The Proprotein Convertase Subtilisin/Kexin Type 9 (PCSK9) Active Site and Cleavage Sequence Differentially Regulate Protein Secretion from Proteolysis*

Background: PCSK9 inhibition shows promise in the treatment of hypercholesterolemia and prevention of atherosclerotic heart disease. Results: PCSK9 proteolysis and secretion have different structural requirements near the protein active site. Conclusion: PCSK9 proteolysis and secretion occur in two separate steps. Significance: The PCSK9 active site can serve as an allosteric modulator of protein secretion, suggesting a new strategy for inhibition. Biologic-based strategies to inhibit proprotein convertase subtilisin/kexin type 9 (PCSK9) show promise as anti-hypercholesterolemic and, therefore, anti-atherosclerotic therapies. Despite substantial effort, no small molecule strategy to inhibit PCSK9 has demonstrated feasibility. In this study we interrogated the chemistry of the PCSK9 active site and its adjacent residues to identify a foothold with which to drug the PCSK9 processing pathway and ultimately disrupt the interaction with the LDL receptor. Here, we develop a system in which we amplify the readout of PCSK9 proteolysis with a highly specific substrate in cells, showing that the PCSK9 catalytic domain is capable of proteolysis in trans. We use this system to show that the substrate specificity for PCSK9 proteolysis is distinct from the specificity for PCSK9 secretion, demonstrating that PCSK9 processing occurs in two separate sequential steps: that of proteolysis followed by secretion. We show that specific residues in the protease recognition sequence can differentially modulate the effects on proteolysis and secretion. Additionally, we demonstrate that the clinically described, dominant negative Q152H mutation restricts proteolysis and secretion independently. Our results suggest that the PCSK9 active site and its adjacent residues serve as an allosteric modulator of protein secretion independent of its role in proteolysis, revealing a new strategy for intracellular PCSK9 inhibition.

and is thought to act as a self-inhibitor, preventing any further proteolytic activity (22). In vitro methods of measuring PCSK9 proteolytic activity struggle from the low intrinsic activity of the protease, requiring high amounts of purified protein and producing low signal-to-noise ratios (23). In addition, the prodomain cannot be removed from the catalytic domain short of denaturing the entire protein, and no group has isolated the proteolytically active proPCSK9 form (24). Last, even if an inhibitor for the PCSK9 protease was found, it would likely need to outcompete an intramolecular reaction to be clinically useful.
Chemical challenges are not atypical for protease inhibitors. Direct renin inhibitors serve as an example that these challenges can be overcome; after decades of focused effort, a structure and modeling approach produced aliskiren, now a clinically approved compound (25,26). Our understanding of PCSK9 processing and maturation suggests that the difficulty in targeting PCSK9 proteolysis is primarily one of active site accessibility. We thus undertook the present study to thoroughly investigate the biochemistry of the PCSK9 active site. Our primary goal was to identify any unusual aspects that would allow us a foothold into a possible small molecule approach to perturb PCSK9 proteolysis or secretion.

EXPERIMENTAL PROCEDURES
Materials and Reagents-Oligonucleotide primers were custom synthesized by Elim Biopharmaceuticals (Hayward, CA). Restriction enzymes and polymerases were purchased from New England Biolabs (Ipswich, MA). Mouse monoclonal antibodies to FLAG (clone M2) and V5 (clone V5-10) and anti-FLAG M2 magnetic beads were obtained from Sigma. Mouse monoclonal antibody to penta-His and Ni-NTA-agarose were obtained from Qiagen (Valencia, CA). Mouse monoclonal antibody to ␤-actin (8H10D10), rabbit monoclonal antibody to GFP (D5.1 XP), which also detects cyan and yellow fluorescent protein, goat anti-rabbit IgG HRP-conjugate antibody, and horse anti-mouse IgG HRP-conjugate antibody were all obtained from Cell Signaling (Danvers, MA). Rabbit polyclonal antibody to transferrin was obtained from Enzo Life Sciences (Farmingdale, NY). Goat anti-mouse IgG IRDye 800CW conjugate and goat anti-rabbit IgG IR Dye 800CW conjugate were obtained from LI-COR Biotechnology (Lincoln, NE).
Plasmid Construction-PCSK9 expression vectors were created by PCR amplification of the appropriate PCSK9 domains from pCMV-PCSK9-FLAG (generously provided by Prof. J. Horton, UT Southwestern) with custom oligonucleotides and subsequent cloning into the BglII and NotI restriction sites of pIRES2-AcGFP1 and pIRES2-DsRed2 (Clontech, Mountain View, CA). V5, HA, and His 6 tags, VFAQSIP insertions, and single site mutations were added by site-directed mutagenesis using partially overlapping primers to generate the constructs as shown in the schematics (27). Truncations of the prodomain C terminus, prodomain, and removal of the vector sequence encoding the fluorescent proteins were also performed by sitedirected mutagenesis using partially overlapping primers. Expression vectors encoding FRET substrates were created by Gibson assembly (28) after PCR amplification of CyPET and YPET from pB33-newCyPET-LVPRGS(Sub)-YPET (gener-ously provided by Dr. T. Shropshire and Prof. P. Daugherty, UC Santa Barbara) and using the PCSK9 expression vectors as a backbone. All plasmids were extensively sequenced to confirm their identities.
Mammalian Cell Culture and Transfections-HEK293T cells (ATCC, Manassas, VA) were cultured in DMEM with 10% FBS. Cells were plated in 6-well plates coated with poly-L-lysine (Sigma) 1 day before transfection to obtain ϳ50 -75% confluency at the time of the experiment. Cells were transfected with Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions using 0.5 g of DNA per construct. The medium was changed 6 -8 h after transfections; for samples in which the conditioned media was evaluated the medium was changed to DMEM with 1ϫ insulin-transferrin-selenium (Invitrogen) rather than DMEM with 10% FBS. After 48 -72 h post-transfections, the cells and conditioned media (if appropriate) were harvested for Western blot analyses. The conditioned media was centrifuged for 5 min at 1000 ϫ g, and the resulting supernatant was concentrated ϳ10ϫ using Amicon Ultra 10 kDa protein concentrators (EMD Millipore, Billerica, MA). The cells were washed twice with ice-cold PBS, lysed in lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.1% Nonidet P-40, 0.5% Triton X-100) for 10 min, and then clarified at 21,000 ϫ g for 15 min. Protein concentration was analyzed by Bradford assay (Bio-Rad). Gel samples were prepared with a SDS-based loading buffer under reducing conditions and heated at 98°C for 5 min. All transfection experiments were repeated at least three times to ensure reproducibility.
Ni-NTA Pulldown Assays-Conditioned media from transfections was adjusted to pH 8.0 with Tris-HCl to a final concentration of 50 mM, and imidazole was added to a final concentration of 10 mM. The medium was centrifuged at 1000 ϫ g, and the supernatant was added to Ni-NTA-agarose beads (Qiagen) previously equilibrated in binding buffer (20 mM Tris-HCl, pH 8.0, 250 mM NaCl, and 10 mM imidazole). Medium was batchbound for 4 h with rotation at 4°C. The beads were washed 3 times with 10ϫ volumes of wash buffer (20 mM Tris-HCl, pH 8.0, 250 mM NaCl, 20 mM imidazole). Elution was performed with 5ϫ volumes of elution buffer (20 mM Tris-HCl, pH 8.0, 250 mM NaCl, 250 mM imidazole) with the supernatant saved and the process repeated. The fractions were pooled and concentrated ϳ10ϫ in Amicon Ultra 10-kDa concentrators. The concentrated samples were then prepared for gel analysis using a SDS-based loading buffer under reducing conditions and heated at 98°C for 5 min.
Anti-FLAG Immunoprecipitations-Transfected cells were lysed on ice for 10 min in lysis buffer containing a protease inhibitor mixture (Roche Applied Science) and clarified at 21,000 ϫ g for 15 min. The resultant supernatant was saved and subjected to Bradford assay. Equivalent amounts of lysate (according to total protein) were added to anti-FLAG M2 magnetic beads (Sigma) previously equilibrated in lysis buffer and bound at 4°C with rotation overnight. The beads were then washed 3 times with 10ϫ volumes of lysis buffer and once with 10ϫ volumes of TBS. Elution was performed by incubation with 10ϫ volumes of 0.1 M glycine, pH 3.0, at room temperature with rotation for 5 min. The process was repeated, and the elution fractions were pooled and concentrated ϳ10ϫ in Ami-can Ultra 10-kDa concentrators. The concentrated samples were then prepared for gel analysis using a SDS-based loading buffer under reducing conditions and heated at 98°C for 5 min.
Western Blot Analyses-Gel samples were loaded onto 4 -12% Bis-Tris SDS NuPAGE precast gels (Invitrogen) for electrophoresis. The size-separated proteins were then transferred to nitrocellulose. The blots were blocked in 5% nonfat dry milk in Tris-buffered saline (20 mM Tris, pH 7.4, 150 mM NaCl) containing 0.1% Tween (TBS-T) for 1 h. The appropriate primary antibodies were hybridized to the blots overnight at 4°C. The blots were washed 3 times for 10 min each with TBS-T, and the appropriate secondary antibodies were hybridized in 5% nonfat dry milk in TBS-T for 1 h at room temperature. The blots were again washed 3 times for 10 min each with TBS-T before either developing with chemiluminescent substrate and imaging with CL-Xposure film (Thermo Fisher Scientific, Rockford, IL) or directly imaging with the ODYSSEY infrared imaging system (LI-COR Biotechnology, Lincoln, NE). Quantification was performed with the LI-COR Image Studio or Image Studio Lite software.

The PCSK9 Catalytic Domain Can Perform Intermolecular
Proteolysis-We first endeavored to develop a robust method to evaluate PCSK9 proteolysis. To maximize both utility and specificity in our system, we focused on proteolysis in the native, cellular environment with a highly specific substrate. Although conversion of wild-type proPCSK9 to mature PCSK9 occurs in an intramolecular fashion (Fig. 1A) (17), McNutt et al. (29) demonstrated that the prodomain and catalytic domain can be assembled in trans (i.e. as two separate polypeptides), which effectively bypasses the need for proteolysis and does not require intrinsic proteolytic activity (Fig. 1, C and D). Furthermore, a secretion-defective PCSK9 mutant can transfer its prodomain to a prodomain-deficient (and also secretion-defective) PCSK9, rescuing the secretion of the latter PCSK9 species (30). As such, we hypothesized that the in trans assembly of the catalytic domain and a specific substrate and subsequently intermolecular proteolysis could also occur.
To evaluate this hypothesis, we designed a series of constructs to evaluate for the specific cleavage of a "substrate," a full-length but proteolytically inactive S386A mutant PCSK9 termed substrate PCSK9 , by a "protease," an orthogonally tagged PCSK9 construct containing the catalytic domain ( Fig. 2A). Given that the S386A mutant is devoid of proteolytic activity, any cleavage of the substrate in our system must, therefore, come from an intermolecular cleavage event. Similarly, we used the S386A mutant in each protease construct throughout our investigation as a negative control. Co-transfection of a fulllength PCSK9 protease, termed protease PCSK9 , with substrate PCSK9 into HEK 293T cells revealed a trace amount of cleaved substrate PCSK9 in the lysate (Fig. 2B, lane 4, open arrows). By contrast, co-transfection of the inactive protease PCSK9-S386A with substrate PCSK9 produced no cleavage product (Fig. 2B, lane 5). Consistent with prior studies concluding that the vast majority of PCSK9 proteolysis is intramolecular (16,17), the scale of intermolecular cleavage seen with protease PCSK9 was far below (only ϳ1%) that of intramolecular cleavage of wild-type PCSK9 (Fig. 2, B, lane 2 compared with lane 4, and D, left, blue). However, the co-transfection of a prodomain-deficient protease, termed protease PCSK9⌬Pro , markedly increased cleavage of substrate PCSK9 (Fig. 2B, lane 6) when compared with protease PCSK9 (Fig. 2B, lane 4) to just Ͼ15% of our wild-type benchmark (Fig. 2D, right, blue). In addition, the absence of cleaved substrate PCSK9 with the cotransfection of inactive protease PCSK9⌬Pro-S386A (Fig. 2B, lane 7) shows that the proteolysis of substrate PCSK9 is highly specific for the active protease PCSK9⌬Pro . Furthermore, because protease PCSK9⌬Pro is significantly more active than protease PCSK9 , these data show that the prodomain is a cis-acting inhibitor of intermolecular proteolysis, consistent with the known PCSK9 structure (22,24,31). We were particularly encouraged by our ability to amplify the detection of intermolecular PCSK9 proteolysis, especially given that the PCSK9 protease is thought to only undergo a single turnover event.
We then sought to evaluate whether the processed intermolecular cleavage products were properly secreted like wild-type PCSK9. To do so we evaluated the conditioned media of our co-transfections by Western blot (Fig. 2C, left). Secretion of protease PCSK9⌬Pro occurred only in the presence of sub-FIGURE 1. Schematics of PCSK9 processing. The known processing pathways of PCSK9 are shown. The signal sequence is shown in dark gray, the prodomain in red, the catalytic domain in light gray, and the CHR domain in blue. A, processing of wild-type PCSK9. The nascent PCSK9 polypeptide is directed to the ER by a signal sequence, which is subsequently removed by a signal peptidase. The proPCSK9 form then undergoes auto-proteolysis, and the mature, cleaved form is transported to the trans-Golgi network (TGN) via a COPII complex involving Sec24a. Secretion from the trans-Golgi network is facilitated by sortilin. The PCSK9 binds the LDL-R on the cell surface, chaperoning the receptor for lysosomal degradation upon internalization. B, processing of proteolytically inactive PCSK9. PCSK9 with an active site mutation (S386A) cannot perform intramolecular proteolysis, is retained in the ER, and thus cannot interact with the LDL-R on the cell surface. C, processing of in trans PCSK9. In an experimental system the prodomain and prodomain-deficient catalytic and CHR domains are expressed as separate polypeptides, each directed independently to the ER. These domains assemble in trans, bypassing the proteolytic step to form a mature PCSK9 that follows the remainder of the prototypic processing pathway and retains activity against the LDL-R. D, processing of in trans proteolytically inactive PCSK9. Because the proteolytic step is bypassed, a proteolytically inactive PCSK9 can be assembled in trans and secreted, retaining activity against the LDL-R. strate PCSK9 and was accompanied by the prodomain from substrate PCSK9 (Fig. 2C, left, lane 6) to levels of ϳ10% that of wildtype PCSK9 (Fig. 2D, right, purple). By contrast, the inactive protease PCSK9⌬Pro-S386A was not secreted in the presence of substrate PCSK9 (Fig. 2C, left, lane 7). A trace amount (Ͻ1% of wild type) of prodomain from substrate PCSK9 was also secreted with protease PCSK9 (Fig. 2C, left, lane 4, and Fig. 2D, left, purple). The C-terminal cleavage product of substrate PCSK9 , which contains the catalytic and cysteine-histidine-rich (CHR) domains, was not secreted under any condition (Fig. 2C, left). FIGURE 2. The PCSK9 catalytic domain can perform proteolysis in trans. A, schematics of the PCSK9 expression constructs. All PCSK9 expression constructs contain the N-terminal signal sequence and the appropriate PCSK9 domains as labeled. Domains within constructs are colored to correspond to the C-terminal recognition tag, orthogonal between protease and substrate constructs: magenta, V5 (N-terminal, substrate); green, FLAG (C-terminal, substrate); orange, His (C-terminal, protease). The P5 through P4Ј sequence spanning the transition from the prodomain to catalytic domain is shown in wild-type PCSK9, substrate PCSK9 , and protease PCSK9 , with the cleavage site indicated by the arrowhead. The S386A mutation in substrate PCSK9 is shown. B, Western blots (WB) of cell lysates from co-transfections of constructs or empty vectors into 293T cells as indicated. Wild-type PCSK9 (lane 2) serves as the positive control. Uncleaved substrate is noted by the black arrow. Cleavage products of substrate PCSK9 are noted by the open arrows. C, Western blots of conditioned media (left) and Ni-NTA pulldown assays (right) from the same experiment as in B. D, mean levels of processed substrate PCSK9 from either proteolysis (blue) or secretion (purple), as normalized to intramolecular proteolysis or secretion of wild-type PCSK9, respectively, from three independent experiments. Error bars represent S.E. E, model for processing of intermolecularly cleaved PCSK9. Protease PCSK9⌬Pro recognizes the prodomain of substrate PCSK9 in the ER in a manner akin to in trans assembly, performs intermolecular proteolysis, and then retains the "donor" prodomain from substrate PCSK9 . Protease PCSK9⌬Pro , with its donor prodomain, is capable of secretion following the prototypic PCSK9 secretion pathway, whereas the C-terminal cleavage product of substrate PCSK9 , without its prodomain, is retained in the ER. TGN, trans-Golgi network. OCTOBER 17, 2014 • VOLUME 289 • NUMBER 42

JOURNAL OF BIOLOGICAL CHEMISTRY 29033
To confirm the association of the V5-tagged prodomain with the His-tagged protease PCSK9⌬Pro , we performed a pulldown experiment with Ni-NTA-agarose beads on the conditioned media to isolate the His-tagged protease PCSK9 and protease PCSK9⌬Pro . This showed that the prodomain from substrate PCSK9 co-precipitated with both proteases (Fig. 2C, right, lanes 4 and 6). Furthermore, there was proportionally much greater V5-tagged prodomain with protease PCSK9⌬Pro than with protease PCSK9 (Fig. 2C, right, compare lane 6 to lane 4). This likely results from a preference of protease PCSK9 to utilize its own prodomain for secretion, which protease PCSK9⌬Pro lacks. Overall, these data suggest that the prodomain is retained by the protease PCSK9⌬Pro after intermolecular cleavage and are consistent with a model whereby the protease PCSK9⌬Pro assembles with the prodomain of substrate PCSK9 , performs intermolecular cleavage, and then non-covalently retains the prodomain in a manner akin to intramolecularly processed PCSK9 (Fig. 2E). The prodomain from substrate PCSK9 then serves as the chaperone guiding the mature protease PCSK9⌬Pro out of the ER through the Golgi and ultimately to the extracellular space. Despite undergoing cleavage, the catalytic and CHR domains of substrate PCSK9 have been stripped of their own prodomain and thus are not secreted.
The Scope of Intermolecular PCSK9 Proteolysis Is Narrow-Having identified an optimized, highly specific substrate for protease PCSK9⌬Pro , we next addressed the scope of the intermolecular cleavage reaction, as knowledge of the substrate specificity is an important first step toward inhibitor design. We asked whether the known cleavage sequence could be recognized and cut if placed elsewhere within the prodomain, anticipating that such engineering might generate a more active protease. We posed this hypothesis based on the typical processing pathway of proprotein convertases; most members of this family of proteases undergo two cleavage events to mature (32). The first event, common to all convertases including PCSK9, cuts the prodomain from the catalytic domain, allowing the prodomain to serve both as a chaperone for delivery of the convertase to its proper cellular compartment and a self-inhibitor of proteolytic activity (33). The second event, which PCSK9 does not perform, cleaves the prodomain at the flexible L loop, causing the prodomain to dissociate from the active site, releasing the self-inhibition on the proteolytic activity of the convertase (34). We, therefore, interrogated the small L loop of PCSK9, located after Lys-125, as well as an additional solventexposed loop of the prodomain at Pro-138 to evaluate whether an additional proteolytic event could occur. We inserted the seven-amino acid PCSK9 cleavage sequence VFAQSIP into our initial substrate construct between Lys-125 and Met-126 to create substrate PCSK9-K125iSELF and in place of Pro-138 and His-139 to create substrate PCSK9-P138iSELF (Fig. 3A). Co-transfection of substrate PCSK9-K125iSELF with protease PCSK9⌬Pro revealed only a trace amount of cleaved product (Fig. 3B, left, open arrows, lane 3) with cleavage clearly less efficient than with unmodified substrate PCSK9 (Fig. 3B, left, lane 2). Because of the low signal for cleavage of substrate PCSK9-K125iSELF in the lysates, we confirmed the presence of the N-terminal cleavage product in the conditioned media (Fig.  3B, right, open arrow, lane 3), indicating that it also permit-ted secretion of the protease. Importantly, this cleavage product migrated at ϳ18 kDa and just above the prodomain of the unmodified substrate PCSK9 , consistent with a cleavage site at the native Gln-152 (anticipated molecular mass of ϳ17 kDa) rather than within the engineered sequence (anticipated molecular mass of ϳ13 kDa) (Fig. 3B, compare  lanes 2 and 3). We did not detect a band at a lower molecular weight corresponding to the smaller cleavage product, indicating that cleavage within the engineered insertion sequence did not occur. By contrast, neither appreciable cleavage nor secretion of the protease occurred for substrate PCSK9-P138iSELF (Fig. 3B, lane 5) or for any conditions with inactive protease PCSK9⌬Pro-S386A (Fig. 3B, lanes 4 and 6), indicating that the insertion in place of Pro-138/His-139 also prevented proteolysis of the native sequence. Because our insertions were purposefully non-conservative, we acknowledge that such mutagenesis may have disrupted the proper folding of the substrate PCSK9 , preventing it from forming a cleavable complex with protease PCSK9⌬Pro . However, because the complex of protease PCSK9⌬Pro and the prodomain from substrate PCSK9-K125iSELF permits secretion, we know this is unlikely for this particular insertion. Although we cannot rule out the possibility that a VFAQSIP site placed elsewhere within the prodomain would be cleaved, our data show that the catalytic domain cannot identify and cleave its known recognition site when that site is placed within the prodomain L loop.
To further address the scope of this reaction, we next asked whether the PCSK9 catalytic domain could identify and cleave a minimal recognition sequence outside of the full PCSK9 protein. We generated minimal substrates consisting of the anticipated PCSK9 cleavage sequences VFAQSIPW and SSVFAQ-SIPWNL flanked by a fluorescent protein pair (CyPET-YPET) (35). Co-expression of either substrate with protease PCSK9⌬Pro revealed no appreciable cleavage product (data not shown), suggesting that the catalytic domain requires at least another region of substrate PCSK9 beyond the cleavage site itself to properly recognize and cleave its known sequence. When taken together with the results of Fig. 3, these data support the concept that the PCSK9 catalytic domain has a narrow specificity window for its known substrate.
The Prodomain Inhibits Intermolecular Proteolysis in trans, Dependent on Its C-terminal Tail-Having established both the capability and the relatively narrow scope of the PCSK9 catalytic domain for intermolecular cleavage, we next sought to identify a positive control for an inhibitor of the intermolecular proteolysis reaction. Based on the occupancy of the active site by the prodomain C terminus (22,24,31) as well as our own results of the prodomain inhibiting intermolecular proteolysis in cis, we hypothesized that exogenous prodomain would serve as a model inhibitor and effectively block intermolecular proteolysis in trans. To test this we performed co-transfections in the intermolecular cleavage system (protease PCSK9⌬Pro acting upon substrate PCSK9 ) with an additional wild-type prodomain construct termed prodomain WT (Fig. 4A). Co-transfection of protease PCSK9⌬Pro with substrate PCSK9 and prodomain WT revealed progressively less cleaved substrate with increasing expression levels of the prodomain WT (Fig. 4B, lanes 4 -7, and the graph below).
We next wished to carry out a structure-activity relationship study of the determinants of inhibition by the C-terminal tail of the prodomain. We created a series of mutant prodomain constructs harboring progressive amino acid deletions from the C terminus. We extended our truncations through the six amino acids spanning the entire binding groove of the active site, from the terminal Gln-152 to Ser-147, to create constructs prodomain ⌬152 through prodomain ⌬147 (Fig. 4A). As anticipated, we found that at equivalent expression levels, the wild-type prodomain was the most effective inhibitor of intermolecular proteolysis (Fig. 4C, lane 4). Furthermore, the loss of even one amino acid from the C terminus markedly abrogated the inhibitory ability of the exogenous prodomain, as evidenced by the presence of more cleavage product (Fig. 4C, compare lane 5 to lane 4, and the graph below). Importantly, further truncations did not result in progressively less effective inhibitors (Fig. 4C, lanes 6-10, and the graph below). Note that the V5-tagged N-terminal cleavage product of substrate PCSK9-K125iSELF migrates at a slightly greater molecular weight than that of substrate PCSK9 , indicating that the cleavage site is at the native Gln-152 rather than the engineered VFAQSIP sequence between Lys-125 and Met-126. OCTOBER 17, 2014 • VOLUME 289 • NUMBER 42

JOURNAL OF BIOLOGICAL CHEMISTRY 29035
The Prodomain C Terminus Regulates Protein Secretion but Is Not Required for Catalytic Domain Binding-One potential explanation for these results is that the entire prodomain is required to bind to the catalytic domain and inhibit the intermolecular cleavage reaction. To evaluate this, we co-transfected our truncated prodomain ⌬ constructs with protease PCSK9⌬Pro to evaluate for the proper association of these domains in cells. Anti-FLAG immunoprecipitations of the lysates revealed that in all cases the V5-tagged truncated prodomains co-immunoprecipitated with the FLAG-tagged protease PCSK9⌬Pro (Fig. 4D, middle, lanes 3-9). These results indicate that the truncated prodomains retain the ability to bind to the catalytic domain, although whether these complexes represent properly folded, mature PCSK9 or an aggregate of misfolded protein is less clear. Remarkably, however, when analyzing the media from these co-transfections, we found no secreted protease PCSK9⌬Pro -prodomain complexes, save for prodomain WT (Fig. 4D, right, lanes 3-9). To further probe the requirement of the C terminus in permitting secretion, we engineered a prodomain construct with an additional alanine inserted after Gln-152, termed prodomain ϩAla . Co-transfection of this construct with protease PCSK9⌬Pro resulted in both intact prodomain:catalytic domain binding (Fig. 4D, middle, lane 10) as well as preserved secretion (Fig. 4D, right, lane 10). Importantly, all of these results held true for both the active protease PCSK9⌬Pro and the inactive protease PCSK9⌬Pro-S386A (Fig. 4E).
The Residues of the Prodomain C Terminus Regulate Proteolysis and Secretion Independently-We were struck by the exquisite sequence requirements of the prodomain ⌬ constructs for secretion. To further evaluate these requirements, we performed an alanine scan of the prodomain C-terminal tail to investigate the effect of mutation of each residue on secretion. We created a series of prodomain constructs with alanine mutations from Ser-148 through Gln-152 along with the corresponding double, triple, and quadruple mutants. We then cotransfected protease PCSK9⌬Pro with the prodomain mutants and analyzed the conditioned media (Fig. 5). We found that secretion was only mildly affected by alanine substitutions at Ser-148, Phe-150, and Gln-152 (Fig. 5B, lanes 5, 7, and 8 and Fig.  5C). By contrast, secretion was markedly abrogated by an alanine substitution at Val-149 (Fig. 5B, lane 6, and Fig. 5C). Similarly, secretion was abrogated for all double, triple, and quadruple mutants with the exception of the double mutants containing S148A, with the S148A/V149A mutant the most significantly affected among that subset (Fig. 5, B, lanes 9 -19, and C). When the results from Figs. 4 and 5 are taken together, the findings suggest that the C terminus of the prodomain is critical for its proper function as a chaperone for PCSK9 secretion, with a strict requirement that all amino acids be present and a more lax requirement for the identity of the individual residues. The strict requirement for Val-149 is not surprising given that the isopropyl side chain points directly into the binding groove of the catalytic domain in the PCSK9 structure.  However, the tolerability at Phe-150 is somewhat surprising given that the phenyl ring of this side chain makes astacking interaction with the indole of Trp-72 and is important for proper intramolecular cleavage of PCSK9 (16,22).
We were intrigued by the differential structural requirements of the proteolytic site vis à vis the requirements for secretion in trans in a system that effectively bypasses the need for proteolysis. We next performed a serial alanine scan of the P5 through P4Ј residues of substrate PCSK9 to determine the tolerance for these mutations in our intermolecular cleavage system. Co-transfections of protease PCSK9⌬Pro with the substrate PCSK9 alanine scan mutants revealed that alanine substitutions at most positions were well tolerated, with the exception of P4 (Val-149) and P3 (Phe-150), where no cleavage occurred (Fig. 6,  B and C). The intolerance for intermolecular cleavage for V149A was not surprising given prior work showing that this mutation abolishes intramolecular proteolysis (16) as well as our own results indicating that this mutation did not permit secretion. However, the lack of intermolecular cleavage of our F150A mutant was unexpected, as this mutant is known to permit a low level of intramolecular proteolysis (16). In addition, our results showed that the F150A modestly permitted secretion in the in trans system, highlighting the disconnect between secretion and intermolecular proteolysis. Overall, our data show that the F150A mutant is incapable of intermolecular proteolysis, but if such proteolysis is bypassed then the F150A mutant is capable of secretion. This also suggests that, at least for our intermolecular system, the process of secretion of mature PCSK9 from the ER does not occur in a one-step concerted mechanism with proteolysis but instead involves two separate events mediated by different complexes (i.e. proteolysis, followed by secretion).
Having identified differential prodomain requirements for secretion and intermolecular proteolysis, we desired to return to a physiologic system that would mimic a natural PCSK9 variant and look for a similar disconnect between secretion and intramolecular proteolysis. We thus focused on Gln-152 given that the Q152H mutation has been documented in a patient cohort as a dominant-negative, loss-of-function phenotype resulting in low serum LDL cholesterol and a reduced incidence of atherosclerotic heart disease (36,37). We began with a vertical mutagenesis strategy at Gln-152 to evaluate the requirements on secretion, as others have demonstrated the requirement of Gln-152 on intramolecular proteolysis previously (37). We generated a series of Q152X mutations in our prodomain constructs and performed co-transfections of prodomain Q152X with protease PCSK9⌬Pro . Comparing the mutations that permitted secretion in our strategy (Fig. 7) to prior reports of mutations permitting proteolysis (37), we identified four phenotypes of residues according to the independent effects on proteolysis and secretion. We then installed representative mutations in both full-length PCSK9 and our substrate PCSK9 constructs to confirm the effects of these mutations on both intra-and intermolecular proteolysis and test alongside secretion. Fig. 8B shows the results of intramolecular proteolysis of the fulllength PCSK9 Q152X (left) and intermolecular proteolysis from protease PCSK9⌬Pro on substrate PCSK9-Q152X (right), whereas Fig.  8C shows the results of secretion from the in trans assembly of protease PCSK9⌬Pro and prodomain Q152X . As expected, the wildtype residue permitted both intra-and intermolecular proteolysis (Fig. 8B, lane 2) as well as secretion when expressed in trans (Fig. 8C, lane 2). The clinically relevant Q152H mutation permitted some proteolysis in both systems, albeit markedly less efficiently than the wild-type (Fig. 8, B, lane 3, and D, light and  dark blue), and was also markedly limited in secretion (Fig. 8, C,  lane 3, and D, purple). The block on secretion, however, was not absolute, as secretion of Q152H (as well as for Q152I) was mildly improved when the mutant prodomain was co-ex- pressed with the inactive protease PCSK9⌬Pro-S386A as compared with the active protease PCSK9⌬Pro (Fig. 8C, compare lanes 3 and  4 with lanes 7 and 8). The Q152I mutation completely abrogated proteolysis in both intra-and intermolecular systems (Fig. 8, B, lane 4, and D) but had only a limited impact on secretion (Fig. 8, C, lane 4, and D, purple). The Q152R mutation completely abolished both proteolysis and secretion (Fig. 8,  B and C, lanes 5, and D). The phenotypes of these mutations are summarized by the schematic in Fig. 8E. Although each of the mutations was less permissive for either proteolysis or secretion as compared with the wild type, the illustrative finding is that certain mutations have differential effects on these two processes. These findings lend further support to our conclusion that proteolysis and secretion proceed in two independent steps mediated by different structural requirements.

DISCUSSION
In this study we began with a desire to learn more about the chemistry of the PCSK9 active site in an attempt to gain a foothold onto a highly validated therapeutic target considered too challenging to drug by binding a small molecule into the active site. We found that the PCSK9 catalytic domain has the capac-ity to perform intermolecular proteolysis on a highly specific substrate, an orthogonally tagged but catalytically inactive proPCSK9 molecule. Furthermore, we found that such cleavage with our protease PCSK9⌬Pro construct was more efficient when compared with the cleavage performed by wild-type PCSK9, allowing our assay to effectively amplify the readout of intermo-  secretion). D, relative levels of intramolecular proteolysis (light blue), intermolecular proteolysis (dark blue), or in trans secretion (purple) of the Q152X mutants in the corresponding experimental system, normalized to the wild-type Gln-152. Data are from three independent experiments, and error bars represent S.E. E, schematic illustrating the phenotypes of the Q152X mutants with regard to permission of proteolysis (either intra-or intermolecular) or isolated effect on secretion (in the in trans system, with proteolysis bypassed). The wild-type glutamine permits both proteolysis and secretion, the histidine mutant permits some proteolysis but mainly precludes secretion, the isoleucine mutant precludes proteolysis but permits secretion, and the arginine mutant precludes both proteolysis and secretion. lecular PCSK9 proteolysis. We suspect that further development of this assay may be particularly useful in future screening strategies aimed at identifying small molecules or genetic targets that inhibit PCSK9 proteolysis as an ultimate means of inhibiting PCSK9's hypercholesterolemic effect. In particular, because this assay focuses on intermolecular proteolysis, it would increase the sensitivity for an initial weak inhibitor that might not be capable of out-competing an intramolecular event.
To our knowledge we present the first biochemical data that the prodomain itself acts as an inhibitor of proteolysis in trans. This is not surprising given that the PCSK9 structure suggests that the prodomain sterically inhibits access to the active site. The marked decrease in inhibition of intermolecular cleavage by a prodomain with a single C-terminal deletion (⌬Q152) further supports the steric hypothesis. However, additional truncations neither make the prodomain a less effective inhibitor nor disrupt binding to the catalytic domain, and as such it is tempting to speculate that another mechanism for this inhibition may be at play. We acknowledge that we have not conclusively addressed other possible mechanisms by which the truncated prodomain may affect the catalytic domain, including interference with proper folding, the formation of non-functional protein aggregates, or the binding of a motif other than the C-terminal tail that might compete against the intact prodomain of the substrate PCSK9 . Exogenous prodomain, in the form of a soluble chimeric prodomain-F c (IgG1) fusion, has been shown to bind to mature PCSK9 independent of the CHR domain and disrupt the interaction with the LDL-R (38). Such an effect may be similar to what we are seeing here, although this remains speculation. As we ultimately chose to pursue the relationship of the prodomain to protein secretion, further evaluation of these mechanisms was outside the scope of this study.
Intriguingly, in our substrate alanine scan we found evidence of a disconnect between our intermolecular cleavage system and intramolecular proteolysis. Notably, our P3 mutant (F150A) does not undergo proteolysis, but our P3Ј mutant (P155A) does, which is essentially the reverse of what has previously been described for intramolecular proteolysis (16). We speculate that this disconnect relates to the method of presentation of the substrate PCSK9 to the active site of the protease. The turn induced by Pro-155 may be required to internally align the cleavage sequence with the substrate binding groove yet may be less important for an intermolecular presentation. Likewise, thestacking interaction of Phe-150 with Trp-72 may be required to drive the initial recognition of the substrate PCSK9 with protease PCSK9⌬Pro but may be less important for an intramolecular process, which has a higher effective molarity to drive the hydrolysis.
Nevertheless, the development of our intermolecular proteolysis assay allowed us to probe the impact of specific point mutations on substrate cleavage while comparing the effects of these same mutations on protein secretion in an in trans assay that effectively bypassed the need for proteolysis. In this manner we specifically isolated the effects of each mutation on proteolysis and secretion and found that in certain cases the results were discordant. We specifically evaluated the effect of muta-tions at Gln-152 because the histidine mutation at this residue has been found in a hypocholesterolemic clinical cohort (36). Biochemical characterization of Q152H has previously shown that this mutant hinders PCSK9 processing by causing decreased intramolecular cleavage, leading to an intracellular accumulation of unprocessed proPCSK9 that interferes with wild-type proPCSK9 processing, producing a dominant-negative effect (37). Our results confirm the findings of the previous work; the Q152H indeed hinders intramolecular processing, and in addition, those proPCSK9 Q152H species that manage to undergo cleavage are then subjected to an additional selection against secretion. Whether it is the proPCSK9 Q152H or the cleaved PCSK9 Q152H that is predominantly responsible for the dominant-negative effect remains outside the scope of this investigation but is certainly of interest.
Within the past two years two specific mediators of PCSK9 secretion have been identified: Sec24a (39) and sortilin (40). These findings lend support to the concept that PCSK9 secretion occurs in a highly regulated manner. A recent report has also suggested that the ectodomain of the LDL-R acts as a chaperone for PCSK9 folding and cleavage, increasing throughput from the proPCSK9 form to the cleaved mature form (41). Furthermore, mutations in the CHR domain, which do not affect self-proteolysis, have been noted to cause PCSK9 loss-of-function by interfering with the secretion pathway (42). Our findings build on this concept, showing that the active site and its adjacent residues can modulate both proteolysis and secretion independently. Indeed, the active site Ser-386 itself is a prime example; we suspect that the general improvement in the in trans secretion of prodomain mutants with the inactive protease PCSK9⌬Pro-S386A , as compared with active protease PCSK9⌬Pro , is related to improved folding of the S386A catalytic domain and subsequently improved prodomain affinity. In support of this hypothesis, we have noticed similar improvements in the yield of a heterologous bacterial expression system, which is not dependent on secretion, with the S386A mutant (data not shown). Whether the residues of the prodomain C-terminal tail are critical for mediating the interactions with Sec24a, sortilin, or the ectodomain of the LDL-R would be an informative next line of inquiry. Our mutants may also help to identify the components of the complexes, such as the putative "cargo receptor" for Sec24a, that mediate these specific secretory events. In so doing we may uncover additional targets for anti-hypercholesterolemic therapies.
The structural information we infer from these mutagenesis experiments may aid in the development of specific secretion inhibitors. Because of the proximity of our mutations to the active site as well as their marked effect on protein secretion, it is reasonable to suggest that a small molecule bound to the active site could serve as an allosteric modulator of the secretion apparatus. Although V149A mutations are expected to abolish a critical isopropyl group that projects into the PCSK9 substrate binding groove, the F150A mutation abolishes astacking interaction with Trp-72, which appears to be critical for intermolecular proteolysis but tolerated by the secretion machinery. Likewise, the proteolytic machinery appears to be at least somewhat tolerant of the histidine imidazole at position 152, whereas the secretory machinery is more tolerant of the isoleucine sec-butyl chain than a charged imidazole. Such differences may allow us to infer different structure-activity relationships upon scaffolds identified as either proteolysis or secretion inhibitors and help to tailor the chemistry of such scaffolds toward inhibiting one pathway versus another. Additionally, our data showing that the C-terminal alanine insertion after Gln-152 retained secretory activity in the in trans system regardless of proteolysis (Fig. 4, D, right, lane 10, compared with E, lane 10) is consistent with prior data showing that a V5 tag inserted after the prodomain C terminus acts similarly in this same system (29). Although we have not tested this C-terminal alanine insertion in either intramolecular or intermolecular proteolysis, these data do raise the question of whether additional mutations in the full-length PCSK9 protein could permit secretion despite precluding proteolysis, allowing for secretion of the proPCSK9 form. Such a mutant might allow for structural data to target the proPCSK9 species directly.
Although our initial goal was to search for a chemically feasible way to inhibit PCSK9 proteolysis, we have uncovered an additional therapeutic target that appears as promising as proteolytic inhibition. Our data suggesting the decoupling of secretion from proteolysis implies that there are two different intracellular stages at which to target PCSK9. Furthermore, an approach to specifically target secretion would theoretically have the same effect as targeting proteolysis; it would prevent PCSK9 from properly reaching the cell surface and interacting with the LDL-R. Although an intracellular mechanism of LDL-R degradation has been noted (43), it is clear from both parabiosis experiments (10) as well as the marked success of anti-PCSK9 antibodies in clinical trials (12,13) that focusing on extracellular PCSK9 is a viable strategy. Because our data as well as that of others (37) suggest that inhibition of Q152H processing is in part mediated by a specific block on secretion, such a strategy would inch even closer to phenocopying this known genetic mutation. Given that this mutation is both dominant negative and is well tolerated (indeed, cardioprotective) in individuals, we see this as further encouragement for such an approach.