Annexin A2 Reduces PCSK9 Protein Levels via a Translational Mechanism and Interacts with the M1 and M2 Domains of PCSK9*

Background: Annexin A2 (AnxA2) is an extracellular endogenous inhibitor of the PCSK9-LDLR protein-protein interaction. Results: AnxA2 mRNA knockdown in Huh7 cells increases PCSK9 protein levels, and its overexpression in HepG2 cells has the opposite effect. Conclusion: AnxA2 inhibits mRNA translation of PCSK9 and interacts with the M1 + M2 domains of PCSK9. Significance: AnxA2 is a negative regulator of PCSK9 protein levels and activity. Annexin A2 (AnxA2) was reported to be an extracellular endogenous inhibitor of proprotein convertase subtilisin kexin type 9 (PCSK9) activity on cell-surface LDL receptor degradation. In this study, we investigated the effect of silencing the expression of AnxA2 and PCSK9 in HepG2 and Huh7 cells to better define the role of AnxA2 in PCSK9 regulation. AnxA2 knockdown in Huh7 cells significantly increased PCSK9 protein levels as opposed to AnxA2 knockdown in HepG2 cells. However, HepG2 cells overexpressing AnxA2 had lower levels of PCSK9 protein. Overall, our data revealed a plausible new role of AnxA2 in the reduction of PCSK9 protein levels via a translational mechanism. Moreover, the C-terminal Cys/His-rich domain of PCSK9 is crucial in the regulation of PCSK9 activity, and we demonstrated by far-Western blot assay that the M1 and M2 domains are necessary for the specific interaction of PCSK9's C-terminal Cys/His-rich domain and AnxA2. Finally, we produced and purified recombinant PCSK9 from humans and mice, which was characterized and used to perform 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate LDL cell-based assays on the stable knockdown HepG2 and Huh7 cells. We also demonstrated for the first time the equipotency of human and mouse PCSK9 R218S on human cells.

Annexin A2 (AnxA2) was reported to be an extracellular endogenous inhibitor of proprotein convertase subtilisin kexin type 9 (PCSK9) activity on cell-surface LDL receptor degradation. In this study, we investigated the effect of silencing the expression of AnxA2 and PCSK9 in HepG2 and Huh7 cells to better define the role of AnxA2 in PCSK9 regulation. AnxA2 knockdown in Huh7 cells significantly increased PCSK9 protein levels as opposed to AnxA2 knockdown in HepG2 cells. However, HepG2 cells overexpressing AnxA2 had lower levels of PCSK9 protein. Overall, our data revealed a plausible new role of AnxA2 in the reduction of PCSK9 protein levels via a translational mechanism. Moreover, the C-terminal Cys/His-rich domain of PCSK9 is crucial in the regulation of PCSK9 activity, and we demonstrated by far-Western blot assay that the M1 and M2 domains are necessary for the specific interaction of PCSK9's C-terminal Cys/His-rich domain and AnxA2. Finally, we produced and purified recombinant PCSK9 from humans and mice, which was characterized and used to perform 1,1dioctadecyl-3,3,3,3-tetramethylindocarbocyanine perchlorate LDL cell-based assays on the stable knockdown HepG2 and Huh7 cells. We also demonstrated for the first time the equipotency of human and mouse PCSK9 R218S on human cells.
Proprotein convertase subtilisin kexin type 9 (PCSK9, gene PCSK9) is a serine protease of the subtilase family (1) and the last member of the proprotein convertase family (eukaryotic subfamily S8B) (2,3). It is mainly expressed in the liver and to a lesser extent in the small intestine and kidney (1,4). PCSK9 plays a unique role in cholesterol homeostasis and has been genetically associated with autosomal dominant hypercholesterolemia, a genetic disorder also caused by mutations in the low density lipoprotein receptor (LDLR) 3 or apolipoprotein B-100 (5). Following the autocatalytic cleavage of the zymogen pro-PCSK9 into PCSK9 within the lumen of the endoplasmic reticulum, the protein exits this compartment as a tight complex with its inhibitory prosegment and is then secreted (1,6,7). Upon binding to the EGF-A domain of the LDLR via the catalytic domain of PCSK9 (8,9), either intracellularly or at the cell surface (10), the PCSK9⅐LDLR complex is routed to endosomes/lysosomes for degradation in acidic compartments (11,12). In mice (4,13) and humans (14,15), the consequence of lower levels of hepatic LDLR is a reduced clearance of LDLcholesterol from the plasma, resulting in hypercholesterolemia. A significant fraction of circulating PCSK9, which mostly originates from liver hepatocytes (4,16), is cleaved therein by the proprotein convertase furin at the RFHR 218 2QA sequence resulting in an inactive shorter form, PCSK9-⌬N 218 , lacking the N-terminal 218 amino acids (17,18). Also, recent studies revealed that the C-terminal Cys/His-rich domain (CHRD) of PCSK9 plays a crucial role in its biological activity (19 -21). The CHRD of PCSK9 is composed of three tandem repeat modules M1, M2, and M3 (7), where the absence of M2 does not affect PCSK9 secretion but seems to prevent the extracellular activity of PCSK9 in enhancing the degradation of cell-surface LDLR (21). Apart from furin, other proteins have been identified to regulate PCSK9 levels such as the transcription factors sterol regulatory element-binding proteins (22) and hepatocyte nuclear factor 1␣ (23), berberine (23,24), and possibly fenofibrates (25). Transcriptional activation of PCSK9 is mainly regulated by nuclear SREBP-2, which binds to sterol regulatory elements and increases PCSK9 transcription (22), as well as hepatocyte nuclear factor 1␣ (23). Those proteins regulate PCSK9 expression level.
Both furin (17) and annexin A2 (AnxA2) (20,26) have been identified to directly inactivate or interact with extracellular PCSK9 to reduce its activity at a post-transcriptional level. AnxA2 is mostly expressed in lungs, aorta, heart, and small intestine. The protein is present as a monomer and as a heterotetrameric complex with p11 (27) that regulates various cellular processes. AnxA2 is localized in the nucleus (28), in the cytosol, and at the cell membrane. This multifunctional protein, which is mostly implicated in vascular fibrinolysis/homeostasis in endothelial cells (27), binds the CHRD of PCSK9 at the cell surface and inhibits its extracellular LDLR-degrading activity. This protein has also been characterized as an mRNA-binding protein and implicated in the translational regulation of itself, prolyl 4-hydroxylase-␣1, c-Myc, and p53 by binding to specific sequences in the 3Ј-untranslated region of their mRNA (29 -32).
To better define the role of AnxA2 in the regulation of PCSK9 function, we investigated the impact of AnxA2 in two commonly used human hepatocellular carcinoma cell lines, Huh7 and HepG2. We produced stable AnxA2 and PCSK9 knockdown (AnxA2-KD and PCSK9-KD) in both cell lines using short hairpin RNA delivered using a lentiviral delivery approach. These modified cell lines allowed the identification of a novel function of intracellular AnxA2 in regulating PCSK9 mRNA translation. We also showed that the M1 and M2 domains are critical for the interaction of PCSK9 and AnxA2. We next used human and mouse PCSK9 in 1,1Ј-dioctadecyl-3,3,3Ј,3Ј-tetramethylindocarbocyanine perchlorate (DiI)-labeled LDL (DiI-LDL) incorporation cell-based assays, using the above cells lines, and hence defined the sensitivity of LDLR in these cells to extracellular PCSK9, revealing marked differences between the roles of AnxA2 on PCSK9 in HepG2 and Huh7 cells.
Lentivirus Production-HEK293 cells were plated in a 100-mm dish at a density of 5 ϫ 10 6 cells in DMEM with 10% FBS. On the 2nd day, HEK293 cells were cotransfected with shRNA constructs in the lentiviral vector pLKO.1/puro (Sigma Mission shRNA) targeting hPCSK9 (TRCN0000075237) or hAnxA2 (TRCN0000056145) and the packaging plasmids pLp1, pLp2, and VSVG (Invitrogen/Sigma) using Lipofectamine 2000 (Invitrogen). Cells were maintained for 48 h, and conditioned media containing lentiviral particles were collected and filtered through a 0.22-m filter. HT-1080 cells were stably transduced with serial dilution of the conditioned media, and 7-14 days prior to puromycin selection, the cell colonies were counted to estimate lentiviral titer.
Knockdown in HepG2 and Huh7 Cells-HepG2 and Huh7 cells were plated, 1 day prior to infection, at 100,000 cells/well in a 6-well plate in complete (10% FBS) Eagle's minimal essential medium and DMEM, respectively. Lentiviral infection was performed as described previously (10) using a multiplicity of infection of three in presence of 4 g/l of Polybrene over a period of 24 h. The media were replaced with fresh complete media without Polybrene followed by the addition of 2 g/ml puromycin for stable selection after 48 h. RNA isolation was performed using the RNeasy kit (Qiagen) and QRT-PCR was carried out with RNA samples (1 g) using SYBR Green and QPCR system Mx3005p (Stratagene). The primers sequences were as follows: hPCSK9 forward, 5Ј-ATCCACGCTTCCT-GCTGC-3Ј, and hPCSK9 reverse, 5Ј-CACGGTCACCTGCT-CCTG-3Ј; hAnxA2 forward, 5Ј-TGAGCGGGATGCTTTG-AAC-3Ј, and hAnxA2 reverse, 5Ј-ATCCTGTCTCTGTGCAT-TGCTG-3Ј; and hLDLR forward, 5Ј-AGCCGTAAGGACAC-AGCACAC-3Ј, and hLDLR reverse, 5Ј-GGAAGACGAGGAG-CACGATGG-3Ј. Results are presented as the means Ϯ S.E. of triplicate experiments for each gene after normalization to nontarget (NT) control.
DiI-LDL Uptake Assay-In a 96-well plate coated with poly-L-lysine (Sigma), HepG2 and Huh7 were plated at 2 ϫ 10 4 cells/ well in complete media. Cells were grown for 24 h before replacing the complete media with serum-free media and incubated for another 24 h. Then, purified recombinant PCSK9 was added to the medium at a final concentration of 0.01 to 20 g/ml for 4 h followed by the addition of 2 mg/ml DiI-LDL (Biomedical Technologies) for 4 h at 37°C. Cells were washed twice with PBS, and DiI-LDL raw fluorescence units were measured with Molecular Geminix XS at excitation/emission of 520/575 nm. Plates were frozen at Ϫ80°C overnight, and cells number/well was determined using Cyquant kit (Invitrogen). Data were analyzed in Prism 6.0 using ratio of DiI-LDL/ Cyquant fluorescence to normalize LDL uptake to cell number. Each value is relative to maximal DiI-LDL incorporation value in the control.
Cycloheximide Time Course-HepG2 and Huh7 cells were plated in a 6-well plate and incubated for 24 h followed by the addition of 40 g/ml cycloheximide (Sigma) with different incubation times from 1 to 8 h. Total cell protein extraction was performed with RIPA, and total cell lysate (30 g) was analyzed by Western blot using rabbit anti-human PCSK9 antibody and mouse anti-human ␤-actin antibody (Cell Signaling).
RNA Immunoprecipitation (RIP) Assay-Immunoprecipitation of endogenous AnxA2 was performed on Huh7 nontarget and knockdown AnxA2 cells and HEK293 cells transiently transfected (48 h) with hPCSK9 ⌬3ЈUTR and hPCSK9 3ЈUTR. Cells were grown at 90% confluence in a 100-mm dish and were washed twice with ice-cold PBS, fixed with 0.75% formaldehyde in PBS, and then washed twice with ice-cold PBS. Cells were harvested and sonicated in 1 ml of RIPA with protease inhibitors. Cell lysates (500 g) were incubated overnight at 4°C with 30 l of protein G magnetic beads (Millipore) coated with 5 g of mouse anti-annexin A2 antibody (BD Biosciences) or without as a control. The beads were collected, and the supernatant was saved for RNA extraction. The beads were washed three times with ice-cold RIPA, and supernatant/beads were incubated 2 h at 60°C with 0.5 mg/ml proteinase K. RNA extraction was performed with TRIzol followed by RNA precipitation with ethanol and glycogen. RNA pellet was incubated with DNase 15 min at room temperature, and RT-PCR was carried out on both supernatant and immunoprecipitated samples using hPCSK9 QPCR primer sequence listed previously.
Purification of Human and Mouse Recombinant PCSK9 from Drosophila Expressions System-Human PCSK9 R218S (hPCSK9-RS), human PCSK9 R218S/D374Y (hPCSK9-RSDY), and mouse PCSK9 R221S (mPCSK9-RS) cDNAs were cloned in the insect expression plasmid pAc5.1/V5-HisA (Invitrogen) (33). Dro-sophila Schneider 2 (S2) cells were cultivated in Schneider's Drosophila medium containing 10% heat-inactivated FBS. S2 cells have been stably cotransfected with the expression plasmid and a selection vector for hygromycin-B using a calciumphosphate transfection kit (Invitrogen). Cells have been selected for 3 weeks with 500 g/ml hygromycin-B. Following the selection, S2 cells have been adapted in free serum media prior to production of ϳ20 liters conditioned media (Wave Bioreactor, GE Healthcare). The following steps of purification were carried out: (i) ultrafiltration (Pellicon 2 maxi/50-kDa cutoff, Millipore) to reduce the volume of media; (ii) ion exchange chromatography; (iii) hydrophobic interaction chromatography; (iv) nickel affinity; and (v) size exclusion chromatography (GE Healthcare columns). Each recombinant protein has C-terminal V5 and His tags. After each purification step, fractions containing recombinant PCSK9 were detected by Western blot and Coomassie Blue staining. For Western blot, rabbit anti-hPCSK9 antibody and mouse anti-V5 antibody (Invitrogen) were used to detect hPCSK9 and mPCSK9 mutants, respectively. Purified proteins were analyzed with Agilent 2100 Bioanalyzer to evaluate protein concentration and purity. Protein purity and concentration were also determined by quantitative amino acid analysis on a Beckman autoanalyzer model 6300 (data not shown).
Mass Spectrometric Analysis of Purified Recombinant Human PCSK9s-Purified recombinant hPCSK9-RS and hPCSK9-RSDY (5 g) were applied to a Gold ProteinChip Array (Ciphergen Biosystems Inc.) with 2 l of saturated 3,5-dimethoxy-4-hydroxycinnamic acid (sinapinic acid, Sigma) in 50% (v/v) acetonitrile (Fisher) ϩ 0.5% (v/v) trifluoroacetic acid (Sigma) and air-dried. Samples were analyzed by time of flight-mass spectrometry (TOF-MS) in a Ciphergen Protein Biology System II. Analyses represent an average of 100 shots, and masses were externally calibrated with all-in-one protein standards (Ciphergen Biosystems Inc.). All data were normalized for total ion current, and peak areas were calculated using the indirect method (with a bracket height of 0.4 and width expansion factor of 2) contained within Ciphergen's ProteinChip 3.1 software.
Enzymatic dephosphorylation was carried out by incubating 10 g of each purified recombinant hPCSK9-RS and hPCSK9-RSDY in the presence of 10 units of shrimp alkaline phosphatase (SAP) (Fermentas) in the provided reaction buffer system for 30 min at 37°C with agitation. Control reactions, in the absence of SAP, identified phosphorylated peaks, whereas those in the absence of recombinant PCSK9 were carried out to identify nonspecific peaks during subsequent MS analyses carried out as described above.
A PNGase F kit (Sigma) was employed to deglycosylate N-linked glycoproteins on recombinant PCSK9 according to the manufacturer's denaturing procedure. 10 g of recombinant PCSK9 was denatured in 1ϫ PNGase buffer ϩ 0.1% (v/v) SDS (Sigma) and 50 mM 2-mercaptoethanol (Sigma) at 100°C for 5 min and allowed to cool. IGEPAL 0.75% (v/v) (Sigma) and 2 l of PNGase F were added to the reaction mix and incubated overnight at 37°C. The control reaction excluded the PNGase F from the reaction. Following deglycosylation, the reaction mixtures were equilibrated three times in 0.1% trifluoroacetic acid using an Amicon Ultra YM60 Microcon (Milli-pore Corp.) by centrifugation at 4,000 ϫ g and analyzed by TOF-MS as described above (34,35).

Translational Effect of Annexin A2 on PCSK9 Protein Levels-
For a better understanding of the role of AnxA2 and PCSK9 in Huh7 and HepG2 cells, we generated stable knockdown of AnxA2 and PCSK9 mRNAs using lentivirus delivery of shRNAs. QPCR analysis showed a reduction of 96 and 97% of AnxA2 mRNA in Huh7 and HepG2 cell lines, respectively (Fig. 1A), and 70 and 75% reductions of PCSK9 mRNA in the same cell lines (Fig. 1B). Western blot analysis of PCSK9 (10,21) in these cells confirmed the efficacy of the mRNA knockdown therein (data not shown). As described previously (20), we observed 5-fold higher levels of AnxA2 mRNA in control (nontarget) Huh7 versus HepG2 cells (Fig. 1A). Western blots performed on total cell lysates and media from stable Huh7 AnxA2-KD cells revealed that the zymogen pro-PCSK9 and mature PCSK9 protein levels were on average ϳ3-fold more elevated in AnxA2-KD cells than in control cells expressing a NT shRNA ( Fig. 2A). The increased PCSK9 protein levels in Huh7 AnxA2-KD cell lysates and media are associated with a 40 -60% lower total LDLR protein level. In agreement, FACS analysis of these cells demonstrated a 40% reduction in cellsurface LDLR in the absence of AnxA2 (data not shown). This is likely due to the increased mature PCSK9 levels in AnxA2-KD cells. Surprisingly, Western blot analysis of PCSK9 in HepG2 AnxA2-KD cells and media did not reveal a significant change in the levels of pro-PCSK9 and its mature form (Fig. 2B). Yet AnxA2-KD HepG2 cells showed an increase of 30 -50% in LDLR levels (Fig. 2B). This contrasts with the observed 40 -60% reduction of total LDLR protein levels in AnxA2-KD Huh7 cells ( Fig. 2A). These results demonstrated that reduction of AnxA2 expression in Huh7 and HepG2 cells has a different consequence on PCSK9 and LDLR protein levels.
We next examined the mechanism underlying the increased levels of PCSK9 protein upon loss of AnxA2 in Huh7 cells. Because QPCR analyses of PCSK9 and LDLR mRNA showed no significant differences in control versus AnxA2-KD cells (Fig.  3A), this ruled out a possible transcriptional effect. Based on endoglycosidase H sensitivity, we have previously shown that most of the intracellular pro-PCSK9 and PCSK9 reside in the endoplasmic reticulum, and upon exiting from this compartment, PCSK9 is rapidly secreted and not stored in cells (6,11). Thus, to probe for a possible post-translational effect of AnxA2, we performed a time course incubation of Huh7 NT and AnxA2-KD cells in the presence of the translational inhibitor cycloheximide from 0 to 6 h. Analysis of the time-dependent decrease in intracellular pro-PCSK9 and PCSK9 protein levels revealed that although in this experiment the absence of AnxA2 resulted in an ϳ2-fold increased protein levels, it did not affect the ϳ1-h half-life of pro-PCSK9 or PCSK9 in these cells (Fig. 3, B and C). Therefore, AnxA2 does not significantly affect the post-translational intracellular degradation rate of PCSK9. These results suggested that in Huh7 cells AnxA2-KD increases PCSK9 protein levels via a translational mechanism.
In that context, it has previously been shown that AnxA2 is also implicated in the mRNA translational regulation of different proteins such as itself, prolyl 4-hydroxylase-␣1, c-Myc, and p53 by binding to specific sequences in the 3Ј-untranslated region of their mRNA (29 -32). Thus, if in Huh7 cells AnxA2 down-regulates the translation of PCSK9 mRNA, this would explain the observed decreased PCSK9 protein levels in AnxA2-KD cells. However, AnxA2-KD in HepG2 cells did not result in a similar reduction of PCSK9 protein levels as observed in Huh7 AnxA2-KD cells (Fig. 2B). This could be due to the 5-fold lower expression levels of AnxA2 in HepG2 versus Huh7 cells (Fig. 1A), making them less sensitive to the loss of AnxA2. Accordingly, when we tested overexpression of AnxA2 in HepG2 cells, we observed a 25-30% reduction in PCSK9 levels, with no effect on PCSK9 and LDLR mRNA levels ( Fig. 3D) or on the ϳ2-h half-life of intracellular PCSK9 (Fig. 3, E and F).
Previous studies (29 -32) have proposed a consensus RNAbinding sequence 5Ј-AA(C/G)(A/U)G-3Ј recognized by AnxA2 that was present at five different positions in the 3ЈUTR of PCSK9 mRNA (Fig. 4A). Therefore, we examined whether AnxA2 could bind endogenous PCSK9 mRNA to regulate its translation by an RNA immunoprecipitation assay using AnxA2 antibody to pull down the AnxA2⅐PCSK9 mRNA complex in Huh7 nontarget and Huh7 AnxA2-KD cells. Surprisingly, PCSK9 mRNA was detected in Huh7 nontarget cells in both immunoprecipitates and supernatants, whereas PCSK9 mRNA was only observed in the supernatant upon loss of AnxA2 in Huh7 cells (Fig. 4B). To verify our premise that AnxA2 could bind one of the five possible sites in the 3ЈUTR of PCSK9 (Fig. 4A), we transiently transfected HEK293 with two constructs (Fig. 4C), containing the full-length 1,247-nucleotide-long 3ЈUTR (hPCSK9 3ЈUTR) or a form that lacked the 3ЈUTR (hPCSK9 ⌬3ЈUTR), and we performed the RIP assay to pull down the AnxA2⅐PCSK9 mRNA complex (Fig. 4D). In HEK293 cells expressing hPCSK9 ⌬3ЈUTR, PCSK9 mRNA was only found in the supernatant as it failed to be coimmunoprecipitated in the negative control (mock IP) or with AnxA2 antibody (IP AnxA2). In contrast, in HEK293 cells expressing hPCSK9 3ЈUTR, PCSK9 mRNA coimmunoprecipitated with AnxA2 but was not immunoprecipitated in a negative control mock IP (Fig. 4D). These preliminary results uncovered an interaction between AnxA2 and endogenous PCSK9 mRNA, and such binding seems to require the 3ЈUTR of PCSK9. Future studies will determine which motifs within the 3ЈUTR are implicated in this interaction and whether such binding results in a direct translation inhibition of the PCSK9 mRNA.
AnxA2 Binds the M1 ϩ M2 Domain of the CHRD-It has been reported that the CHRD of PCSK9 is needed for the intracellular and extracellular LDLR degradation induced by PCSK9 (19 -21, 36) and that its M2 module is critical only for the extracellular activity of PCSK9 on LDLR (21). The regulatory mechanism behind the role of the M2 domain remains unclear. Because the N-terminal repeat domain 1 of AnxA2 binds the CHRD (20, 26), we thought to define the CHRD module(s) implicated in this interaction. As reported previously, we used a far-Western blot assay testing the interaction of a secreted form of PCSK9 with a nitrocellulose-bound form of AnxA2 or AnxA1 as a negative control (20,26). Accordingly, the media of HEK293 cells overexpressing V5-tagged PCSK9, a form lacking the CHRD (L455X) or the CHRD alone (Fig. 5A), were used to probe the interaction of these proteins (Fig. 5B) with either HA-tagged AnxA2 or AnxA1. The latter were produced in HEK293 cells, and the lysates were separated by SDS-PAGE and then blotted onto a nitrocellulose membrane (Fig. 5C). As reported previously (20), incubation of HepG2 cells with PCSK9 or its CHRD obtained from the conditioned medium of HEK293 cells resulted in an efficient binding of PCSK9 or the CHRD to AnxA2 but not to AnxA1 (Fig. 5D). In contrast, PCSK9 L455X lacking the CHRD does not bind either protein (Fig. 5D).
Next, we tested whether single point natural mutations within the C terminus of PCSK9 affect its ability to interact with AnxA2. These included GOF (E482Q, E482R, R469W, R496W, and H553R) and loss-of-function (R434W and Q554E) mutants of PCSK9 within the CHRD or its adjoining N-terminal hinge domain (2,37). None of these natural mutants affected the ability of PCSK9 to bind AnxA2, which had similar results to mutants E482Q and E482R presented in Fig. 5E. To define which of the three modules of the CHRD is critical in this interaction, we used previously generated constructs lacking one or two of these domains (Fig. 6, A-C) (21). Far-Western blot anal-ysis revealed that the most critical domains were M1 and M2, as the constructs lacking one or both domains exhibit no binding, whereas the lack of M3 does not influence binding of PCSK9 to AnxA2 (Fig. 6D). The sum of our results demonstrate the direct interaction of AnxA2 to PCSK9 C-terminal domains M1 and/or M2 likely responsible for the extracellular inhibition of PCSK9 activity by secreted AnxA2, whereas intracellular AnxA2 seems to be responsible for the translational repression of PCSK9.

JOURNAL OF BIOLOGICAL CHEMISTRY 17737
Expression and Purification of Recombinant Human and Mouse PCSK9-Previous studies have shown the effect of exogenous PCSK9 on LDLR degradation on cells by analyzing the following: (i) total cell lysates by Western blot (10,22,38); (ii) measuring the LDL uptake using fluorescent DiI-LDL (39) or (iii) by immunofluorescence (11,40), and (iv) surface LDLR levels by FACS (41,42). However, measurement of LDL internalization by purified PCSK9 has never been used on stable knockdown cell lines lacking either PCSK9 or AnxA2.
To analyze the LDL incorporation profile in our different cell lines, we expressed and purified recombinant hPCSK9 in S2 cells. We previously showed that in vivo and/or ex vivo human and mouse PCSK9 are inactivated by cleavage within their catalytic domain at RFHR 218 2 (for hPCSK9) and RFHR 221 2 (for mPCSK9), mostly by the proprotein convertase furin (17, 18).
Thus, a naturally occurring mutant R218S or R221S (43), which is resistant to furin cleavage (17,18), was selected for our study, as preliminary data showed extensive cleavage (ϳ40% of total hPCSK9 level) of wild type PCSK9 by an endogenous furin-like enzyme in S2 cells (Fig. 7A). As described under "Experimental Procedures," S2 cells were transfected with a plasmid vector, pAc5.1/V5-HisA, for high level constitutive expression of hPCSK9-RS. Starting from 20 liters of S2 crude media containing the secreted hPCSK9-RS, the protein was purified to homogeneity following five purification steps (see "Experimental Procedures"). The purified protein was analyzed by SDS-PAGE (Fig. 7B), as well as by Western blot and bioanalyzer analyses (data not shown). Following SDS-PAGE of the purified material and analysis by either Western blot or Coomassie Brilliant Blue staining, we observed the presence of two ϳ65and ϳ14-kDa

IP AnxA2
Sup. IP AnxA2  proteins, corresponding to the mature/active form of PCSK9 and its tightly bound prosegment, respectively. The apparent molecular masses are similar to those calculated from the amino acid composition of PCSK9 (amino acids 153-692 ϩ 6 His), which are predicted to be 62 and 13.8 kDa (amino acids . Quantitative analysis of the purified material with a bioanalyzer showed an overall 96% purity. The success of this initial approach led us to use similar steps to express and purify the GOF hPCSK9-RSDY and mPCSK9-RS. The D374Y mutant has been reported to have at least 10-fold more affinity for LDLR than the WT (7), a property corroborated in a cell-based assay with the RSDY mutant (see below). Here also, at each step of the purification, we were able to eliminate various contaminating proteins (data not shown) until we reached a purity of Ͼ98% after the size exclusion chromatography step for hPCSK9-RSDY and mPCSK9-RS (Fig. 7B). The production yield for every recombinant PCSK9 was ϳ0.5 mg/liter of S2 crude media. Characterization of Recombinant hPCSK9s-Edman degradation of purified hPCSK9-RS and -RSDY and their amino acid composition confirmed their purity. Indeed, the N-terminal sequence determined (data not shown) starts at the expected Ser 153 of PCSK9, just following the autocatalytic cleavage site of its zymogen at VFAQ 152 2SIP (Fig. 8A), as reported in human cells (6). Thereafter, hPCSK9-RS and -RSDY were analyzed by mass spectrometry to identify their post-translational modifications (PTMs). In human cells, PCSK9 undergoes three major PTMs (Fig. 8A) (34); and (iii) N-glycosylation of Asn 533 (1).

Huh7 non-Target
To identify phosphorylation, both human recombinant proteins were incubated with and without SAP, and TOF-MS was performed on each sample. In Fig. 8B (upper panel), the observed molecular mass revealed that the prosegment of hPCSK9-RS secreted from S2 cells is either singly sulfated (39%; SO 4 2Ϫ /13,829.4 Da) and or doubly sulfated ϩ phosphorylated (43%; SO 4 2Ϫ ϩ PO 4 2Ϫ /13,909.5 Da). Interestingly, the double mutant hPCSK9-RSDY is predominantly singly sulfated (58%; Fig. 8C) with lower sulfation ϩ phosphorylation of the prosegment (38%). After SAP incubation, we observed a major peak  (E482Q and E482R). C, cell lysates of HEK293 expressing AnxA1-HA or AnxA2-HA. D and E, direct binding of PCSK9 and AnxA2 was analyzed by far-Western blot assay. Thirty micrograms of total cell lysates from HEK293 cells expressing AnxA1-HA or AnxA2-HA were resolved by SDS-PAGE, transferred onto a PVDF membrane (bait proteins), and incubated 4 h with concentrated (10 -15-fold) conditioned media from HEK293 cells containing 2 g of PCSK9 or its derivatives (prey proteins). Specific binding of prey proteins on AnxA2 (bait) was detected with HRP-conjugated anti-V5 mAb. Cell lysates ␤-actin levels are shown. n.s. for nonspecific band. JUNE (Fig. 8B, lower panel and Table 1). To identify the N-glycosylation at Asn 533 , overnight incubation was performed on hPCSK9 with and without PNGase F. Before treatment (Fig. 9A, upper panel), two single peaks are observed that correspond to hPCSK9-RS mature form (59,841.9 Da) and the mature form with the prosegment that remained as a complex even after electron bombardment in the MS (73,694.5 Da). After treatment with PNGase F (Fig. 9A,  lower panel), there is a difference of 1,000 -1,100 Da from all the peaks observed for hPCSK9-RS protein compared with MS analyses without treatment ( Table 2). This mass loss likely corresponds to the removal of the N-glycosylation modification on PCSK9 by PNGase F. Most notably, loss of N-glycosylation considerably reduced the level of the complex prosegment⅐PCSK9 that flew in the MS, suggesting that N-glycosylation stabilizes this heterodimer. The same effect was observed in the MS analyses of hPCSK9-RSDY, and the molecular masses are listed in Table 2. However, removal of N-glycosylation by incubating the hPCSK9-RS mutant with PNGase F did not affect its biological activity, as it was still able to reduce DiI-LDL uptake levels with identical potency (Fig. 9B). We conclude that although the stability of the complex prosegment⅐PCSK9 seems to be reduced in the absence of N-glycosylation, it still remains as an active complex in solution. Thus, mass spectral analyses confirmed that both human recombinant proteins were secreted from S2 cells and post-translationally modified like the ones characterized in human cells (6,17).

Regulation of PCSK9 by Annexin A2
Cell-based Assay with Purified PCSK9-The DiI-LDL cellbased assay is a powerful tool to compare the activity of different PCSK9 isoforms, as demonstrated previously (39,41,44). Accordingly, DiI-LDL uptake was used to probe the ability of the above purified PCSK9s to enhance the degradation of the LDLR. As shown in Fig. 10, when exogenous PCSK9 is added to the media, DiI-LDL internalization is reduced in a dose-dependent manner. This assay allowed us to examine the potency of each purified protein on HepG2 cells (Fig. 10). The data showed that hPCSK9-RSDY is ϳ30-fold more effective in reducing DiI-LDL uptake with an EC 50 of 1.7 Ϯ 0.1 nM (0.12 g/ml) compared with 49 Ϯ 6 nM (3.5 g/ml) for hPCSK9-RS.
Interestingly, although PCSK9 from humans and mice exhibits an overall 78% sequence homology, the activity of mouse PCSK9 has never been directly compared with human PCSK9 on LDLR degradation. DiI-LDL incorporation in human HepG2 cells revealed no significant difference between the activity of PCSK9 from these two species, with hPCSK9-RS exhibiting an EC 50 of 49 Ϯ 6 nM (3.5 g/ml) and mPCSK9-RS giving a value of 36 Ϯ 6 nM (2.6 g/ml) (Fig. 10), and with PCSK9-RSDY still being ϳ30-fold more efficient than either hPCSK9-RS or mPCSK9-RS to reduce DiI-LDL uptake. These results showed for the first time that when added extracellularly to human cells, human and mouse PCSK9 exhibits similar potencies to enhance the degradation of the LDLR.
LDL Uptake Assay in Knockdown Cell Lines-Following the production, purification, and characterization of the three different recombinant PCSK9s, our goal was then to characterize and compare the LDL incorporation profile of our different knockdown cell lines in the DiI-LDL cell-based assay. LDL incorporation profile was performed on stable PCSK9-KD and AnxA2-KD in Huh7 and HepG2 cells, as well as in HepG2 overexpressing AnxA2-HA (Fig. 11). We tested the dose-dependent effect of 0.1-10 g/ml of the double mutant hPCSK9-RSDY on the levels of DiI-LDL incorporation into HepG2 PCSK9-KD cells as compared with control NT cells. hPCSK9-RSDY exhibits a similar EC 50 value of 0.23 and 0.22 g/ml in PCSK9-NT and PCSK9-KDcells, respectively (Fig. 11A). Moreover, the maximal reduction of LDL uptake in both cell lines occurs at similar concentrations of hPCSK9-RSDY. However, basal LDL incorporation in the absence of recombinant protein is higher in HepG2 PCSK9-KD cells compared with NT cells. Thus, in HepG2 PCSK9-KD cells, basal incorporation is 55% higher than in HepG2 NT cells (Fig. 11A), in agreement with the reported  higher levels of cell-surface LDLR in the absence of endogenous PCSK9 (10). Equivalent results were obtained in HepG2 NT and HepG2 AnxA2-KD cells with EC 50 value of 0.20 and 0.19 g/ml, respectively. Once more, maximal reduction in LDL uptake was achieved with a similar concentration of hPCSK9-RSDY, and basal LDL incorporation was 40% lower in HepG2 AnxA2-KD versus nontarget cells (Fig. 11B). This agrees with the fact that AnxA2 can inhibit PCSK9 activity, and hence its removal would result in lower levels of basal cell-surface LDLR and consequently of DiI-LDL incorporation, as observed. In agreement, an opposite effect on basal DiI-LDL uptake was observed in HepG2 cells stably overexpressing AnxA2, which exhibit a 50% higher uptake than control HepG2 cells (Fig.  11C). LDL incorporation profile from Huh7 PCSK9-KD cells showed similar results to those in HepG2 PCSK9-KD cells, except that the basal DiI-LDL incorporation is higher by only 25% in Huh7 PCSK9-KD cells (Fig. 11D) versus 55% HepG2 PCSK9-KD cells (Fig. 11A). Similar to HepG2 cells, in Huh7 cells no significant difference was detected between hPCSK9-RSDY EC 50 of 0.70 g/ml for nontarget cells and 0.84 g/ml for shPCSK9 cells. This is likely due to the fact that the GOF D374Y is 10 -30-fold more potent than the endogenous wild type PCSK9 in either cell line, and hence the activity of endogenous PCSK9 would not effectively compete with the exogenous highly active hPCSK9-RSDY.
FACS and Western blot analyses showed a 40 -50% decrease in cell-surface and total LDLR in Huh7 AnxA2-KD cells ( Fig.  2A). Yet basal DiI-LDL incorporation in Huh7 AnxA2-KD cells showed a 20% higher uptake (Fig. 11E). No significant difference was observed from a calculated EC 50    11D). Because total protein levels ( Fig. 2A) and cell-surface LRP1 levels (data not shown) remain unchanged, increased basal DiI-LDL uptake in Huh7 AnxA2-KD cells cannot be related to nonspecific binding to cell-surface LRP1.

DISCUSSION
Here, we showed that PCSK9 and LDLR protein levels are differentially sensitive to AnxA2 mRNA knockdown in Huh7 versus HepG2 cells. Reduction of AnxA2 expression in Huh7 cells resulted in increased PCSK9 protein levels both in cells and media, whereas no difference was observed in HepG2 AnxA2-KD cells. Also, LDLR total and cell-surface protein levels were reduced as a direct consequence of increased PCSK9 protein levels in Huh7 AnxA2-KD cells. Because AnxA2 mRNA levels are 5-fold lower in naive HepG2 cells versus Huh7, we sought to determine whether overexpression of AnxA2 in HepG2 cells would affect protein levels of PCSK9. HepG2 cells stably overexpressing AnxA2 resulted in a reduction of PCSK9 protein levels as opposed to an increase in Huh7 AnxA2-KD cells. Based on QPCR analysis, no significant changes were observed in PCSK9 mRNA levels suggesting that removal or overexpression of AnxA2 in Huh7 or HepG2 cells, respectively, did not modify the transcription of PCSK9 or LDLR genes. Additionally, we excluded a possible effect of AnxA2 on PCSK9 protein stability at a post-translational step because its half-life was not changed in cycloheximide time course experiments.
Thus, we hypothesize that AnxA2 might be involved in a negative translational regulation of PCSK9 mRNA. In fact, previous studies have shown that AnxA2 can bind to higher order structures of 3ЈUTR of specific mRNAs such as AnxA2 itself, collagen prolyl 4-hydroxylase-␣(I), N-methyl-D-aspartate R1, and c-Myc (29 -32, 45, 46). These studies identified and suggested a consensus motif 5Ј-AA(C/G)(A/U)G-3Ј that was frequently present in mRNA regions binding AnxA2. AnxA2 is also associated with a subpopulation of messenger ribonucleoprotein complexes (46) and can reduce its own translation (45). Therefore, it is possible that AnxA2 plays a role in both localization and/or activation/repression of translation during transport of specific mRNAs, including that of PCSK9. Trans-acting proteins on 3ЈUTR mainly regulate mRNA transport or their stability, whereas those interacting with the 5ЈUTR are mostly involved in initiation of translation (45). Based on RIP assay results, endogenous PCSK9 mRNA in Huh7 cells is coimmunoprecipitated with AnxA2, which supports the hypothesis of PCSK9 regulation through AnxA2 binding with PCSK9 mRNA. Based on the suggested consensus sequence motif, our analysis reveals the presence of five potential binding sites in the 3ЈUTR of PCSK9 mRNA (Fig. 4A). Removal of the PCSK9 3ЈUTR prevented the coimmunoprecipitation of AnxA2⅐PCSK9 mRNA as compared with the coimmunoprecipitation of PCSK9 containing its 3ЈUTR in HEK293 cells. Therefore, AnxA2 clearly forms a com- plex with the 3ЈUTR of PCSK9 mRNA. Future experiments will be required to confirm whether AnxA2 binds to one or more site(s) in the 3ЈUTR of PCSK9 mRNA to inhibit its translation, as it does for its own translation (45). Furthermore, AnxA2 may also inhibit PCSK9 mRNA transport and/or reduce its stability. Interestingly, cholesterol profiles of AnxA2 Ϫ/Ϫ mice (26) revealed that the loss of AnxA2 is associated with 2-fold higher levels of circulating PCSK9 and 40% higher circulating LDL-cholesterol. Moreover, QPCR analysis did not reveal a significant change in PCSK9 mRNA levels in liver and other tissues. Accordingly, we now hypothesize that loss of AnxA2 expression in mice could result in both a decrease in PCSK9 endogenous inhibition in some tissues (26), but may also be associated with a loss of repression of PCSK9 mRNA translation (this work) leading to the reported 2-fold increased circulating PCSK9 levels (26).
Also, we sought to identify the module within the CHRD of PCSK9 that is responsible for the specific interaction with AnxA2 using previously generated CHRD deletions (⌬M1, ⌬M2, ⌬M3, ⌬M12, ⌬M23, and ⌬M13) using a far-Western blot assay (21). Our results suggest that M3 is the only dispensable module for the specific interaction of PCSK9 with AnxA2. None of the GOF or loss-of-function single point mutants within the M1, M2, or hinge domains analyzed modified the ability of PCSK9 to bind AnxA2 (Fig. 6). However, deletion of either M1 or M2 abrogated the interaction between the two proteins. The specific amino acids responsible for PCSK9-AnxA2 interaction within the M1 ϩ M2 domain remain to be characterized. Interestingly, monoclonal antibodies targeting the M3 (47) or M1 (48) domain of PCSK9 resulted in a maximal 50% inhibition of PCSK9 activity on LDLR. Thus, it is possible that targeting both M1 and M2 domains may result in a more efficient inhibition of PCSK9 activity than only one of them individually.
In this study, we also have presented the production and purification to homogeneity of three recombinants PCSK9s (hPCSK9-RS, hPCSK9-RSDY, and mPCSK9-RS). Sequencing of those mutants demonstrated the presence of prodomain autoprocessing and their resistance to furin cleavage at Arg 218 for hPCSK9 and Arg 221 for mPCSK9. MS analysis of hPCSK9-RS and -RSDY under SAP and PNGase F incubation revealed different molecular mass heterogeneity of secreted PCSK9 as described previously (29). Our results further indicated that secreted recombinant PCSK9 from Drosophila S2 cells resulted in the same PTMs (Tyr sulfation, Ser-phosphorylation, and Asn glycosylation) as observed in secreted PCSK9 from human cell lines.
Using a cell-based assay, we confirmed that recombinant proteins were biologically active and reduced DiI-LDL uptake in a concentration-dependent manner with an EC 50 of 49, 1.7, and 36 nM for hPCSK9-RS, hPCSK9-RSDY, and mPCSK9-RS, respectively. The fact that mouse and human PCSK9 exhibits similar potencies toward enhancing the cellular degradation of the LDLR (Fig. 10) validates the work done using mice overexpressing either mouse or human PCSK9, either as transgenes (4,49) or following recombinant adenovirus expression (22,50). The EC 50 value of hPCSK9-RS produced in Drosophila S2 cells is similar to the one reported for wild type PCSK9 pro-duced in HEK293 cells (EC 50 of ϳ60 nM) (44,51). As shown previously, the mutant GOF D374Y in human recombinant PCSK9 increased the potency by ϳ30-fold with an EC 50 of 1.7 nM similar to Fisher et al. (51) who obtained an EC 50 of 3.5 nM. This increased potency is linked to PCSK9-D374Y higher affinity toward the LDLR (7).
It was previously shown that HepG2 cells are 2-3-fold more sensitive to PCSK9 action on endogenous LDLR than Huh7 cells (41), leading us to postulate that the lower levels of AnxA2 in HepG2 cells (Fig. 1A) may in part be responsible for this observation. This was further supported by the higher DiI-LDL incorporation in HepG2 cells overexpressing AnxA2 (Fig. 11C). This is the rationale behind the routine use of HepG2 (as opposed to Huh7) cells to screen for PCSK9 activity, as also done in this work. However, we were intrigued by the observation that DiI-LDL incorporation was 20% higher in Huh7 AnxA2-KD (Fig. 11E) but 40% lower in HepG2 AnxA2-KD (Fig.  11B). Morel et al. (52) showed that knockdown of AnxA2 expression inhibited endosomal-lysosomal sorting revealing that AnxA2 can regulate endosomal trafficking to lysosomes. It is thus possible that in cells expressing high levels of AnxA2 (e.g. Huh7 cells), the latter may also enhance the rate of degradation of the endocytosed DiI-LDL⅐LDLR complex.
In conclusion, the data presented in this work confirmed the inhibitory property of secreted AnxA2 on PCSK9 activity and defined the M1 ϩ M2 domain as the CHRD binding region of PCSK9 to AnxA2. We further propose that cytosolic AnxA2 could inhibit the translation of PCSK9 mRNA and enhance the sorting of the endocytosed DiI-LDL⅐LDLR complex to lysosomes.