Proprotein Covertases Are Responsible for Proteolysis and Inactivation of Endothelial Lipase*

Plasma lipoprotein metabolism is tightly regulated by several members of the triglyceride lipase family, including endothelial lipase (EL) and lipoprotein lipase (LPL). Our previous work suggested that EL is proteolytically processed. In this report, we have used a combination of epitope tagging, mutagenesis, and N-terminal sequencing to determine the precise location of the cleavage site within EL. The cleavage occurs immediately after the sequence RNKR, a known recognition sequence for the proprotein convertase (PC) family. We demonstrate that some PCs, but not all, can proteolytically cleave EL at this site and thereby directly regulate EL enzymatic activity through modulating EL cleavage. Furthermore, specific knockdown of individual PCs proves that PCs are the proteases that cleave EL in human endothelial cells. Interestingly, a homologous site in LPL is also cleaved by PCs. This action is unusual for PCs, which are traditionally known as activators of pro-proteins, and highlights a potential role of PCs in lipid metabolism through their proteolytic processing of lipases.

Site-specific proteolysis is crucial in regulating many fundamental biological pathways, including the sequential initiation of activation of blood coagulation factors and activation of caspases and digestive enzymes (1)(2)(3). The proprotein convertase (PC) 5 family is composed of at least nine members and belongs to the subtilisin superfamily of serine endoproteases (4). The basic amino acid-specific PCs catalyze the proteolytic maturation of a strikingly diverse collection of substrates at paired basic amino acid processing sites to generate biologically active molecules in the secretory pathway (5). Some PC substrates are proteins involved in lipid metabolism. Furin, the first discovered and best characterized PC, was shown to cleave the low density lipoprotein receptorrelated protein (6), and a mutation of the processing site of chicken low density lipoprotein receptor-related protein impaired efficient exit from the endoplasmic reticulum (7). Subtilisin-kexin isozyme-1/Site-1 protease (SKI-1/S1P), an early Golgi-localized PC (8), cleaves membranebound sterol regulatory element-binding proteins and releases the active subunit of sterol regulatory element-binding proteins to maintain cholesterol homeostasis (9).
PCs were considered to be redundant because a large body of in vitro work had shown that several PCs could cleave the same substrate. Recently, lessons from the studies of neural apoptosis-regulated convertase 1 (NARC-1), a newly discovered PC (10,11), reinforced the importance of individual PCs in regulating plasma lipid metabolism. However, the substrate of NARC-1 is not yet identified.
Endothelial lipase (EL) is a recently described modulator of lipoprotein metabolism that belongs to the triglyceride lipase family, which also includes lipoprotein lipase (LPL) and hepatic lipase (HL) (12). Like LPL and HL, the EL protein is secreted and binds to heparan sulfate proteoglycans on the endothelial cell surface, where it interacts with lipoproteins (13)(14)(15). Both overexpression and loss of function studies have shown that EL is important in the regulation of HDL cholesterol levels (16 -19). It may also have a role in apoB-containing lipoprotein metabolism (20). EL protein was detected in endothelial cells, smooth muscle cells, and macrophages in human atherosclerotic lesions (21). Unlike LPL and HL, EL is up-regulated in response to cytokines, shear stress, and cyclic stretch (22,23). Furthermore, EL deficiency reduced the development of atherosclerosis in apoE Ϫ/Ϫ mice (24).
Interestingly, two major forms of the EL protein, the full-length and a smaller form, are detected in conditioned media from endothelial cells by Western blotting using an anti-hEL peptide antibody generated against an N-terminal peptide (16). Expression of EL cDNA in vitro and in vivo results in generation of the same two proteins (25). We hypothesized that the smaller form of EL is the product of a specific proteolytic cleavage of the full-length protein. In this report, we document the site of this processing, elucidate the specific sequence requirements, demonstrate that specific PCs are responsible for the cleavage of EL, show that cleavage reduced EL activity, and provide evidence that LPL is also cleaved by PCs.

EXPERIMENTAL PROCEDURES
Materials-We obtained horseradish peroxidase-conjugated goat anti-mouse and anti-rabbit IgG (affinity-purified) from Jackson Immu-noResearch Laboratories, monoclonal anti-myc (IgG fraction) from the culture medium of hybridoma clone 9E10 (American Type Culture Collection), human embryonic kidney HEK293 cell line from American Type Culture Collection, BisTris NuPage Gels (10% resolving gel; 4% stacking gel) from Invitrogen, peptides, anti-FLAG-M2 (anti-FLAG) antibody from Sigma, Lipofectamine Transfection reagent from Invitrogen, SuperScript Plasmid System for cDNA Synthesis and Plasmid Cloning from Roche, and carbonyl cyanide m-chlorophenyl hydrazone, cholchicine, monensin, and brefeldin A from Calbiochem. * This work was supported in part by a Scientist Development Grant from the American Heart Association (to W. J.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental experimental procedures and Fig. S1. 1  Cell Culture-Monolayers of HEK293 cells were cultured in 5% CO 2 at 37°C in Dulbecco's modified Eagle's medium containing 100 units/ml penicillin and 100 g/ml streptomycin supplemented with 10% (v/v) fetal bovine serum. For transient transfection, cells were set up on day 0 (1.3 ϫ 10 5 cells/well on a 6-well plate). On day 2, cells were transfected with 1 g of the indicated plasmid using Lipofectamine in 1 ml of Opti-Dulbecco's modified Eagle's medium (Invitrogen). Five hours after transfection, 1 ml of Dulbecco's modified Eagle's medium containing 100 units/ml penicillin and 100 g/ml streptomycin supplemented with 20% (v/v) fetal bovine serum was added into each well. Twenty-four hours after transfection, the cells were switched to Dulbecco's modified Eagle's medium in the presence of 10 units/ml of heparin (Sigma). The culture media were collected at 48 h post-transfection. The media collected with heparin are referred to as conditioned media. After incubation for 48 h, the cells were harvested, resuspended in 450 l of buffer containing 150 mM NaCl, 1.0% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 50 mM Tris-Cl, pH 8.0, with protease inhibitor mixture (Roche). The lysate was centrifuged at 16,000 ϫ g for 30 min, and the supernatant was collected. For adenoviral infection, 3000 particles per cell were used for COS-7 cells, and 800 particles per cell were applied for HUVECs.
Western Blotting-SDS-PAGE and immunoblot analysis were carried out as described (23). The primary antibodies used were a rabbit polyclonal antibody against human EL peptide (amino acids 8 -23) and mouse monoclonal antibodies against myc or FLAG. Treatment of samples with heparin-Sepharose beads was carried out as previously reported (16). F, N, and C denote the full uncleaved precursor form, the cleaved N-terminal fragment of EL, and the cleaved C-terminal fragment of EL, respectively.
PC Processing of EL and LPL in HEK293 Cells-HEK293 cells stably expressing myc-tagged EL or transiently expressing myc-tagged LPL were transiently transfected with pIRE2-EGFP recombinant cDNAs coding for each of the convertases: furin, PACE4, PC5A, PC5A-⌬C (lacking the Cys-rich domain), PC5B, PC7, SKI-1, and NARC-1. Two days post-transfections, cells were pulse-labeled for 3 h in the presence of [ 35 S]Met ϩ Cys. Cell media were subjected to immunoprecipitated with the anti-myc antibody, and the immunoprecipitates were resolved on 8% SDS-PAGE gels and then autoradiographed, as reported (8, 26 -28).
Reverse Transcriptase-PCR-One microgram of total RNA was converted into cDNA using the SuperScript first-strand synthesis system for reverse transcription-PCR (Roche Molecular Biochemicals). All primers designed were identical to the previous report (29).
Lipase Activity Assays-The lipoprotein binding assays were performed as previously described (30). Both triglyceride lipase and phospholipase activities were assayed as previously described (31).

RESULTS
EL Is Proteolytically Cleaved-Conditioned media from various types of human endothelial cells were treated with heparin-Sepharose beads, and proteins bound to the beads were subjected to immunoblotting using an anti-hEL antibody that detects only the N-terminal epitope of EL (Fig. 1B). All of the endothelial cells tested expressed EL as demonstrated by the presence of an immunoreactive 68-kDa form of EL. In addition, the media also contained a 40-kDa protein recognized by this same antibody (Fig. 1A). The ratio of 40/68-kDa forms of EL differed considerably among the different endothelial cells, suggesting that the processing of EL is cell-type dependent.
We hypothesized that the 40-kDa protein was derived from fulllength EL by proteolytic cleavage. To identify the C-terminal cleavage product of EL, we prepared cDNAs encoding full-length EL with either a C-terminal myc-His tag (ELmyc) or a FLAG tag version of human EL (ELflag) (see supplemental experimental procedures). After transfection in HEK293 cells, the anti-N-terminal EL antibody detected fulllength ELmyc or ELflag slightly larger than the untagged EL ( Fig. 1C) as well as N-terminal 40-kDa fragments that were identical in size to untagged EL and to that found in endothelial cell media (Fig. 1C, lane 5). This suggests that EL is proteolytically cleaved even when expressed in cell types other than endothelial cells. The anti-myc antibody detected the full-length ELmyc as well as a smaller 33-kDa band, and the anti-FLAG antibody detected the full-length ELflag as well as a 33-kDa band (Fig. 1D). This suggests that the 33-kDa band represents the tagged C-terminal cleavage product. In side by side comparative studies, we demonstrated that the myc tag has no effect on EL processing (data not shown).
To better understand the kinetics of cleavage of EL, stable HEK293 cells expressing human ELmyc were pulse-labeled in a chemically defined medium with [ 35 S]methionine. EL was immunoprecipitated from media obtained at different time points using a polyclonal rabbit anti-adhEL antibody that recognizes both N-and C-terminal regions of EL (23). As expected, more EL protein was recovered from the media in the presence of heparin. Control antibody did not pull down similar size bands (data not shown). Both in the presence and absence of heparin, the full-length ELmyc appeared in the medium first, with the smaller fragments appearing later (Fig. 1E). The ratio of 40/68-kDa forms of EL changed as a function of the incubation time, suggesting that the processing of EL is time-dependent.
Processing of EL Requires a PC Consensus Sequence-To narrow down the site of cleavage, we prepared a series of truncated ELmyc constructs ( Fig. 2A) (supplemental experimental procedures). An initial study verified that EL is cleaved between amino acids 297 and 346 (data not shown). cDNAs encoding the full-length EL and the truncated versions were transiently transfected into HEK293 cells. All of the truncated EL proteins as well as full-length EL were easily detected in the media with the anti-N-terminal EL antibody (Fig. 2B). The shorter truncation mutants ⌬323 and ⌬327 were easily detected by anti-myc antibody, but the longer ⌬330 and ⌬334 mutants were not detected with the same antibody (Fig. 2C). Thus, the loss of the C-terminal myc epitope through cleavage occurred between residues 327 and 330. This placed the cleavage in the hinge region between the N-and C-terminal domains of EL. In contrast to the limited processing of full-length of EL, all truncated forms of EL were either completely cleaved or completely uncleaved, indicating the importance of the C terminus of EL in regulating this process.
To identify the precise cleavage site, we overexpressed full-length EL using a plasmid encoding ELmyc in HEK293 cells by transient transfection, purified proteins containing a His epitope from cell culture medium using a nickel-nitrilotriacetic acid column, separated these proteins by SDS-PAGE, and transferred them to membranes. The C terminus of ELmyc was confirmed by Western blotting. N-terminal sequencing (supplemental experimental procedures) of this protein revealed that the cleavage of EL occurs immediately after residue 330, the second arginine of an RNKR 330 sequence (Fig. 2D).
To determine the amino acid residues that are important for cleavage of EL, we made a systematic series of mutations in which alanine was individually substituted for the amino acids around the RNKR sequence. Fig. 2E shows representative results. Only two single alanine substitutions reduced EL cleavage: R327A (P4 mutant; partial cleavage) and the R330A (P1 mutant; no cleavage). Mutation of the positively charged arginine to lysine at position 327 still permitted EL cleavage, but substitution of lysine for arginine at position 330 abolished cleavage. In addition, two multiple alanine replacements abolished EL cleavage and all included the R330A substitution. These data indicate that arginine 327 at the P4 position relative to the cleavage site and arginine 330 at the P1 position are the only two residues in the region required for effective proteolytic processing, although we cannot rule out the importance of the P6 Lys 325 that may compensate for the absence of a P4 arginine (8) and explain the partial cleavage seen in the R327A mutant. There is no dramatic change of EL secretion even though the cleavage is greatly inhibited by mutagenesis (data shown in Fig. 6). Deletion of asparagine 328 or insertion of alanine within the RNKR sequence did not affect secretion of the EL protein into the culture medium, but dramatically reduced EL cleavage (Fig. 2F). These data indicate that an intact RNKR sequence length is also required for effective EL cleavage.
Processing of EL Occurs Both Intracellularly and Extracellularly-A stable HEK293 cell line expressing ELmyc was used to determine the subcellular location of the proteolytic cleavage (Fig. 3A). Carbonyl cyanide m-chlorophenyl hydrazone, which interferes with transport of glycoproteins out of the rough endoplasmic reticulum to the Golgi (32), blocked secretion of full-length of EL into the medium but did not result in intracellular accumulation of the 40-kDa fragment in the cell lysate, indicating that proteolytic pathways associated with the endoplasmic reticulum do not participate in processing EL. Monensin, which interferes with intra-Golgi vesicular transport (33), and brefeldin A, which induces the fusion of the rough endoplasmic reticulum with the early Golgi compartment (34), also did not result in intracellular accumulation of the N-terminal fragment. These results demonstrate that proteolytic processing of the EL is unlikely to occur prior to the trans Golgi network. In contrast, colchicine, which interferes with post-Golgi transport (35), resulted in substantial accumulation of the N-terminal fragment in the cell lysate, indicating that processing of EL can occur in the late Golgi. We could not detect an intracellular cleavage product of EL in HUVECs, which might be below the limit of detection using our antibody (data not shown).
As another approach to study the intracellular proteolytic processing of EL, the endoplasmic reticulum retention sequence Lys-Asp-Glu-Leu (KDEL) (36) was inserted at the C terminus of EL. Cells expressing the ELkdel construct had no secretion of EL into the media (data not shown) but had substantial EL in the cell extract (Fig. 3C). Endoglycosidase H sensitivity of EL further proved that ELkdel did not reach the Golgi (Fig. 3C). In the cells expressing wild type EL, the N-terminal fragment was clearly seen in the cell extract (Fig. 3C). In contrast, no N-terminal fragment was seen in cells expressing ELkdel. This suggests that cleavage occurs distal to the early Golgi and is consistent with the chemical inhibitor studies.
Overall, these results suggest that proteolytic processing of EL is either a late intracellular or post-secretion event or both. When we treated the stable HEK293 cell line expressing ELmyc with 10 units/ml heparin at 4°C for 30 min, mainly full-length EL is recovered from the conditioned medium (Fig. 3D, 0 h), suggesting that the secreted form of EL is largely full-length. To test whether extracellular proteolysis can occur, the medium containing full-length ELmyc was incubated at 37°C in the absence or presence of HUVEC for up to 48 h. In the absence of cells, no full-length EL was proteolytically cleaved (data not shown). In the presence of cells, the full-length ELmyc was gradually converted to the 33-kDa C-terminal fragment with t1 ⁄ 2 about 12 h (Fig. 3D). Thus, HUVEC are able to secrete, or expose at the cell surface, the enzyme that cleaves EL.
Proprotein Convertases Are Responsible for EL Cleavage-Because the sequence RNKR is a typical PC recognition site and PCs are known to be Amino acid alignment of the hinge region between the N and C terminus of human EL, mouse EL, and rat EL is shown at the top. Bold letters denote amino acid residues that are identical in all three ELs. Residues individually mutated to alanine in human EL are indicated below the alignment. The cleavage of each mutant relative to that of wild-type human EL, as determined by immunoblot analysis of transfected cells, is shown at the right. A value of 0 denotes undetectable cleavage; a value of 100% denotes cleavage of the mutant protein equivalent to that of wild-type human EL. F, immunoblot analysis of FLAG-tagged EL in HEK293 cells transfected with wild-type and RNKR mutants. The RNKR sequence between position 327 and 330 in human EL was mutated to the indicated sequences. EGFP, enhanced green fluorescent protein; F, N, and C, full uncleaved precursor form, the cleaved N-terminal fragment of EL, and the cleaved C-terminal fragment of EL, respectively. active after the late Golgi, we used PC inhibitors to study their effects on the proteolytic processing of EL. DPC (decanoyl-Arg-Val-Lys-Argchloromethyl ketone) is a modified peptide that binds covalently to the substrate binding site of PCs and inhibits their activity (37). To determine whether DPC affects the cleavage of EL, HEK293 cells transfected with a plasmid encoding hEL were treated with DPC for 24 h. As seen in Fig. 4A, DPC caused a reduction of cleavage of EL, without accumulation of the full-length EL. In contrast, EL cleavage was not affected by other serine protease inhibitors, including aprotinin (0.1 M), leupeptin (1 M), Pefabloc SC (1 mM), or a protease inhibitor mixture obtained from Roche (data not shown).
The prosegments of furin and PC5A are potent autoinhibitors of PC activity by noncovalently attaching to PCs, and profurin as well as proPC5A inhibit both furin and PC5A (27,38). We investigated the ability of overexpressed profurin and proPC5A to inhibit the processing of EL. As seen in Fig. 4B, overexpression of either profurin or proPC5A resulted in a complete inhibition of EL cleavage.
Another PC inhibitor, ␣1-PDX, was generated by mutating the reactive-site loop of ␣ 1 -antitrypsin to contain the minimal consensus sequence for PCs cleavage (-R-I-P-R-) and to act as a competitive inhibitor (39). We investigated the ability of overexpressed ␣1-PDX to inhibit the processing of EL. The processing of EL was apparent when COS-7 cells were infected with AdhEL alone, but co-expression of ␣1-PDX completely blocked EL cleavage (Fig. 4C). Importantly, the expression of ␣1-PDX also significantly reduced the processing of endogenous EL in HUVEC as compared with untreated or Ad-enhanced green fluorescent protein-infected samples (Fig. 4C).
Because the RNKR sequence is the site of cleavage, a synthetic peptide corresponding to this site should competitively inhibit the processing of EL. We studied the effects of two peptides on the processing of EL after expressing ELmyc in HEK293 cells: peptide A, which contains the intact RXKR sequence, and as a control, peptide B, a mutant version containing an RNAA sequence. EL cleavage was substantially reduced by peptide A, whereas peptide B had no effect on cleavage (Fig. 4D). Because these peptides are not expected to enter the cells, this strongly suggests that substantial processing occurs either at the cell surface or in the medium.
Only Some PCs Have the Ability to Cleave EL-To determine which member(s) of the PC family cleave EL, we transiently expressed, in the stable HEK293 cell line expressing ELmyc, each of the PCs individually. PACE4, PC5A, and furin were all found to generate substantial C-terminal fragment from full-length EL (Fig. 5A), but not PC5B, PC5⌬C (without the C-terminal Cys-rich domain) (5), PC7, SKI-1/S1P, or NARC-1. The C-terminal cleavage products of EL were immunoprecipitated and subjected to N-terminal protein sequencing. All contained EL sequence immediately following arginine 330. Thus, these three PCs cleaved EL at the same site where EL is normally cleaved in HEK293 cells. Note that loss of the cysteine-rich domain of PC5A (PC5A-⌬C) resulted in less efficient processing (Fig. 5A), although both PC5A and PC5A-⌬C are expressed at similar levels (data not shown).
We tested whether cleavage by furin, PC5A, or PACE4 required the intact RNKR recognition sequence of EL. For this purpose, we transiently co-expressed wild-type EL and mutant versions of EL (ELR327A and ELR330A) with furin, PC5A, or PACE4 in HEK293 cells. Expression of furin, PC5A, or PACE4 increased cleavage of wild-type EL, whereas ELR327A showed reduced cleavage and ELR330A totally abolished cleavage by these PCs (Fig. 5B). A minor, additional EL band was observed as indicated by an asterisk (Fig. 5B), which may occur through a second cleavage when PCs are overexpressed.
The mRNA expression of PCs was assessed by reverse transcriptase-PCR in HUVEC, human pulmonary microvascular endothelial cells, human dermal microvascular endothelial cells, and human uterine microvascular endothelial cells. PC-specific primers covering at least two exons were designed for all members of the PC family. Furin, PC5A, and PC7 were expressed in all these cells, PACE4 was expressed in human pulmonary microvascular endothelial cells and human uterine microvascular endothelial cells, PC1 was expressed in human uterine microvascular endothelial cells, human pulmonary microvascular endothelial cells, and human dermal microvascular endothelial cells, and PC2 was not expressed in any of these endothelial cells (Fig. 5C). Thus, the PCs that are able to cleave EL are in fact normally expressed in endothelial cells.
To determine the physiological contribution of individual PCs in processing EL in endothelial cells, we generated several lentiviral vec-  1, 3, and 5) or with (lanes 2, 4, and 6) endoglycosidase H (endo H) at 37°C for 2 h. Then, they were subjected to SDS-PAGE and immunoblotted with anti-hEL peptide antibody. D, proteolysis of ELmyc as a function of time. Partially purified ELmyc protein was incubated for 0, 6, 12, 24, or 48 h with HUVEC at 37°C. The samples were then centrifuged, and the supernatants were subjected to Western blot analysis using anti-myc antibody. Experiments were repeated three times. F, N, and C, full uncleaved precursor form, the cleaved N-terminal fragment of EL, and the cleaved C-terminal fragment of EL, respectively.
tors expressing small hairpin RNA against luciferase, furin, or PC5. The efficacy of these constructs in knocking down their corresponding targets was tested by co-transfection in HEK293 cells (data not shown). Then we transduced HUVEC using these lentiviruses and examined the expression of EL in the conditioned media. As shown in Fig. 5D, cleavage of EL was substantially reduced, but not all, in cells infected with Lenti-small hairpin furin and lenti-small hairpin PC5. Therefore, both furin and PC5 play essential roles in the processing of EL in HUVEC.
EL Processing by PCs Reduces Its Lipase Activity-We studied the effect of cleavage on EL enzymatic activity. PC5A, which is expressed in endothelial cells (40,41), cleaved EL in a dose-dependent fashion and reduced lipase (triglyceride lipase and phospholipase) activities in a dose-  The cells were pulse-labeled with [ 35 S]Met/Cys for 2 h and then chased for 4 h. Culture media were collected without heparin and immunoprecipitated with the anti-myc antibody and analyzed by SDS-PAGE on 8% Tricine gels. B, immunoblot analysis of the culture medium of HEK293 cells transfected with the indicated constructs with anti-hEL peptide or anti-FLAG antibody. There is one additional band appearing after furin, PC5A, or PACE4 overexpression, designated by the asterisk. C, total RNA was extracted from the indicated endothelial cells, and reverse transcription-PCR analysis was performed using PC1-, PC2-, furin-, PC5A-, PACE4-, and PC7-specific primers under the conditions described under "Experimental Procedures." D, immunoblot analysis of EL protein in media from HUVEC after transduced with lenti-small hairpin luciferase, lenti-small hairpin furin, lenti-small hairpin PC5, or untreated cells. Experiments were repeated three times. EGFP, enhanced green fluorescent protein; HPMEC, human pulmonary microvascular endothelial cells; HDMEC, human dermal microvascular endothelial cells; HutMEC, human uterine microvascular endothelial cells; F, N, and C, full uncleaved precursor form, the cleaved N-terminal fragment of EL, and the cleaved C-terminal fragment of EL, respectively. dependent manner. In contrast, profurin prevented the cleavage of EL and modestly increased EL lipase activities (Fig. 6, B and C).
To confirm that the effects of PC5A and profurin on EL lipase activities are dependent on an intact RNKR sequence, we expressed wildtype EL and a mutant version of EL (ELR330A) with profurin or PC5A in HEK293 cells. As expected, profurin prevented the cleavage of EL and increased EL lipase activities, whereas PC5A cleaved almost all fulllength EL and reduced activities markedly, but neither had any effect on the cleavage or enzymatic activity of ELR330A (Fig. 6D). Therefore, the cleavage of EL at the RNKR site by PCs specifically reduces its enzymatic activity. After normalized to the overall immunoreactive mass, we found that the specific activity of this uncleavable mutant was slightly higher than wild-type EL (data not shown).
LPL Is Also Processed by PCs at a Conserved RAKR Sequence-LPL is another member of the triglyceride lipase family closely related to EL. In the region between the N and C terminus of human LPL, we identified a RAKR sequence at residues 321-324, which is also conserved among all known LPL proteins in various species (supplemental Fig. S1). To test whether LPL is proteolytically cleaved, we expressed parental LPL, LPL with C-terminal myc tag (LPLmyc), and its mutants (R324A, R333A) in HEK293 cells. LPL protein was detected by the anti-LPL antibody, which only recognizes the C terminus of LPL. As shown in Fig. 7, in the medium wild-type LPL mainly appears as a 20-kDa band and a minor 55-kDa band (full-length). Additional myc tag at its C terminus results in a shift of both bands, further suggesting that full-length LPL is proteolytically cleaved. Mutant LPLR324A with the C-terminal myc tag showed almost no processing of LPL, giving mainly a 58-kDa band (full-length in addition with the myc tag). Another adjacent mutant, LPLR333A with the C-terminal myc tag, had no effect on LPL processing. When LPL was co-expressed with furin, most of the LPL protein was converted to 20 kDa; when LPL is co-expressed with profurin, all of the LPL appeared as a 55-kDa protein. Of note, unlike EL, the prevention of cleavage of LPL affects its secretion into the culture medium.
To prove that the cleavage of LPL is occurring at the RAKR sequence, we applied to LPL the same approach we used for EL. We purified the C-terminal fragment of LPL using the nickel-nitrilotriacetic acid column and sequenced it. The cleavage of LPL is after position 324, the second arginine in the RAKR sequence (Fig. 7B).
Finally, we tested the ability of various PCs to cleave LPL in a stable HEK293 cell line expressing LPLmyc. Pulse-chase experiments were performed under conditions similar to those used for EL. As illustrated in Fig. 7C, cells transfected with the empty vector (Control) exhibited two bands with an apparent molecular mass of 60 and 25 kDa corresponding to the full-length and C terminus of LPL. Transfection with vectors encoding each of the PCs revealed that furin and PACE4, and to a lesser extent PC5A, could process LPL, but not the other PCs tested. A third member of the triglyceride lipase family, HL, cannot be cleaved by PCs. 6

DISCUSSION
Our results indicate that the EL protein is proteolytically cleaved at a specific RNKR site during and/or after its secretion and that this cleav-  age is brought about by several members of the PC family. These findings are consistent with our early observation that a 40-kDa fragment of mouse EL was detected in mouse liver lysate (17). In contrast to many PC substrates, intracellular proteolytic cleavage is clearly not a prerequisite for EL activation. Rather, cleavage of EL by PCs results in a less active enzyme, representing a unusual "inactivation" function for PCs.
We found that EL is cleaved at the peptide bond following arginine 330 in the hinge region between two putative domains of EL. The only residues specifically required for efficient cleavage are arginine 327 and 330. The integrity of the recognition sequence RXXR is also important. This RXXR sequence is completely conserved in the human, mouse, and rat ELs. The specific location of this sequence is critical, because a similar RXXR sequence is also found at several other sites within the EL protein, but is not cleaved.
We demonstrated that EL is a substrate of PCs. However, EL cleavage is not mediated by SKI-1/S1P and NARC-1, which are the only PCs known to be involved in lipid metabolism and specifically affect the metabolism of apoB-containing lipoproteins. Furin, PC5A, and PACE4 (but not several other PCs) are able to cleave wild-type EL but not the P1 R330A site mutant. Furthermore, several specific inhibitors of PCs inhibit EL cleavage. We show that several PCs that cleave EL are coexpressed with EL in endothelial cells, consistent with previous observations that furin and PC5A are expressed ubiquitously by endothelial cells (40,41). Knockdown experiments further indicate that PCs are the endogenous proteases in endothelial cells that cleave EL. Because the expression pattern of PCs is different in a range of endothelial cells, it is conceivable that several PCs, including furin, PC5A, and PACE4, might spatially regulate EL cleavage in a tissue-specific manner under various physiological and pathogenic conditions.
Our data suggest that EL cleavage probably occurs both intracellularly, most likely in the trans Golgi network, and at the cell surface. It is noteworthy that only a small fraction of EL is cleaved intracellularly even after inhibition of protein transport with colchicine. Several lines of evidence support substantial extracellular processing of EL. When heparin was used to release cell surface-bound EL, the EL was mainly in the full-length form. Further incubation of the medium in the absence of cells led to no additional cleavage; however, there was additional processing when the same medium was post-incubated in the presence of cells. Treatment with the water-soluble peptide A containing the EL cleavage sequence dramatically blocked EL processing. Therefore, our data suggest that EL is cleaved by PCs at least in part on the cell surface. The role of furin, PC5A, and PACE4 in the processing of their substrates on the cell surface remains elusive. Furin is a type-one transmembrane protein, and, until recently, its only known role on the plasma membrane was activation of opportunistic pathogens (42). However, furin was shown to cleave and inactivate the Tat protein of HIV-1, a transcriptional activator of the provirus (43) at the cell surface. In addition, membrane type-1 matrix metalloproteinase was recently shown to be cleaved and activated by furin at the cell surface (44). PACE4 and PC5A are secretory proteins and have been shown to localize in the extracellular matrix, presumably through binding to heparan sulfate proteoglycans (45). PC5A was shown to cleave neural adhesion molecule L1 on the surface of endothelial cells (46). Therefore, we speculate that the extracellular matrix may provide a microenvironment for the interaction of PCs and EL.
Based on a presumed three-dimensional structure similar to pancreatic lipase, the only member of the triglyceride lipase for which a structure has been determined (47), full-length EL protein is probably composed of an ␣ helical N-terminal domain and a ␤ sheet C-terminal domain joined by a hinge region. Cleavage of EL at the hinge region would be predicted to generate two smaller fragments with intact secondary structure. We do not know whether the cleavage products of EL themselves have any distinct biological functions. Notch-1, for example, is cleaved by PCs, and the cleavage products mediate distinct signaling pathways (48). The N-terminal fragment of EL co-purifies with the C-terminal fragment of EL after cleavage (data not shown), indicating the existence of intermolecular interactions of the two domains of EL, and raising the question whether cleavage of EL plays a role in this interaction. Despite extensive cleavage of EL under some experimental conditions, some EL activity remained, suggesting that complexes other than full-length EL may be active. Moreover, the EL protein is incompletely processed in various endothelial cells as shown in Fig. 1A. Thus, intermolecular interaction of EL molecules combined with limited cleavage of EL may generate several intermediates that may have different lipolytic properties. We did not observe a higher lipase activity of the uncleavable ELR330A mutant. It is possible that this alanine replacement might have global effects on protein structure thereby affecting the overall activity in addition to affecting EL cleavage. We are currently testing other amino acid substitutions at the position.
Post-transcriptional regulation of LPL activity has been observed for FIGURE 7. Cleavage of human LPL by PCs. A, processing of LPL was analyzed in HEK293 cells transiently transfected with the indicated plasmids. Following expression, aliquots of cell medium were analyzed by immunoblotting with anti-LPL antibody. F and C denote the full-length LPL and C-terminal fragment, respectively. B, NH 2terminal sequence of C-terminal fragment of LPL. C, HEK293 cells stably expressing LPLmyc were transiently transfected with the indicated constructs. The cells were pulse-labeled with [ 35 S]Met/Cys for 2 h and then chased for 4 h. Culture media were immunoprecipitated with the anti-myc antibody and analyzed by SDS-PAGE on 8% Tricine gels. Experiments were repeated three times.
more than 3 decades (49), but the mechanisms of this regulation are uncertain. In the region between the N and C terminus of human LPL, we identified a RAKR sequence at residues 320 -324, which is also conserved among all known LPL proteins in various species (supplemental Fig. S1). Here, we provide one potential mechanism of post-transcriptional regulation, showing that LPL is also proteolytically cleaved by PCs. Early observations by several groups showed that LPL protein was degraded into two major fragments, but this was felt to be an ex vivo artifact (50). In this study, we show that LPL cleavage is not because of nonspecific degradation, but rather a specific proteolytic event mediated by PCs. The processing of LPL has some common features compared with EL cleavage. Both of them occur at a typical conserved PC recognition sequence RXKR within the hinge region of the molecule, and the cleavage is sensitive to PC inhibitors. However, they are also quite different in several ways. First, individual PCs have differential abilities to cleave EL or LPL. For EL, the efficiency of cleavage is PACE4 Ͼ PC5A Ͼ furin, whereas for LPL, the efficiency of cleavage is furin Ͼ PACE4 Ͼ PC5A. Second, the cleavage of LPL by PCs is overall less efficient than of EL. Third, unlike EL, the prevention of cleavage of LPL reduces its secretion. Therefore, whereas both EL and LPL are cleaved by members of the PC family at a conserved RXKR site, the physiologic implications of the cleavage may differ.
In summary, these studies demonstrate that PCs are implicated in the cleavage of EL at a conserved Arg-Asp-Lys-Arg 330 sequence and underscore the potential importance of this processing in regulating EL function. Furthermore, PCs are able to cleave LPL as well. Therefore, control of lipase activity by PCs may be a general physiological control mechanism. In this context, modulation of PC-mediated processing of EL and LPL may provide a novel potential therapeutic strategy for regulating lipase activity and influencing different classes of lipid metabolism. This work establishes a novel link between the PC family and lipid metabolism, advances our understanding of the processing of lipases, and may facilitate the development of therapeutic approaches to treat dyslipidemia and atherosclerosis.