reverse, 5′-GTT CAC ACC CAT GAC GAA CAT G-3′.

Internalization Assays

HepG2 cells (in some cases stably transfected with an expression vector for PCSK9) were seeded at 2 × 10⁵ cells/well in a 24-well plate (precoated with 1 mg/ml gelatin), in medium containing 10% lipoprotein-depleted serum (LPDS) for 16 h. Cells were washed twice with Opti-MEM (Gibco) and treated with Lp(a) purified from human plasma (5–10 μg/ml) or r-apo(a) variants (100–200 nm) in the presence of 0, 1, 10, or 20 μg/ml purified recombinant PCSK9 in Opti-MEM for 4 h at 37 °C. For experiments using LDLR-blocking monoclonal antibodies, cells were pretreated for 30 min with 50 μg/ml 5G2 or 7H2 followed by incubation with Lp(a) or apo(a) in the continued presence of 50 μg/ml antibody for 2 h at 37 °C. In some experiments, cells were co-treated with 10 μg/ml lactacystin (Cayman), 150 μg/ml E-64d (Cayman), or vehicle (dimethyl sulfoxide) along with Lp(a) or apo(a) for 4 h at 37 °C. Concanamycin A (Cayman) was dissolved in 100% ethanol and used at a final concentration of 50 nm for 16 h followed by co-treatment with Lp(a) for 4 h. For all internalization experiments, HepG2 cells were extensively washed: three times with PBS, 0.8% BSA; two times with PBS, BSA, 0.2 m ϵ-ACA for 5 min each; two times with 0.2 m acetic acid, pH 2.5, containing 0.5 m NaCl for 10 min each; two times with PBS. The cells were then lysed with lysis buffer (50 mm Tris, pH 8.0, 1% Nonidet P-40, 0.5% sodium deoxycholate, 150 mm NaCl, 1 mm EDTA, 0.1% SDS, 1 mm PMSF, and 150 μg/ml benzamidine).

For experiments with fibroblasts, cells were seeded in a 24-well plate at 1.4 × 10⁵ cells/well in medium containing 10% LPDS for 16 h. Cells were washed twice with Opti-MEM and treated with Lp(a) (5 μg/ml) or apo(a) (100 nm) in the presence or absence of 20 μg/ml PCSK9 in Opti-MEM for 4 h at 37 °C. Cells were extensively washed (three times with PBS, 0.8% BSA; two times with PBS containing 10 μg/ml heparin for 10 min; one time with PBS, BSA, 0.2 m ϵ-ACA for 5 min; two times with 0.2 m acetic acid, pH 2.5, containing 0.5 m NaCl for 10 min; one time with 0.5 m HEPES, pH 7.5, 100 mm NaCl for 10 min; and one time with PBS) and then lysed with lysis buffer. Concentrations of lysate samples were determined by BCA assay with BSA as a standard and analyzed by Western blotting.

Western Blotting

Cell lysates were subjected to SDS-PAGE on 5–15% (Lp(a)-treated cells) or 7–15% (apo(a)-treated cells) polyacrylamide gradient gels, respectively. The gels were transferred onto PVDF membranes (Millipore) and immunoblotted with either mouse anti-human apo(a) a5 antibody (39), mouse-anti human β-actin (Sigma), rabbit anti-human LDLR (GeneTex), or goat-anti-human clathrin heavy chain (Santa Cruz Biotechnology). After incubation with the appropriate horseradish peroxidase-linked secondary antibody, immunoreactive bands were visualized with SuperSignal® West Femto Maximum Sensitivity Substrate (Thermo Scientific) and quantified using Alpha View software (Alpha Innotech).

LDLR Degradation Assay

HepG2 cells were seeded at 2 × 10⁵ cells/well in a 24-well plate in medium containing 10% LPDS for 16 h. PCSK9 (20 μg/ml) with 0, 100, or 250 μg/ml plasma-purified Lp(a) or LDL or 0, 100, or 250 nm apo(a) was added in Opti-MEM, and the cells were incubated for 4 h. Cells were washed three times with PBS and lysed. Concentrations of samples were determined by BCA assay and LDLR levels were analyzed by Western blotting.

Binding Study

Saturation binding curves were generated by incubating LDL or Lp(a), at 0.5 mg/ml, with increasing amounts of PCSK9-Alexa Fluor 488 (25–1200 nm) in binding buffer (25 mm HEPES, pH 7.4, 150 mm NaCl, 2 mm CaCl₂, 1% BSA) for 1 h at 37 °C. Glycerol was added to the samples to a final concentration of 10%, and the samples were subjected to electrophoresis on 0.7% agarose gels (UltraPure Agarose, Invitrogen) for 2 h at 40 V in 90 mm Tris, pH 8.0, 80 mm borate, 2 mm calcium lactate. In-gel scanning and quantification of the amount of labeled PCSK9 free and bound to Lp(a) or LDL was achieved with a FluorChem Q imager (Alpha Innotech). Dissociation constants (K_D) were determined by fitting the data to an equation describing a rectangular hyperbola by nonlinear regression using Graph Pad Prism version 6.

Purification of LDL and Lp(a) and Preparation of Lipoprotein-deficient Serum (LPDS)

Blood was collected from a healthy human volunteer (with written informed consent) with no detectable Lp(a) into BD Vacutainers containing sodium polyanethol sulfonate and acid citrate dextrose. The blood was centrifuged at 2,000 × g for 15 min at 4 °C, and LDL was isolated from plasma through sequential ultracentrifugation (1.02 g/ml < d < 1.063 g/ml); the centrifugation steps were at 45,000 × g for 18 h at 4 °C. The isolated LDL was extensively dialyzed against 150 mm NaCl, 5.6 mm Na₂HPO₄, 1.1 mm KH₂PO₄, 0.01% EDTA (pH 7.4). LPDS was prepared through the addition of NaBr to FBS (ATCC) to a final density of 1.21 g/ml followed by ultracentrifugation as described above. The top fraction was removed, and the infranatant fraction containing LPDS was extensively dialyzed against HEPES-buffered saline (20 mm HEPES, pH 7.4, 150 mm NaCl). Lp(a) was prepared from a single donor with high Lp(a) and a single 16-kringle apo(a) isoform as described previously (40). Concentrations of LDL and Lp(a) were determined by a BCA assay using BSA as a standard.

Immunofluorescence

HepG2 cells were seeded on gelatin-coated coverslips in the wells of 24-well plates at 1.25 × 10⁵ cells/well for 16 h in medium containing 10% LPDS. Cells were washed twice with Opti-MEM (Gibco) and treated with Lp(a) purified from human plasma (5 μg/ml) in the presence of 20 μg/ml purified recombinant PCSK9 in Opti-MEM for 4 h at 37 °C. Cells were washed three times with PBS, 0.8% BSA; two times with PBS, BSA, 0.2 m ϵ-ACA for 5 min each; and three times with PBS. The cells were then fixed with 3.7% paraformaldehyde for 20 min at room temperature. Cells were permeabilized with 0.2% Triton X-100 in PBS for 5 min and blocked with 5% normal goat serum containing 0.1% Triton X-100 (blocking buffer) for 30 min. Mouse anti-human apo(a) (a5) antibody (39) (1:50) was incubated in blocking buffer for 45 min at 37 °C; washed three times for 5 min with PBS, 0.1% BSA; incubated with Alexa Fluor 595-conjugated goat anti-mouse IgG (0.5 μg/ml) in blocking buffer for 30 min at 37 °C; and washed three times with PBS, 0.1% BSA with the final wash containing 4′,6-diamidino-2-phenylindole (DAPI). After this, coverslips were mounted to slides using anti-fade fluorescence mounting medium (Dako). Immunofluorescence microscopy was performed with a Leica DMI6000B inverted fluorescence microscope with a ×63.0 oil immersion objective with a numerical aperture of 1.4 and refractive index of 1.52. The microscope was fitted with a Leica DFC 360FX camera using A4 (DAPI) and Txr (Alexa Fluor 595) filters. Images were acquired using LAS AF software and processed with Corel Draw Graphics Suite X6.

Purification of LDLR-blocking Monoclonal Antibodies

Anti-human LDLR monoclonal antibodies 5G2 and 7H2 (a gift from Dr. Ross Milne, University of Ottawa Heart Institute) were purified from ascites fluid using Protein G-Sepharose 4 Fast Flow affinity chromatography according to the manufacturer's recommendations (GE Healthcare). Concentrations of antibodies were determined using a BCA assay with BSA as a standard.

Statistical Methods

Comparisons between data sets were performed using a two-tailed Student's t test assuming unequal variances.

RESULTS

PCSK9 Inhibits Lp(a) and Apo(a) Internalization

PCSK9 can target the LDLR for degradation in an intracellular pathway by targeting the LDLR from the trans-Golgi network directly to lysosomes (41). Conversely, extracellular PCSK9 targets the LDLR for degradation through binding of PCSK9 to the EGF-A domain of the LDLR and subsequently targeting the complex to lysosomes for degradation (42, 43).Herein, we evaluated the role of both the intra- and extracellular PCSK9-mediated degradation of LDLR in Lp(a)/apo(a) internalization by HepG2 cells. Overexpression of PCSK9, which would stimulate both the intracellular and extracellular pathway of targeting the LDLR for degradation, resulted in a significant decrease in the amount of Lp(a) internalized by HepG2 cells (Fig. 1A). Similar results were obtained for the internalization of a physiologically relevant r-apo(a) species (17K) that contains eight identically repeated KIV₂ domains (Fig. 1B).

FIGURE 1.

PCSK9 reduces the internalization of both Lp(a) and apo(a) in HepG2 cells. A and B, HepG2 cells stably transfected with empty vector or PCSK9-v5 were grown for 16 h in LPDS medium followed by treatment with either 10 μg/ml Lp(a) (A) or 200 nm apo(a) (B) in the presence or absence of 200 mm ϵ-ACA for 4 h. The cells were extensively washed to remove any bound Lp(a)/apo(a) and lysed to determine the relative amount that was internalized, using β-actin as an internal control. Error bars, S.E. with n ≥ 3 independent experiments. *, p < 0.05; **, p < 0.01. C and D, HepG2 cells were grown in LPDS medium for 16 h followed by treatment with various concentrations of PCSK9 or PCSK9 D374Y in the presence of either 10 μg/ml Lp(a) (C) or 200 nm 17K (D). The cells were extensively washed to remove any bound Lp(a)/apo(a) and lysed to determine the relative amount that was internalized, using β-actin as an internal control. The data shown are the means of at least three independent experiments; error bars, S.E. Representative Western blots are shown. *, p < 0.05; **, p < 0.01.

Interaction of apo(a) and Lp(a) with cell surface receptors has been shown to be mediated, at least in part, by the binding of lysine-binding kringles in apo(a) to lysine-containing receptors (12, 44). The addition of a lysine analog, ϵ-ACA, markedly inhibited the uptake of both Lp(a) and r-apo(a) (Fig. 1, A and B), although PCSK9 still significantly reduced the uptake of Lp(a). Likewise, mutating the strong lysine binding site present in KIV₁₀ of 17K (17KΔLBS₁₀ variant) results in a significant decrease in its ability to be internalized (Fig. 1B). Interestingly, however, PCSK9 is able to significantly decrease internalization of either 17K or 17KΔLBS₁₀ in the absence, but not in the presence, of ϵ-ACA (Fig. 1B). Because ϵ-ACA cannot totally abolish the ability of PCSK9 to decrease internalization of Lp(a), we can conclude that there must be a component of the binding and internalization of Lp(a) that is not dependent on the binding to cell surface lysines.

To specifically determine the role of the extracellular PCSK9 degradation pathway, HepG2 cells were exposed to exogenous, purified PCSK9 or a GOF mutant of PCSK9 (D374Y) in the presence of Lp(a) or apo(a). Treatment of HepG2 cells with various concentrations of wild type (WT) PCSK9 resulted in a significant decrease in Lp(a) and 17K internalization (Fig. 1, C and D). The GOF mutant was found to have a more robust effect on Lp(a) (at 1 μg/ml) and 17K (at 1 and 10 μg/ml) internalization compared with WT PCSK9.

PCSK9 Does Not Bind to Lp(a)

It has been previously reported that PCSK9 can bind to LDL in vitro consistent with a one-site binding model with a K_D of ∼325 nm (45). Furthermore, the binding of PCSK9 to LDL inhibits its ability to target the LDLR for degradation in HuH7 human hepatoma cells (45). Hence, we determined whether Lp(a) can bind to PCSK9 in vitro and if Lp(a)/apo(a) can inhibit the ability of exogenous PCSK9 to target the LDLR for degradation. We found that LDL can bind to PCSK9 in vitro with a K_D of ∼130 nm, a value close to that reported previously (45) (Fig. 2A). On the other hand, little or no binding of Lp(a) to PCSK9 was detected (Fig. 2, A and C). Treatment of HepG2 cells with exogenous PCSK9 results in a substantial decrease in LDLR levels, whereas co-treatment of PCSK9 with LDL results in recovery of LDLR levels (Fig. 2D). These findings are also in agreement with previously reported data (42). However, co-treatment of Lp(a) or 17K with PCSK9 results in no significant recovery in LDLR levels (Fig. 2, E and F). Together, these results suggest that Lp(a) does not bind to PCSK9 and therefore cannot block the ability of PCSK9 to target the LDLR for degradation.

FIGURE 2.

PCSK9 binds to LDL in vitro but not to Lp(a). LDLR levels are recovered from PCSK9 degradation by treatment with LDL but not Lp(a) or apo(a). Various concentrations of PCSK9-Alexa Fluor 488 were incubated with 0.5 mg/ml purified LDL or Lp(a) for 1 h at 37 °C. Samples containing PCSK9 and either LDL or Lp(a) were resolved on 0.7% agarose gels. A, bound and free levels of PCSK9 were quantified and fit to a saturation curve by nonlinear regression using GraphPad Prism 6 to give a mean K_D for LDL of 128 ± 40 nm (n = 3). B and C, representative gel images for LDL (B) or Lp(a) (C). D–F, HepG2 cells were grown for 16 h in LPDS medium followed by treatment with various concentrations of LDL (D), Lp(a) (E), or 17K r-apo(a) (F) in the presence or absence of 20 μg/ml PCSK9 for 4 h. The relative LDLR levels were determined and normalized to β-actin. The data shown are the means of at least three independent experiments; error bars, S.E. Representative Western blots are shown. *, p < 0.05; **, p < 0.01 compared with control in the absence or presence of PCSK9.

Lp(a)/Apo(a) Internalization Involves Clathrin-mediated Endocytosis and Internalized Lp(a)/Apo(a) Is Targeted to Lysosomes

PCSK9 has been previously shown to target the LDLR for degradation via clathrin heavy chain-mediated endocytosis and subsequent targeting to lysosomes (46, 47). We therefore determined whether Lp(a) and/or apo(a) can undergo the same degradation pathway. Knockdown of clathrin heavy chain in HepG2 cells results in a significant decrease in Lp(a) and apo(a) internalization (Fig. 3). In both cases, whereas PCSK9 treatment results in a dose-dependent decrease in internalization in the absence of clathrin heavy chain knockdown, no further decrease resulting from PCSK9 is observed in the presence of clathrin heavy chain knockdown (Fig. 3). These results indicate that the PCSK9-regulated internalization of Lp(a)/apo(a) is dependent on clathrin-coated pits.

FIGURE 3.

Lp(a) and apo(a) internalization is dependent on clathrin coated pits in HepG2 cells. A and B, HepG2 cells were transfected with control or clathrin heavy chain siRNA for 48 h followed by incubation of cells in LPDS medium for 16 h. HepG2 cells were then treated with 5 μg/ml Lp(a) (A) or 200 nm 17K (B) in the presence or absence of PCSK9 for 4 h. Internalization of Lp(a)/apo(a) was measured as described in the legend to Fig. 1. The data shown are the means of at least three independent experiments; error bars, S.E. *, p < 0.05; **, p < 0.01. C and D, clathrin heavy chain mRNA was quantified using quantitative RT-PCR (C) as well as total clathrin heavy chain protein levels compared with β-actin (D), as determined through Western blot analysis, following clathrin heavy chain siRNA treatment to determine knockdown efficiency. The data shown are the means of at least three independent experiments; error bars, S.E.

The degradation pathway that Lp(a)/apo(a) undergoes was further evaluated through inhibitors of both the lysosomal and proteosomal pathway. Treatment of HepG2 cells with the lysosomal inhibitor E-64d or concanamycin A resulted in increased intracellular accumulation of Lp(a) and apo(a) (Fig. 4). However, treatment with a proteosomal inhibitor, lactacystin, resulted in no change in intracellular accumulation of Lp(a) or apo(a). These results indicate that Lp(a)/apo(a) is internalized through clathrin-mediated endocytosis and is subsequently targeted for lysosomal degradation.

FIGURE 4.

Lp(a) and apo(a) degradation in HepG2 cells occurs in lysosomes and not proteosomes. A and B, HepG2 cells were grown for 16 h in LPDS medium followed by treatment with either DMSO, E-64d (150 μg/ml), or lactacystin (10 μm) in the presence or absence of 20 μg/ml PCSK9 as well as Lp(a) (A) or 17K (B) for 4 h. Internalization of Lp(a)/apo(a) was measured as described in the legend to Fig. 1. The data shown are the means of at least three independent experiments; error bars, S.E. Representative Western blots are shown. C and D, HepG2 cells were grown for 16 h on gelatinized coverslips in LPDS medium with either vehicle or concanamycin A (50 nm) followed by treatment with Lp(a) in the presence of either vehicle, lactacystin (10 μm), E-64d (150 μg/ml), or concanamycin A (50 nm), in the presence or absence of 20 μg/ml PCSK9, for 4 h. The cells were fixed with PFA, permeabilized with Triton X-100, and stained for apo(a) (red) with a monoclonal anti-apo(a) a5 antibody followed by an Alexa Fluor 595-linked secondary antibody. Nuclei were stained (blue) using DAPI, and cells were visualized by fluorescence microscopy. Representative images are shown. Bar, 25 μm. E, quantification of fluorescence relative to total cell number. The data shown are the means of at least three independent experiments, in which apo(a) internalization was averaged across three different fields of view; error bars, S.E. *, p < 0.05; **, p < 0.01.

PCSK9 Regulates Lp(a) Internalization through the LDLR

Previous studies have shown that apo(a) can be internalized into HepG2 cells through the LDLR or through lysine-dependent interactions with plasminogen receptors (12). We therefore wanted to examine which of these routes might be sensitive to PCSK9, particularly in view of our findings that ϵ-ACA had different effects on the internalization of Lp(a) and apo(a) (Fig. 1). Apo(a) is not itself a ligand for the LDLR, but r-apo(a) present in HepG2 cell medium binds non-covalently (and, ultimately, covalently) to apoB-containing lipoproteins secreted by the HepG2 cells (5, 48), which allows the complex to be internalized by the LDLR in a “piggyback” manner (12, 49). The weak lysine binding sites in KIV type 7 and 8 mediate these non-covalent interactions (6, 49), and therefore, for internalization studies, we utilized a r-apo(a) variant in which both of these lysine binding sites were mutated (17KΔLBS_7,8) (6). We found that 17KΔLBS_7,8 was poorly internalized in HepG2 cells (Fig. 5A); although its internalization did not appear to be affected by PCSK9, this conclusion has to be tempered by the fact that the internalization of this species is at our limit of detection. Nonetheless, it is clear that prevention of the association of apo(a) and apoB100-containing lipoproteins in the medium of HepG2 cells decreases the ability of apo(a) to be internalized by these cells.

FIGURE 5.

The LDLR contributes to Lp(a) and apo(a) internalization in HepG2 cells. A, HepG2 cells were grown in LPDS medium for 16 h followed by treatment with various concentrations of PCSK9 in the presence of 17K or 17KΔLBS_7,8 (200 nm). Internalization of apo(a) was measured as described in the legend to Fig. 1. Error bars, S.E. with n ≥ 3 independent experiments. **, p < 0.01. B and C, HepG2 cells were transfected with expression plasmids encoding either LDLR or LDLRΔCT for 48 h and were then grown for 16 h in LPDS medium. The cells were incubated with 10 μg/ml Lp(a) (B) or 200 nm apo(a) (C) in the presence or absence of 20 μg/ml PCSK9 for 4 h after which internalization of Lp(a)/apo(a) was measured. Error bars, S.E. with n ≥ 3 independent experiments. *, p < 0.05; **, p < 0.01. D and E, HepG2 cells were grown in LPDS medium for 16 h followed by pretreatment with monoclonal LDLR blocking antibodies (mAb 5G2 or 7H2; 50 μg/ml) for 30 min. Cells were then incubated with 5 μg/ml Lp(a) (D) or 100 nm apo(a) (E) in the presence or absence of 10 μg/ml PCSK9 and in the continuing presence of mAbs for 2 h, after which internalization of Lp(a)/apo(a) was measured. The data shown are the means of at least four independent experiments; error bars, S.E. *, p < 0.05; **, p < 0.01 for comparison of absence of antibody and PCSK9.

To determine more directly whether the LDLR plays a role in Lp(a)/apo(a) internalization, the LDLR or the LDLR lacking its carboxyl tail (LDLRΔCT) was overexpressed in HepG2 cells. The ΔCT deletion occurs where the autosomal recessive hypercholesterolemia (ARH) adaptor protein binds and is important for recruiting the complex into clathrin-coated pits (50, 51). Overexpression of LDLR in HepG2 cells results in a dramatic increase in Lp(a) internalization (Fig. 5B) and only a modest and not statistically significant increase in apo(a) internalization (Fig. 5C). Treatment of the cells overexpressing the LDLR or LDLRΔCT with PCSK9 leads to a significant decrease in Lp(a) internalization (Fig. 5B).

Treatment of HepG2 cells with a blocking monoclonal LDLR antibody was also utilized to confirm that the LDLR is involved in Lp(a) catabolism and its regulation by PCSK9. Two LDLR-blocking monoclonal antibodies, 5G2 and 7H2, were used; both were previously shown to specifically block the binding of LDL to the LDLR in cultured human fibroblasts (52). Lp(a) internalization was markedly decreased by the addition of either antibody, and PCSK9 had no effect on Lp(a) internalization in the presence of the antibodies (Fig. 5D). Furthermore, we found that 7H2 likewise markedly decreased 17K internalization (Fig. 5E). On the other hand, PCSK9 did not decrease 17K internalization in the presence of the antibody, and internalization of 17KΔLBS_7,8 appeared to be insensitive to both PCSK9 and the antibody (Fig. 5E). These results indicate that the LDLR mediates internalization of Lp(a) through the LDL component and in a manner that is regulated by PCSK9.

The role for the LDLR was also explored using primary fibroblasts isolated from individuals with or without FH. The three cell lines studied were GM01386 (fully functional LDLR), GM01355 (clinically affected with severe hypercholesterolemia with LDLR activity found to be partially negative), and GM00701 (LDLR activity <1% compared with normal cells). Lp(a) internalization substantially decreases in cells with a defective LDLR, and the internalization was unaffected by PCSK9 in these cells (Fig. 6A). Conversely, no significant difference in 17K internalization is observed between LDLR-defective and normal fibroblasts, and there is no effect of PCSK9 on 17K internalization by any cell line (Fig. 6B). PCSK9 was able to dramatically decrease the LDLR content of those fibroblasts that contained immunoreactive receptor (Fig. 6C). These findings underscore the requirement for apo(a) to couple to apoB-containing lipoproteins in order to internalize through the LDLR in a PCSK9-regulable manner because these fibroblasts do not express apoB-containing lipoproteins.

FIGURE 6.

Lp(a), but not apo(a), internalization decreases in FH fibroblasts lacking LDLR. A and B, FH fibroblasts used were GM01386 (fully functional LDLR), GM01355 (clinically affected with severe hypercholesterolemia with LDLR activity found to be partially negative), and GM00701 (LDLR activity <1% compared with normal cells). The FH fibroblasts were grown in LPDS medium for 16 h followed by incubation with 5 μg/ml Lp(a) (A) or 100 nm apo(a) (B) in the presence or absence of 10 μg/ml PCSK9 for 4 h. Internalization of Lp(a)/apo(a) was measured as described in the legend to Fig. 1. The data shown are the means of at least five independent experiments; error bars, S.E. Representative Western blots are shown. *, p < 0.05; **, p < 0.01 for comparison of control in the absence of PCSK9 and ϵ-ACA. C, the LDLR content of the respective fibroblast cell lines in the presence or absence of PCSK9 treatment was assessed by Western blot analysis using a monoclonal antibody against LDLR.

DISCUSSION

Elevated plasma Lp(a) levels have been recently shown to be effectively reduced with the use of two different monoclonal antibodies against PCSK9 (25–30). This therapy was conceived to lower LDL levels because inhibition of PCSK9 leads to up-regulation of the LDLR. The ability of PCSK9-based therapies to lower plasma Lp(a) challenges the existing dogma that the LDLR does not play a major, if any, role in Lp(a) catabolism. Indeed, we propose, based on our findings, that PCSK9 inhibition leads to a combination of supraphysiological hepatic LDLR abundance and dramatic lowering of LDL that unmasks LDLR as a significant route of clearance of Lp(a) (Fig. 7).

FIGURE 7.

Model for receptor-mediated catabolism of apo(a) and Lp(a). Apo(a) and apoB-containing lipoproteins are independently secreted from hepatocytes and form a non-covalent, and subsequently covalent, complex as Lp(a). Apo(a) can be internalized by plasminogen receptors, and the apo(a) component of Lp(a) mediates clearance of the particle by plasminogen receptors. Apo(a) can only be internalized by the LDL receptor when in a complex with LDL. ϵ-ACA can inhibit binding of apo(a) or Lp(a) to plasminogen receptors as well as Lp(a) assembly, the latter through inhibition of non-covalent interactions between apo(a) and apoB-100. However, ϵ-ACA cannot prevent binding of pre-formed Lp(a) to the LDL receptor. In the presence of inhibitors of PCSK9, there is a substantial increase in DL receptor number such that clearance of Lp(a) through this route is of a much greater magnitude.

We found that PCSK9 is indeed able to inhibit Lp(a) internalization in HepG2 cells (Fig. 1). This effect was observed whether PCSK9 was ectopically overexpressed (and hence active both intracellularly and extracellularly) or added as a purified protein to the culture medium along with Lp(a) (hence acting exclusively extracellularly). Notably, we also found that PCSK9 can stimulate internalization of apo(a) itself (Fig. 1). However, we conclude that the effect of PCSK9 on apo(a) internalization is dependent on the ability of free apo(a) to associate with apoB-containing lipoprotein particles in the culture medium, with internalization of the resultant complex being sensitive to PCSK9 (Fig. 7). This conclusion is based on the fact that internalization of apo(a) by fibroblasts, which do not express apoB, is insensitive to PCSK9 (Fig. 6). Moreover, internalization of the 17KΔLBS_7,8 variant, which cannot associate non-covalently with apoB-containing lipoproteins (6), appears to be insensitive to the effects of PCSK9 (Fig. 5). Given these findings and a previous report that demonstrated that apoB-100, not apo(a), is the ligand in Lp(a) for the LDLR (12), we suspected that the LDLR, the major target of PCSK9, was mediating the PCSK9-sensitive component of Lp(a) and apo(a) internalization.

Importantly, apo(a), due to its structural similarities to plasminogen, may also potentially bind to and be internalized by plasminogen receptors, which contain carboxyl-terminal lysine residues (12). Previous results have shown that removal of the strong lysine binding site in r-apo(a) (the 17KΔLBS₁₀ variant) results in an inability to effectively bind to fibrin surfaces (7). In the current study, 17KΔLBS₁₀ internalization is significantly reduced, but not abolished, compared with wild-type 17K in HepG2 cells (Fig. 1). Treatment of HepG2 cells with the lysine analogue ϵ-ACA resulted in a significant decrease in both Lp(a) and apo(a) internalization, and PCSK9 was still able to inhibit Lp(a) (but not apo(a)) internalization in the presence of ϵ-ACA. Thus, removal of the strong lysine binding site in apo(a) affects its ability to be internalized through lysine-dependent plasminogen receptors but not through non-covalent interactions with apoB and subsequent binding to LDLR. However, treatment with ϵ-ACA abolishes the ability of both apo(a) and Lp(a) to bind to lysine-dependent plasminogen receptors as well as the ability of apo(a) to couple to the apoB component of LDL (53) (Fig. 7); the latter effect accounts for the inability of PCSK9 to inhibit apo(a) uptake in the presence of ϵ-ACA (Fig. 1B).

Recently, it has been reported that LDL can bind to PCSK9 and inhibit its ability to target the LDLR for degradation (45). This effect may serve to limit the extent to which PCSK9 can act to lower hepatic LDLR abundance. We therefore analyzed whether Lp(a) can bind to PCSK9 in order to determine if (i) less Lp(a) is being internalized due to its ability to bind to PCSK9 and thus prevent its internalization or (ii) Lp(a) binding to PCSK9 leads to a reduced ability for PCSK9 to target the LDLR for degradation or other receptors, limiting the ability of Lp(a) to be internalized through those receptors. Through in vitro binding experiments, we have shown that PCSK9 cannot bind to Lp(a) (Fig. 2). In addition, neither Lp(a) nor apo(a) inhibits the ability of PCSK9 to target the LDLR for degradation in HepG2 cells. Therefore, it is possible that the apo(a) component of Lp(a) is interfering with the interaction of PCSK9 and apoB. Notably, the exact site at which the apoB component of LDL binds to PCSK9 is currently unknown.

We also explored the degradation pathway of Lpa(a)/apo(a) through PCSK9. Previous work has shown that PCSK9 can target the LDLR for degradation through clathrin-mediated endocytosis and subsequent lysosomal degradation (42, 43). We show here, through knockdown of clathrin heavy chain, that Lp(a) and apo(a) are also internalized through clathrin-mediated endocytosis. Treatment with PCSK9 results in no further decrease in Lp(a)/apo(a) internalization following clathrin heavy chain knockdown. This indicates that the receptors that internalize Lp(a)/apo(a), which can be regulated by PCSK9, are dependent on clathrin-mediated endocytosis. Furthermore, treatment with the lysosomal inhibitor E-64d or concanamycin A, but not a proteosomal inhibitor, lactacystin, results in a significant accumulation of intracellular Lp(a) and apo(a) with or without PCSK9 treatment. E-64d and lactacystin are highly selective, potent, and irreversible inhibitors of their respective target proteases (54, 55). E-64d inhibits calpain and the cysteine proteases cathepsins F, K, B, H, and L and acts by forming a thioether bond with the active site cysteine of target proteases without affecting cysteine residues in other enzymes. Lactacystin covalently modifies the amino-terminal threonine of specific catalytic subunits of the proteasome. Conversely, concanamycin A specifically blocks lysosomal acidification through inhibition of V-type ATPase (56). Taken together, these results indicate that Lp(a)/apo(a) internalization (whether regulated by PCSK9 or not) occurs, in part, through clathrin-mediated endocytosis and Lp(a)/apo(a) is subsequently targeted for lysosomal degradation. These findings are consistent with a role for LDLR in PCSK9-regulated Lp(a) catabolism.

The contribution of the LDLR to Lp(a) catabolism has been controversial. Compared with LDL, plasma Lp(a) concentrations are much less responsive to conventional lipid-lowering therapies, such as statins. Indeed, some studies have shown an increase, no effect, or a decrease in plasma Lp(a) levels with statins (57). More recent studies have found that statins modestly, but significantly, reduce Lp(a) in subjects with dyslipidemia or heterozygous FH (58, 59). Moreover, in some studies of FH kindreds with a null LDLR, elevated plasma Lp(a) levels are observed in affected individuals (21, 22, 60), although this result has not been unanimously observed (22, 61, 62). Although overexpression of the LDLR in mice significantly increased Lp(a) clearance (20), plasma clearance studies in Ldlr^−/− mice and human FH patients reported no significant difference in Lp(a) clearance compared with the presence of normal LDLR, although a non-significant decrease in fractional catabolic rate in the absence of the LDLR of about 10% was reported in both studies (23, 24). Plausible evidence therefore exists to indicate that the LDLR does participate in Lp(a) catabolism, which may account for the ability of PCSK9 inhibitors to lower plasma Lp(a). Accordingly, we directly examined the role of the LDLR in the regulation of Lp(a) catabolism by PCSK9.

The following lines of evidence from our study very strongly support the concept of the LDLR being a PCSK9-regulable clearance receptor for Lp(a). (i) The GOF PCSK9 mutant D374Y, which can target the LDLR for degradation more rapidly, was more effective than WT PCSK9 in inhibiting both Lp(a) and apo(a) internalization in HepG2 cells (Fig. 1, C and D). (ii) Overexpression of LDLR (and LDLRΔCT) dramatically increases Lp(a) clearance (Fig. 5B). The addition of PCSK9 in the context of LDLR overexpression decreased internalization, but the difference did not reach statistical significance. It is possible that the dose of PCSK9 added was not sufficient to influence the very high concentrations of ectopically expressed LDLR. (iii) The addition of blocking monoclonal antibodies against LDLR decreased Lp(a) internalization, and PCSK9 had no effect in the setting of LDLR blockade with the antibodies (Fig. 5D). (iv) Human fibroblasts lacking the LDLR showed decreased internalization of Lp(a) that was unaffected by the addition of PCSK9 (Fig. 6A).

It is notable that the LDLR lacking the cytoplasmic tail, which interacts with the ARH adaptor protein to promote endocytosis, retains a considerable fraction of the wild-type receptor's ability to internalize Lp(a) (Fig. 5B). It has been previously shown that PCSK9 cannot target the LDLR for degradation in primary hepatocytes isolated from Arh^−/− mice (63). However, PCSK9 can target the LDLR for degradation upon removal of the cytoplasmic tail in CHO cells (64), and the ARH adaptor protein is not necessary in PCSK9-mediated LDLR degradation in fibroblasts (47). These results suggest a potential PCSK9-interacting partner in mediating endocytosis of the LDLR-PCSK9 complex in HepG2 cells.

Less of an increase is observed with apo(a) internalization following LDLR overexpression (Fig. 5C), indicating the requirement for apo(a) coupling to apoB for recognition by this receptor. Although HepG2 cells were deprived of LDL by growth in LPDS, these cells do express apoB and secrete apoB-containing lipoprotein particles in the LDL density range (12). Formation of non-covalent complexes between these particles and apo(a) could be a rate-limiting process and therefore may account for why less of an increase in internalization is observed for apo(a) compared with Lp(a) with LDLR overexpression. Our results also show that Lp(a) internalization is significantly reduced in human FH fibroblasts with a defective LDLR compared with fibroblasts with WT LDLR function (Fig. 6A). Fibroblasts do not express apoB, and therefore the internalization of apo(a) cannot be affected by PCSK9 in either the control or LDLR-defective fibroblast cells (Fig. 6B). Collectively, these results definitively indicate a role for the LDLR in internalization of Lp(a) through the apoB component rather than apo(a).

PCSK9 has been reported to down-regulate other members of the LDLR, specifically the VLDLR and the apoER2 receptor (35). It is not known if these are ligands for Lp(a) in vivo, but it does not appear that they are playing a role in Lp(a) internalization in our experiments, at least with respect to the PCSK9-dependent component. This conclusion stems from our observations that Lp(a) internalization is insensitive to PCSK9 in the presence of antibodies that block the LDLR (Fig. 5D) or in fibroblasts lacking functional LDLR (Fig. 6A).

Previously reported clinical studies have shown that antibodies that target PCSK9 lower Lp(a) to a lesser extent (∼30% decrease) than LDL (∼70% decrease) (25–30). Because LDL concentrations are higher than Lp(a) on a particle number basis, LDL can compete with Lp(a) for binding to the LDLR. It is notable that all study subjects receiving PCSK9-inhibitory antibodies were also receiving an optimal dose of statin (27), and an Lp(a) lowering effect was still observed, possibly because statins increase PCSK9 expression (65). Thus, by increasing hepatic LDLR to supraphysiological levels, possibly along with profound lowering of LDL levels, clearance of Lp(a) through the LDLR assumes a greater importance. This is validated by a previous finding where overexpressing the LDLR in mice results in an increase in Lp(a) catabolism (20). Our results clearly suggest that the effects of PCSK9-inhibitory antibodies on Lp(a) levels in vivo are the consequence of greater LDLR-mediated catabolism of Lp(a) (Fig. 7). Therefore, although the LDLR may not be a major route of clearance of Lp(a) under most circumstances, its importance may increase in the setting of supraphysiological levels of the LDLR, such as is the case with the use of inhibitory antibodies against PCSK9. Definitive proof of this concept will ultimately require further studies in an in vivo setting.