Dual Mechanisms for the Fibrate-mediated Repression of Proprotein Convertase Subtilisin/Kexin Type 9*

Proprotein convertase subtilisin/kexin type 9 (PCSK9) is associated with familial autosomal dominant hypercholesterolemia and is a natural inhibitor of the LDL receptor (LDLr). PCSK9 is degraded by other proprotein convertases: PC5/6A and furin. Both PCSK9 and the LDLr are up-regulated by the hypocholesterolemic statins. Thus, inhibitors or repressors of PCSK9 should amplify their beneficial effects. In the present study, we showed that PPARα activation counteracts PCSK9 induction by statins by repressing PCSK9 promoter activity and by increasing PC5/6A and furin expression. Quantification of mRNA and protein levels showed that various fibrates decreased PCSK9 and increased PC5/6A and furin expression. Fenofibric acid (FA) reduced PCSK9 protein content in immortalized human hepatocytes (IHH) as well as its cellular secretion. FA suppressed PCSK9 induction by statins or by the liver X receptor agonist TO901317. PCSK9 repression is occurring at the promoter level. We showed that PC5/6A and furin fibrate-mediated up-regulation is PPARα-dependent. As a functional test, we observed that FA increased by 30% the effect of pravastatin on the LDLr activity in vitro. In conclusion, fibrates simultaneously decreased PCSK9 expression while increasing PC5/6A and furin expression, indicating a broad action of PPARα activation in proprotein convertase-mediated lipid homeostasis. Moreover, this study validates the functional relevance of a combined therapy associating PCSK9 repressors and statins.

Proprotein convertase subtilisin/kexin type 9 (PCSK9) 3 is associated with autosomal dominant hypercholesterolemia (1) and is a natural inhibitor of the LDL receptor (LDLr) (2). Its autocatalytic activity is necessary to its processing and for the mature form to reach the LDLr and induce its degradation (3). This catalytic activity is not involved in LDLr degradation per se. Indeed, cellular or secreted mature PCSK9 probably acts as a chaperone and prevents the LDLr from being recycled by binding to its EGFA domain (4 -10). Accordingly, hepatic transient overexpression of wild-type PCSK9 in mice (11)(12)(13)(14) or elevated circulating levels in parabiotic mice (15) decrease the LDLr expression and lead to hypercholesterolemia. It has been shown that the proprotein convertase furin and to a lesser extent PC5/6A cleave PCSK9 into a secreted, truncated, inactive form (16). Interestingly, some "gain of function" mutations do not necessarily induce a higher affinity of PCSK9 for the LDLr, but decrease PCSK9 sensitivity to this degradation (4). In contrast, PCSK9 deficiency lowers plasmatic cholesterol concentration and confers protection against cardiovascular disease (17)(18)(19)(20). Thus, treatments inhibiting PCSK9 synthesis, processing, or binding to the LDLr could be useful for hypercholesterolemic patients who do not reach therapeutic goals. In particular, these inhibitors could be added to statins to amplify their effect on the LDLr activity. Surprisingly, statins also increase PCSK9 expression via SREBP-2, a pathway that exerts a break on their efficiency (21)(22)(23). This negative feedback has been shown in vivo in mice deficient for PCSK9 and is supported by studies in patients bearing nonsense mutations for PCSK9 and highly responsive to statins (19,24). Various positive and negative regulatory pathways of the PCSK9 gene have been identified. PCSK9 is up-regulated in mice by the agonist of the liver X receptor (LXR), TO901317 (23). We recently demonstrated that PCSK9 is also up-regulated by insulin as well as the LXR agonist T0901317 via SREBP-1c, which binds a response element located at Ϫ336 bp from the ATG in the promoter region (25). However, we showed that fenofibrate, a ligand for the peroxisome proliferator-activated receptor ␣ (PPAR␣), decreases PCSK9 mRNA and protein quantity in the liver of wild-type mice, but not in PPAR␣ Ϫ/Ϫ mice (26). Fenofibrate is commonly used in clinical practice for its hypotriglyceridemic effects.
Although the benefit of fenofibrate used as a monotherapy is debated (27), a significant decrease of the LDL-C has been observed in patients treated with an association of fenofibrate and simvastatin (28). Furthermore, the "epidemic" of combined hyperlipidemia related to the insulin-resistant states make the association of statins with other drugs targeting triglycerides and HDL levels, as fenofibrate does, desirable (29).
Here we show that various fibrates repressed PCSK9 expression in immortalized human hepatocytes (IHH). We show for the first time that fibrates also increase the synthesis of PC5/6A and furin in a PPAR␣-dependent fashion, suggesting a dual mechanism of action on PCSK9 both at the transcriptional and protein levels. Fenofibric acid (FA), the active form of fenofibrate, reduced the protein content of PCSK9 in IHH as well as its cellular secretion and prevented its accumulation caused by statins or the LXR agonist TO901317. These repressive effects were reproduced at the PCSK9 promoter level. To test the functional relevance of these findings, we verified that FA amplified the effect of pravastatin on LDLr activity.

MATERIALS AND METHODS
Chemicals-Fenofibric acid was purchased from ABCR (Karlsuhe, Germany); other chemicals and drugs came from Sigma. Drugs were dissolved in Me 2 SO, which was also used in controls.
Cell Culture-Human immortalized hepatocytes are described elsewhere (30). They were cultured on collagencoated flasks in William's E medium in the presence of a 10% fetal calf serum. The day before the experiment, the cells were plated in 6-well plates at 1 ϫ 10 6 cells/well. The cells were exposed to various treatments in the presence of a 5% lipoprotein-deficient serum (LPDS) unless notified. Media and drugs were renewed after 24 h of culture. When the secretion was to be analyzed, cells were incubated in 1 ml of medium without serum.
Western Blots-Proteins from 300 l of cell culture medium were precipitated with acetone. Cells were scraped and homogenized in 1ϫ phosphate-buffered saline, 1% Triton X-100. 75 g of proteins were analyzed by Western blot as described elsewhere (25), using a polyclonal rabbit IgG directed against the CRSRHLAGASQELQ peptide (Neosystem, Strasbourg, France), an epitope of the C-terminal domain of human PCSK9, or with a polyclonal anti-PC5 antibody (Abcam, Cambridge UK), a polyclonal anti-furin (abcam), and with the monoclonal anti-␤-actin AC-15 antibody (Sigma).
Real-time PCR-Real-time PCR analysis were performed as described elsewhere (25) with SYBR green PCR Master Mix (Applied Biosystems, Courtaboeuf, France). Oligonucleotides used for real-time PCR are described under supplemental data. Because primers targeting PC5 detect both PC5/6A and 6B, we refer to these mRNA as PC5.
Promoter Analysis-We previously identified the PCSK9 proximal promoter (25). Briefly, long (Ϫ4 bp to Ϫ1460 bp) and short (Ϫ4 bp to Ϫ1027 bp) versions of the proximal promoter (wt) were inserted into the PGL3 basic vector, and versions mutated (mut) for a SREBP response located at Ϫ336 bp were generated. We used long or short hPCSK9wt or hPCSK9mut (0.150 g per well) together with pRL-CMV-Renilla (5 ng/well) in HepG2 cells and long hPCSK9wt (0.5 g per well) with TK-Renilla (10 ng/well), in IHH cells.
The day before the experiment, HEPG2 cells were plated at a density of 300,000 cells per well and IHH cells at a density of 350,000 cells per well, out of 12-well plates, in Dulbecco's modified Eagle's medium, glucose (1 g/liter) (HepG2) or William's medium E (IHH), SVF (10%), 10 g/ml of streptomycin, 100 units/ml penicillin. The transfection was performed with Lipofectamine-2000, according to the manufacturer's instruction (Invitrogen, Cergy Pontoise, France), and cells were maintained in the same culture medium without serum for 20 h. The analysis was performed with the Dual-Luciferase Reporter Assay System (Promega, Charbonnières, France). TK-Renilla was a generous gift from K. E. Berge. IHH cells were a generous gift from H. Moshage and D. Bernuau.
siRNAs-The following predesigned siRNAs were used as duplexes for PCSK9-knockdown experiments: sense sequence: GGUCUGGAAUGCAAAGUCAdTdT and antisense sequence: UGACUUUGCAUUCCAGACCdTdT. Nontargeting siRNA or negative siRNA were used as a control (Eurogentec, Liege, Belgium). Transfections were carried out with Lipofectamine-2000 (Invitrogen) over 7 h, with a final concentration of 200 nM siRNA in 6-well-plates according to the manufacturer's recommended procedures. The cells were then incubated in fresh medium and grown without serum for 48 h to increase the basal activity of the LDLr.
For PPAR␣ knockdown, a set of duplex siRNA was used (On target plus siRNA smart pool, Dharmacon) under the same conditions as above. After transfection, cells were left in fresh medium with LPDS for 24 h and exposed to vehicle or FA for 8 h.
LDLr Activity-After the incubation period time, cells were washed two times in ice-cold phosphate-buffered saline and then treated with 20 g/ml 125 I-LDL without or with 500 g/ml unlabeled LDL in William's E medium, 5% LPDS, 4% bovine serum albumin, 50 mM HEPES, for 4 h at 4°C. Cells were washed three times in phosphate-buffered saline containing 1% bovine serum albumin and rinsed three times with phosphatebuffered saline. Cells were lysed with NaOH 1 N for 30 min, and the radioactivity counted. 125 I-LDL was corrected for cellular protein. The specific binding was calculated by subtracting the normalized 125 I-LDL radioactivity in the presence of excess unlabeled LDL from radioactivity in its absence.
Statistics-Each experiment is representative of at least two independent experiments with a minimum of triplicates per condition. All values are reported as means Ϯ S.D. The statistical significance was analyzed using a Student's unpaired t test. Values of p Ͻ 0.05 were considered significant.

PCSK9 Expression Is Repressed by Various PPAR␣ Ligands in
Human Immortalized Hepatocytes-To assess whether PCSK9 repression by fenofibrate could be extended to other PPAR␣ agonists and to the human gene, we exposed IHH cells for 24 h to Wy14643 (100 M) or to 250 M FA, clofibrate, and gemfibrozil (Fig. 1A). Cells were cultured with LPDS (5%) to reduce the amount of natural PPAR␣ ligands associated to lipoproteins. As expected, PPAR␣ itself responded positively to the Fibrates Repress PCSK9 and Increase PC5/6A and Furin Expression APRIL 11, 2008 • VOLUME 283 • NUMBER 15 treatments (respectively, ϩ1000%, p Ͻ 0.05, ϩ400%, p Ͻ 0.01, ϩ377%, p Ͻ 0.01, ϩ300% p Ͻ 0.01). Each agonist induced a dramatic decrease of PCSK9 mRNA quantity, (Ϫ75% for clofibrate and Ϫ95% for FA, p Ͻ 0.05), down to almost undetectable levels for Wy14643 and gemfibrozil.
Next, we characterized FA-dependent PCSK9 repression in IHH. We observed a dramatic decrease of the quantity of PCSK9 mRNA within 12 h of treatment at the dose of 250 M (86%, p Ͻ 0.05), with a concomitant 181% (p Ͻ 0.05) increase of the quantity of PPAR␣ mRNA, as expected. Increasing doses of FA resulted in a dose-dependent decrease of PCSK9 mRNA quantity in IHH exposed to FA for 24 h, with an EC 50 of 200 M (p Ͻ 0.05) (Fig. 1C). PPAR␣ was responsive to 100 M FA (ϩ100%, p Ͻ 0.01) and doses above that. FA also repressed PCSK9 in the other human cell line HepG2 (supplemental data, Ϫ40% after 24 h of exposure to 250 M FA, p Ͻ 0.01).
FA Prevents the Induction of PCSK9 in Response to Statins and the LXR Agonist TO901317-To determine whether PCSK9 up-regulation by statins could be affected by FA, we exposed IHH to pravastatin (10 M, PV) and/or FA (250 M) over 24 h ( Fig. 2A). As a positive control, FA treatment increased the quantity of carnitine palmitoyl transferase-I (CPT-1) mRNA (ϩ219%, p Ͻ 0.001) (31). Pravastatin did not interfere with this response. Pravastatin increased the quantity of PCSK9 mRNA, (ϩ96%, p Ͻ 0.01). Similar results were obtained with lovastatin (10 M, data not shown). As published, the increase of the quantity of the LDLr mRNA was weaker in response to pravastatin than for PCSK9 (21) (ϩ26%, p Ͻ 0.05,). FA decreased the quantity of PCSK9 mRNA (Ϫ90%, p Ͻ 0.001). Interestingly, FA counteracted the statin-induced expression of PCSK9 (Ϫ90% for PV ϩ FA versus control, p Ͻ 0.001). FA increased the quantity of LDLr mRNA (ϩ44% p Ͻ 0.05 and ϩ59%) but did not impair its response to pravastatin (ϩ44%, p Ͻ 0.05).
We recently showed that PCSK9 is up-regulated by the LXR agonist TO901317 via SREBP-1c, in murine cells (25). We verified these results in IHH cells and investigated the effect of FA on this pathway (Fig.  2B). After 48 h of exposure, the quantity of CPT-1 mRNA augmented by 163% in the presence of FA (p Ͻ 0.05), showing the integrity of the PPAR␣-related pathway. The effect of FA on PCSK9 mRNA was still important (Ϫ76% compared with control, p Ͻ 0.01). TO901317 (1 M) also increased PCSK9 mRNA (ϩ151%, p Ͻ 0.05) in these human cells. Interestingly, FA suppressed the LXR agonist-dependent induction of PCSK9. However, FA did not suppress SREBP-1c induction in response to TO901317 (Fig. 2B). The LDLr responded positively to FA and TO901317 (ϩ116% and ϩ149%, respectively, p Ͻ 0.01), and when both compounds were added together, an amplification of their effect was observed (ϩ308% compared with control, p Ͻ 0.01, and ϩ89%, p Ͻ 0.05 when compared with FA alone).
FA Impairs the Stimulation of PCSK9 Secretion and Activity by Pravastatin-Next, we verified whether FA impairs PCSK9 protein accumulation in response to pravastatin treatment (Fig.  3B). A Western blot was performed with proteins from cells or the culture medium after they were exposed to FA (250 M) or pravastatin (10 M) for 48 h in the absence of serum. Within the cells, pravastatin increased pro-PCSK9 and PCSK9 total content by 47% (p Ͻ 0.05). FA decreased PCSK9 expression by 65% (p Ͻ 0.05), and no change was observed compared with control when both drugs were added together. In the medium, pravastatin increased the PCSK9 content by 45% (versus control, p Ͻ 0.05). FA diminished PCSK9 quantities by 68% (versus control, p Ͻ 0.01). Interestingly, the repressing effect of FA on PCSK9 protein content was maintained when both drugs were added (Ϫ63%, p Ͻ 0.01).  (16), we decided to test whether PPAR␣ agonists would increase their expression in IHH, concomitantly with the analysis described above. A 24-h-long treatment with Wy14643 (100 M), gemfibrozil (250 M), and fenofibric acid (250 M) increased PC5 mRNA quantity by 600% (p Ͻ 0.001), 370% (p Ͻ 0.001), 116% (p Ͻ 0.05), respectively, and furin expression was raised by 275% (p Ͻ 0.05), 990% (p Ͻ 0.001), 160% (p Ͻ 0.001), respectively (Fig. 1A). The same tendency was observed with clofibrate although not statistically significant. At a dose of 250 M, the kinetics of induction showed a maximal effect of FA after 12 h of exposure for both genes (ϩ200% for PC5 and ϩ400% for furin, p Ͻ 0.05, Fig. 1B). In a 24-h-long treatment and dose-dependent study, a plateau was observed for PC5/6A starting with the lowest dose of FA tested, 100 M (ϩ120%, p Ͻ 0.05) (Fig. 1C). Interestingly, a dose-dependent effect was observed for furin.
Similar results were observed in HepG2 cells (supplemental data). After 24 h of exposure to 250 M, furin mRNA levels were increased by 340% (p Ͻ 0.001) and PC5 by 100% (p Ͻ 0.001).
We verified that these changes in PC5 and furin expression were also present at the protein level (Fig. 3A). Western blot analysis showed that upon FA treatment (250 M for 48 h), both the mature PC5/6A isoform and furin protein quantities were increased by 130 and 70%, respectively (p Ͻ 0.001). The PCSK9 content was reduced by 60% (p Ͻ 0.001).
Next, we verified the effect of statins and the LXR agonist on furin and PC5 expression. No effect of pravastatin or T0901317 was detected on PC5 or furin expression or FA-dependent repression at the mRNA level (Fig. 2, A and B).
We checked whether PPAR␣ is involved in the response of PC5 and furin to fibrate (Fig. 4A). Cells were transfected with nontargeting siRNA or a pool of PPAR␣-targeting siRNAs and cultured for 24 h in medium containing 5% LPDS and then exposed to FA for 12 h. This resulted in the knockdown of 80% PPAR␣ mRNA (p Ͻ 0.001). FA alone increased by 100% the PPAR␣ mRNA quantity, but, as desired, this short treatment was unable to overcome the effect of the siRNAs on the PPAR␣ target gene CPT-1. Indeed FA alone increased the mRNA quantity by 81% (p Ͻ 0.05) although the up-regulation was lost in the presence of the siRNAs. A very similar profile was observed for PC5 and furin strongly suggesting that they constitute genuine PPAR␣ target genes. We observed an important and reproducible decrease of PCSK9 expression in cells depleted in PPAR␣ (Ϫ70%, p Ͻ 0.001) and no additional effect of FA with this knockdown. Because SREBPs are major regulators of PCSK9, we checked their expression under these conditions. Both SREBP-1c and SREBP-2 were significantly down-regulated in PPAR␣ knockdown cells (respectively, Ϫ75%, p Ͻ 0.001 and Ϫ30%, p Ͻ 0.05, p ϭ 0.06 for SREBP1a), suggesting that this   APRIL 11, 2008 • VOLUME 283 • NUMBER 15 might be the cause of PCSK9 mRNA fall-down. Their expression levels were not affected by FA alone with or without SiPPAR␣.

Fibrates Repress PCSK9 and Increase PC5/6A and Furin Expression
FA Abolishes PCSK9 Proximal Promoter Response to Statins-We showed that the PCSK9 proximal promoter contains a response element for SREBP-1c located at Ϫ336 bp from the ATG (25), and it is known that in vivo and in HepG2 cells, PCSK9 is regulated by statins and SREBP-2 (21-23). First, we verified whether PCSK9 response to statins was mediated by the proximal promoter. To reproduce the experimental conditions of the literature quoted above, we chose lovastatine and HepG2 cells as a model (Fig. 4B). We transfected the cells with either a long version or a short version of PCSK9 proximal promoter (long wt, short wt), or with the equivalent constructs mutated for the SREBP response element (long mut, short mut). After transfections, cells were cultivated in the presence of lovastatine (10 M), in the presence or not of the cholesterol precursor mevalonolactone (2.5 mM mevalonate). As we previ-ously showed (25), the short version of the promoter reproducibly exhibited a higher activity than the long version (p ϭ 0.07, ns), suggesting that a repressor element might exist between Ϫ1460 bp and Ϫ1027 bp. Lovastatine increased the wild-type promoter activity by 450% (p Ͻ 0.001) for the long version and by 512% (p ϭ 0.012) for the short version, showing that there is probably no response element involved in this pathway between Ϫ1460 bp and Ϫ1027 bp. The activation by lovastatine was significantly reduced by 66% with the concomitant addition of mevalonate (long wt "lovastatine" versus"lovastatine ϩ mevalonate," p Ͻ 0.001), confirming that the pathway involved is indeed cholesterol-related. Mutating the SREBP response element impaired severely the activity of the promoter itself, suggesting that it is a critical element for PCSK9 expression (long wt versus long mut: Ϫ65%, p Ͻ 0.001, short wt versus short mut: Ϫ68%, p Ͻ 0.05). The mutated versions of the promoter were still responsive to statins but to a lesser extent than for the wild-type version (183% increase for long mut versus long wt, p Ͻ 0.01, and ϩ118% fold increase for short mut versus short wt), suggesting that this element is the major site accountable for the effect of lovastatine.
Next, we verified whether FA would repress the activation of PCSK9 promoter by statins (Fig.  4C). To parallel the data of the present article, we extended our results to IHH and showed that the activity of the long version of the promoter is increased by 260% in response to pravastatine (10 M, p Ͻ 0.001) and reduced by 47% by FA alone (p Ͻ 0.001). Interestingly, the concomitant addition of FA (250 M) to pravastatin suppressed the activation of the promoter, suggesting that this is the mechanism underlying PCSK9 repression by PPAR␣ ligands.
FA Amplifies the Effect of Pravastatins on LDLr Activity-PCSK9 down-regulation by FA should result in a better efficacy of statins on LDLr activity. First, we validated that our cellular model was appropriate to evaluate interactions between the LDLr pathway and PCSK9. Indeed, we observed that PCSK9 knockdown (Ϫ40 to Ϫ60% cellular and secreted protein content depending on experiment) by siRNA resulted in a 200% increase in LDLr activity (Fig. 5A). Next, we verified our hypothesis by incubating IHH cells in the presence of FA (250 M) for 24 h followed by a 24-h treatment with pravastatin . Fenofibric acid transcriptionally regulates PCSK9, PC5, and furin. A, IHH cells were transfected with nontargeting siRNA (Si negative) or siRNA directed against PPAR␣ (SiPPAR␣) and cultured as described under "Materials and Methods." Gene-specific expression was measured by real-time PCR analysis. Results represent the mean Ϯ S.D. *, p Ͻ 0.05; **, p Ͻ 0.01. B, HepG2 cells were transfected with a construct containing a long or short version of the human PCSK9 proximal promoter, wild type or mutated for the SREBP response element located at Ϫ336 bp from the ATG (wt, mut) ("Materials and Methods"). A, cells were exposed to lovastatine (10 M) and/or mevalonolactone (2.5 mM). B, IHH cells were transfected with the long version of the PCSK9 wild-type promoter and immediately exposed to PV (10 M) and/or fenofibric acid (250 M). Results represent the mean Ϯ S.D. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.

Fibrates Repress PCSK9 and Increase PC5/6A and Furin Expression
alone or in combination (5 M) (Fig. 5B). Cells were washed and exposed to 125 I-LDL as described under "Materials and Methods" at 4°C. As expected, pravastatin increased the amount of specifically bound 125 I-LDL by 500% (p Ͻ 0.01). FA increased the binding by 170% (p Ͻ 0.05). In agreement with our hypothesis, when both drugs were added together, an additional 30% increase in LDLr activity was observed compared with pravastatin alone (varying from 18 to 46% in a total of five independent experiments, data not shown).

DISCUSSION
PCSK9 is a natural post-transcriptional inhibitor of the LDLr pathway, and its deficiency results in very low LDL-C levels and a protection against cardiovascular disease (2). Here we show that various fibrates, ligands for PPAR␣, repressed PCSK9 expression in IHH. Fibrates also increased the expression of PC5/6A and furin in a PPAR␣-dependent fashion. These proprotein convertases degrade PCSK9, suggesting the existence of a dual mechanism of repression of PCSK9 targeting both the mRNA synthesis and the protein degradation. Fenofibric acid, the active form of fenofibrate, prevented PCSK9 accumulation within the cell or in the medium due to statins or the LXR agonist TO901317. These repressive effects were reproduced at the promoter level. As a functional test of the relevance of our findings, we showed that FA amplified the effect of pravastatin on LDLr activity.
PPAR␣ is an essential participant in lipoprotein metabolism. It displays normolipidemic properties by governing the expression of many genes involved in fatty acid catabolism and trans-port, reverse cholesterol transport via HDL, as well as enzymes involved in the remodelling of lipoproteins (32). Its activity in peripheral tissues has also been involved in the pleïotropic effects of statins (33,34). Our data support a broader role for PPAR␣ in LDL metabolism (Fig. 6). However, it is striking that one would expect a more potent effect of fibrates on LDL-C in humans, considering the intensity of the repression on PCSK9 we observed and the relative 28% decrease of LDL-C observed in patients heterozygote for nonsense mutations (18). Furthermore, we observed that FA alone also increases LDLr activity and it is mRNA cellular quantity, a trait that was previously found for gemfibrozil (35).
Fenofibrate activates lipases and promotes LDL remodelling into particles with optimal binding capacities for the LDLr (36). The overall limited effect of fibrates on LDL-C suggests the existence of a negative feedback pathway similar to PCSK9 upregulation by statins. If this hypothesis is true, identifying this pathway could help with improving PPAR␣ agonists and making them more efficient toward the LDL-C. This is supported by the observation that significant reductions of LDL-C, following PPAR␣ activation promotes PCSK9 repression by inhibiting its transcription and increasing PC5 and furin expression. Fibrates also directly increase the synthesis of the LDLr and promote a remodelling of LDLs, reducing the number of dense proatherogenic particles and increasing the number of large LDLs with optimal binding capacities for the LDLr. Interestingly, PC5 and furin cleave the lipoprotein lipase (triglyceride lipase) and degrade the phospholipase endothelial lipase (43), an enzyme known for hydrolyzing phospholipids from HDLs but that also converts large LDL into smaller LDL with less affinity for the LDRr. This promotes higher plasmatic concentrations of HDL-C (44) and lower LDL-C. In humans, the relative low efficiency of PPAR␣ on LDL clearance compared with statins suggests the existence of a negative feedback pathway. APRIL 11, 2008 • VOLUME 283 • NUMBER 15

JOURNAL OF BIOLOGICAL CHEMISTRY 9671
fibrate administration, were still reported in clinical trials. This was the case for gemfibrozil in the Helsinki Heart study (37). Although less efficient than statins (38), the decrease of plasma LDL-C levels observed with fenofibrate in short term studies reached up to 30% (39). In the Fenofibrate Intervention and Event Lowering in Diabetes study (FIELD), the decrease compared with the control group was 12% after 4 months of treatment and was only 5.8% after 5 years. But the results are questionable because of a strong drop in treatment with statins that contributed to the decrease of LDL-C concentrations in both placebo and fenofibrate groups (27). However, our data suggest that fenofibrate could be the most useful in combination with statins. In patients with combined dyslipidemia, co-administration of fenofibrate with fluvastatin or simvastatin resulted in an additional decrease of 10% of LDL-C compared with the monotherapy (28,40,41). A large cardiovascular end point trial, ACCORD (Action to Control Cardiovascular Risk in Diabetes), is currently in progress and will evaluate the beneficial effect of such a combined therapy. Large differences in lipid profiles were observed in patient carriers of PCSK9 nonsense mutations, suggesting a complex pathway of regulations (18,42). It is unknown whether variations of the individual responses to the combined treatment correspond to PCSK9 mutations or polymorphisms, and whether the repression of PCSK9 is sustainable. Evaluating the amount of circulating PCSK9 in patients before and after treatment by fibrates would be needed to test these hypotheses and verify our results in vivo in humans.
Here we show for the first time that PC5/6A and furin are positively regulated by fibrates in a PPAR␣-dependent way. Although it would be very informative to show the relative contribution from each pathway to the fibrate-mediated inhibition of PCSK9 (transcriptional or enzymatic), we have not been able to successfully knockdown PC5 and furin at the same time.
PC5/6A and furin are known for cleaving lipoprotein lipase and endothelial lipase (43). Interestingly, while this manuscript was under review, Jin et al. (44) showed that hepatic cleavage and inhibition of the endothelial synthase by furin and PC5/6A modulate HDL metabolism. In particular, their inhibition resulted in an increase of HDL-C levels. We propose that the regulation of both these genes by fibrates and PPAR␣ contributes to the beneficial effect of these drugs on HDL-C (Fig. 6) (37,38).
Although our findings were confirmed in the human HepG2 cells (supplemental data), we found no regulation of PC5 or furin in vivo in mice (data not shown) after a treatment by fenofibrate, suggesting strong species specificity. We designed oligonucleotides annealing to the junction of human exons 7 and 8, thus targeting both PC5/6A and PC5/6B transcripts. Both transcripts are present in the liver (45,46). PC5/6B encodes a larger protein with a transmembrane domain and is unable to cleave PCSK9, contrary to the shorter, soluble PC5/6A (16). Our Western blots confirm that the PC5/6A isoform is up-regulated by fenofibric acid.
We were surprised to observe that PCSK9 expression fell in response to the knockdown of PPAR␣ because we expected the opposite. No additional decrease was observed in the presence of fenofibric acid, in agreement with a role for PPAR␣. It is also reminiscent of what we observed in vivo in mice at the mRNA level, but it does illustrate the complexity of the regulatory pathways at play, in particular when they result in a repressive effect (26). We propose that at least in vitro, the down-regulation of SREBPs is responsible for these effects, although more work is needed to investigate the abundance of their nuclear forms under these conditions. LXR is a very promising therapeutic target because of its anti-atherogenic properties. However, in vivo, LXR agonists also exhibit hypertriglyceridemic and lipogenic properties (47). Whether PCSK9, which acts on VLDL production and triglyceridemia under specific pathophysiological conditions, is involved in this process is unknown (26,48). Here we showed that the LXR agonist TO901317 is an activator of PCSK9 in human cells. As published recently, it also induced LDLr transcription (49). Interestingly, we found an additive effect of both drugs on LDLr regulation. This suggests that LXR agonists might potentiate the effect of PCSK9 inhibitors on LDLr activity.
To our knowledge, this study is the first formal demonstration that PCSK9 promoter activity is activated by statins, although it was already known that statins and SREBP-2 increase PCSK9 expression (21)(22)(23). The decrease of activity and the lower sensitivity to statins consequent to the mutation within the SRE located at Ϫ336 bp strongly suggest that this sequence mediates the promoter response to statins, as it does with insulin and SREBP-1c (25). It would be interesting to genotype individuals with abnormal LDL-C levels for this region of thegene.AlthoughourexperimentssuggestthatPCSK9,fibratedependent repression is occurring transcriptionally via the inhibition of the promoter activity, the exact mechanism is still to be determined. It has been shown in rat hepatocytes that clofibrate and Wy14682 decrease the amount of nuclear active SREBP-2 (50). The hypothesis that this down-regulation is responsible for PCSK9 repression would be difficult to reconcile with the concomitant induction of the LDLr we observe.
It has recently been shown that PCSK9 probably acts as a molecular chaperone on the LDLr by binding to its extracellular domain (4,5,10). Several strategies have been proposed or discussed in the literature to inhibit PCSK9, including inhibitors of its catalytic activity or peptides preventing its binding to the LDLr as well as antisense oligonucleotides to shut down its synthesis (2,51). Our results suggest that PCSK9 expression could be successfully repressed at the mRNA level, a strategy that may be useful for patients suffering from mutations that considerably enhance the affinity of PCSK9 for the LDLr and therefore makes patients less responsive to statins, like the D374Y (52).
In conclusion, various fibrates repressed human PCSK9 expression in a transcriptional fashion and increased PC5/6A and furin expression in a PPAR␣-dependent fashion. By defining PPAR␣ as being a simultaneous regulator of three proprotein convertases, this study identifies a new class of targets for this nuclear receptor and reinforces their role in lipid homeostasis. Based on the recent findings of Jin et al., we also propose that part of the beneficial effects of fibrates on HDL-C is mediated by this regulatory pathway. Moreover, this study supports the functional relevance of a combined therapy associating PCSK9 repressors and statins.