Development of protein-based inhibitors of the proprotein of convertase SKI-1/S1P: processing of SREBP-2, ATF6, and a viral glycoprotein.

Processing of membrane-bound transcription factors such as sterol regulatory element-binding proteins (SREBPs) and the ER-stress response factor ATF6, and glycoproteins of some hemorrhagic fever viruses are initiated by the proprotein convertase SKI-1/S1P. So far, no cellular protein-based inhibitor of the hydrophobic-amino acid specific SKI-1 is known. The prosegment of the basic-amino acid specific convertases (e.g. furin and PC5) or alpha(1)-PDX, a variant of alpha(1)-antitrypsin (alpha(1)-AT) exhibiting an RIPR(358) sequence at the reactive site loop, were shown to potently inhibit these secretory proteinases. Accordingly, we tested the SKI-1-inhibitory potential of various point mutants of either the 198 amino acid preprosegment of SKI-1-(1-198) or alpha(1)-AT. Transient transfections data showed that, out of numerous mutants studied, the R134E prosegment mutant or the alpha(1)-AT reactive site loop variants RRVL(358), RRYL(358) and RRIL(358) are the best specific cellular inhibitors of SKI-1. The observed inhibition of the processing of endogenous SREBP-2, exogenous ATF6 and a PDGF-A (RRLL(86)) variant were >55% and reach approximately 80% in stable transfectants. We also show that SKI-1 forms SDS-stable complexes with these alpha(1)-AT variants, but not with wild-type alpha(1)-AT or alpha(1)-PDX. Finally, these inhibitors were also shown to affect the processing and stability of the Crimean-Congo hemorrhagic fever virus glycoprotein.

Subtilisin-, kexin-, and furin-based studies established that the prosegment of these enzymes could act both as an intramolecular chaperone and a potent inhibitor (28 -31). The prodomain of PCs acts as a competitive inhibitor (31)(32)(33)(34), whereas the prodomain of the yeast kexin behaves as a mixed inhibitor with an IC 50 of ϳ160 nM (29). The wild-type prosegment of SKI-1 was also shown to inhibit this enzyme in vitro, albeit at micromolar concentrations (35). Finally, it was shown that ex vivo overexpression of the preproregions of furin (ppfurin), PC7 (ppPC7), and PC5 (ppPC5) resulted in potent but moderately selective cellular inhibition of their parent enzyme (31,33,36).
SKI-1 plays a crucial role in the regulation of lipid metabolism and cholesterol homeostasis through the processing of the sterol regulatory element-binding proteins, SREBP-1 and SREBP-2, which occurs in the cis/medial Golgi (38,39). Other type-II membrane-bound substrates include ATF6 that plays a major role in the unfolded protein response (UPR) to enhance the protein folding or refolding capacity of the secretory pathway (40,41), and the basic leucine zipper transcription factor Luman, the cellular counterpart of herpes simplex virus VP16 (42). Brain-derived neurotrophic factor (BDNF) is a soluble substrate and the study of its processing led to the initial cloning of SKI-1 (5). Mutation of proplatelet-derived growth factor A (pro-PDGF-A) at its furin-cleavage site (RRKR 86 ) into RRLL 86 (pro-PDGF-A*) resulted in a SKI-1 artificial substrate (43). Finally, SKI-1 was shown to play a major role in the processing of surface glycoproteins of infectious viruses such as Lassa (44,45), lymphocytic choriomeningitis (LCMV) (46,47) and Crimean Congo hemorrhagic fever (CCHF) (48) viruses. CCHF is a tick-borne member of the genus Nairovirus, family Bunyaviridae. The mature virus glycoprotein Gn is generated by proteolytic cleavage from Pre-Gn at the RRLL site by SKI-1 in the ER-cis Golgi compartments (48).
The three SREBP isoforms in mammals that belong to the basic helix-loop-helix leucine zipper (bHLH-Zip) family of transcription factors are designated SREBP-1a, SREBP-1c, and SREBP-2 (39). Synthesized as membrane-bound precursors, they are cleaved in an SREBP-cleavage-activating protein (SCAP)-and insulin-induced gene (Insig)-dependent fashion. When cellular cholesterol levels are high, Insig proteins bind and retain the SCAP-SREBP complex in the ER. When cells are deprived of sterols, Insig separates, allowing the transport of the SREBP-SCAP complex to the Golgi. Therein, a two-step proteolytic process (SKI-1/S1P and S2P) (49), releases the SREBPs from cell membranes, allowing their active N-terminal segments (nSREBP) to translocate to the nucleus, where they activate transcription of more than 25 mRNAs coding for proteins required for the biosynthesis and uptake of cholesterol and unsaturated fatty acids (50).
ATF6 is an ER type II leucine zipper transmembrane transcription factor held in the ER by Bip under normal conditions, with its N-terminal DNA binding domain facing the cytosol and its C terminus in the ER lumen (51). Accumulation of improperly folded proteins in the ER, calcium depletion by thapsigargin or inhibition of glycosylation by tunicamycin leads to an ER-stress response resulting in Bip dissociation from ATF6. The latter is then translocated in a SCAP-independent fashion to the Golgi where it is first cleaved by SKI-1/S1P and then by S2P. This releases the cytosolic N-terminal domain of ATF6 (nATF6), which subsequently reaches the nucleus to activate ER stress target genes (41,52).
SREBPs and their processing enzymes, SKI-1/S1P and S2P are certainly important targets for drug development. A sensi-tive SKI-1-specific fluorogenic assay has been developed based on the processing of a quenched fluorogenic peptide mimicking the glycoprotein processing site of Lassa (45) and CCHF (48) viruses. It was shown in vivo that SKI-1/S1P plays a crucial role in the processing of SREBPs in liver and is necessary for normal rates of triglyceride and sterol synthesis. Thus, whereas homozygote SKI-1/S1P (Ϫ/Ϫ) results in a lethal phenotype, conditional knockout in mouse liver results in 64 -83% decrease in the rates of cholesterol and fatty acid biosynthesis in hepatocytes (53). However, analysis of SKI-1 mRNA distribution revealed a widespread expression that suggests that this convertase plays other roles, as in neurons, bone and cartilage development (5). Indeed, it was recently shown that SKI-1 plays a critical role in cartilage development in zebrafish (54).
The critical implication of SKI-1 in various cellular functions and in certain pathologies emphasizes the importance of understanding the function of this convertase and of developing specific inhibitors that could modulate its activity in disease states. Whereas SKI-1 inhibition was recently achieved with 300 M of the general serine protease inhibitor AEBSF (55), it was not a specific SKI-1 inhibitor. In the present study, we introduced SKI-1 recognition motifs into the reactive site loop (RSL) of ␣ 1 -AT (P1-P4 positions) as one approach to the development of protein-based inhibitors. We also optimized the prosegment-based inhibition of SKI-1 and identified a unique R134E mutant exhibiting a potent inhibitory activity. These inhibitors represent protein-based inhibitors designed to specifically block intracellular SKI-1 activity.
Transient Transfection and Immunoblot Analysis of Endogenous Hamster SREBP-2-Chinese hamster ovary cells (CHO-K1) were set up at a density of 4 ϫ 10 5 cells per 60-mm plate for transfection. On day 1, cells were transiently transfected with 6 g of cDNA using Lipo-fectAMINE 2000, and cultured overnight in medium B (1:1 mixture of Ham's F-12 medium and Dulbecco's modified Eagle's medium containing 100 g/ml streptomycin sulfate) supplemented with 5% fetal calf serum. On day 2, cells were washed with PBS and then switched to medium with 5% lipoprotein deficient serum (LPDS) with 50 M compactin and 50 M sodium mevalonate in the absence or the presence of sterols for 18 h. Thereafter, the cells received N-acetyl-leucinal-leucinal-norleucinal (ALLN; Sigma) at a final concentration of 25 g/ml, a calpain and proteasome inhibitor that blocks the degradation of the mature form of SREBP-2, and the cells harvested 1-h later. Cells were then washed twice with PBS, lysed in 200 l of SDS buffer (10 mM Tris, pH 7.5, 100 mM NaCl, 1% SDS, and 25 g/ml ALLN), passed repeatedly through a 25-gauge needle, and centrifuged. Cell lysates were separated by electrophoresis on 6% SDS-polyacrylamide gel and the separated proteins were transferred to Hybond-C nitrocellulose membranes (Amersham Biosciences). These were then incubated with a mouse monoclonal antibody IgG-7D4 (dilution 1:200) directed against the N-terminal domain of hamster SREBP-2 (amino acids 32-250). The membranes were washed, and the immunoreactive proteins detected with horseradish peroxidase-conjugated anti-mouse IgG using an enhanced chemiluminescence (ECL) Western blotting detection system kit (Amersham Biosciences) according to the manufacturer's instructions.
Inhibitory Effect of Various Mutants on the Processing of ATF6 -Monolayers of CHO-K1 cells were set up (4 ϫ 10 5 cells/60-mm plate) and cultured in (1:1 mixture of Ham's F-12 medium and Dulbecco's modified Eagle's medium containing 100 units/ml penicillin and 100 g/ml streptomycin sulfate) supplemented with 5% fetal calf serum. The cells were co-transfected with a total of 6 g of DNA in a ratio of (1:4 substrate: inhibitor). With 3ϫ FLAG-ATF6 and mutant cDNA constructs using LipofectAMINE 2000. Two days post-transfection cells received a direct addition of 25 g/ml ALLN in the absence or presence of 2 g/ml tunicamycin (4 h) or 300 nM thapsigargin (1 h). Cells were then washed twice with PBS, lysed in 200 l of SDS buffer, passed 15-20ϫ through a 25-gauge needle, and centrifuged. The lysates were analyzed by immunoblotting using a 1:5000 dilution of anti-FLAG M2 monoclonal antibody (Stratagene). Horseradish peroxidase-conjugated goat anti-mouse IgG was used as secondary antibody at a 1:10,000 dilution, and the signal was visualized by the ECL chemiluminescence kit.
CCHF Virus Pre-Gn to Gn Processing and Inhibition-SW13 cells were maintained in DMEM supplemented with 10% fetal calf serum. SW13 cells stably expressing ␣ 1 -PDX and RRVL 358 were developed and maintained in the presence of G418 (300 g/ml). CHO-K1 derived cells expressing ␣ 1 -AT, RRVL 358 , RRYL 358 , and R134E were maintained in 1:1 proportion of DMEM and Ham's F-12 medium supplemented with 5% fetal calf serum. Hyperimmune mouse ascites fluid (HMAF) for CCHF virus proteins were kindly provided by T. Ksiazek, Centers for Disease Control and Prevention (Atlanta, GA). The antipeptide antibody used in this study was raised in rabbits against KLH-conjugated peptide sequences present in the mature Gn (amino acids 540 -551 of CCHF virus 1bAr 10200 strain: EIHGDNYGGPGD) and was done under contract with Research Genetics Inc, Huntsville, AL. NuPAGE ready-made gels (7% or 3-8%) and recommended buffers were purchased from Invitrogen.
Because the expression of CCHF virus glycoprotein required T-7 RNA polymerase to drive the T-7 promoter, vvT-7 vacccinia virus expression system was used as performed before (48). Briefly, 5 ϫ 10 5 cells were seeded onto 6-well plates a day before transfection and were infected for 1 h with vaccinia virus expressing T-7 RNA polymerase. Upon removal of the virus, cells were transfected with plasmid DNA encoding the wild type CCHF virus glycoprotein or in combination with other inhibitor expressing plasmids at 1:1 or 1:4 ratio. After 12 h, cells were labeled for 30 min with [ 35 S]cysteine and chased for 3 h. While using cells stably expressing the inhibitors, we infected them with vaccinia virus and transfected with 5 g of plasmid DNA expressing CCHF virus glycoprotein. Immunoprecipitated proteins were resolved in NuPAGE gels and proteins were visualized by autoradiography. The protein levels were quantified using a PhosphImager (ABI Prizm) using ImageQuant program. In experiments involving infection with CCHF virus, SW13 cells stably expressing ␣ 1 -PDX or RRVL 358 were infected with CCHF virus and labeled with 35 S in the Biosafety Level 4 laboratory at the CDC, Atlanta. Proteins were immunoprecipitated using HMAF, resolved in 7% NuPAGE, and bands visualized by autoradiography.
CTCACCCACGATATCATCACC CTTCGGCCAGTAACGTTAGGGG (43), we had shown that SKI-1 can process intracellularly an RRLL 86 mutant of pro-PDGF-A (pro-PDGF-A*) into PDGF-A (Fig. 1A), paving the way for a convenient ex vivo assay of SKI-1 inhibition using an anti-V5 Western blot analysis of secreted PDGF-A. The choice of CHO-K1 cells was dictated by the availability of CHO cells lacking SKI-1 (SRD-12B, called here SKI (Ϫ) cells) (56). In CHO-K1 cells, endogenous SKI-1 processes pro-PDGF-A*, whereas in SKI (Ϫ) cells such processing does not occur except in the presence of overexpressed SKI-1 (SKI (ϩ) ) cells (Fig.  1B). We used this assay to test the inhibition of intracellular SKI-1 activity in CHO-K1 cells by a series of mutants of either ␣ 1 -AT (Fig. 1C) or the preprosegment of human SKI-1 (ppSKI-1; Fig. 1D).

Inhibition of Pro-PDGF-A* (RRLL 86 ) Processing by SKI-1 in CHO-K1 Cells-Previously
In the ␣ 1 -AT variants, we mutated the reactive site loop sequence AIPM 358 (P1-P4 positions) to mimic the reported SKI-1 specificity for the motif (R/K)X(hydrophobic)Z2 (5-7), with a preference for the RRLL2 cleavage site reported for the SKI-1 prosegment site C (5), and the glycoproteins of Lassa virus (44,45) and CCHV virus (48). Co-expression of pro-PDGF-A or pro-PDGF-A* with either ␣ 1 -AT, ␣ 1 -PDX, or various serpin mutants are shown in Fig. 2. Western blots using an ␣ 1 -AT antibody revealed that ␣ 1 -AT and its mutants were expressed at similar levels (not shown). Under normal conditions, the 24-kDa pro-PDGF-A is very efficiently processed into a 16-kDa PDGF-A product. This process is completely inhibited by ␣ 1 -PDX, and most of the mutants analyzed did not have any significant inhibitory effect. However, a series of mutants RRIL 358 , RRYL 358 , RRVL 358 , and RRLL 358 , showed Ͻ 30% inhibitory effect ( Fig. 2A, bottom panels). The small amount of inhibition observed might be due to the presence of a dibasic motif in the reactive site loop that may inhibit the basic convertases. In contrast, the processing of pro-PDGF-A* is inhibited by the mutants RRLL 358 , RRVL 358 , RRYL 358 , RRIL 358 , and RRLE 358 , whereas ␣ 1 -AT, ␣ 1 -PDX and the other mutants tested were not inhibitory ( Fig. 2A, upper panels). These results emphasize that serpins with a P2 hydrophobic and P4 Arg are very efficient in blocking SKI-1 activity, consistent with the SKI-1 recognition motif (5-7).
The mutations of the prosegment of SKI-1 (ppSKI-1) were selected to evaluate the importance of B/BЈ primary autocatalytic zymogen processing sites (amino acids 130 -137), and the segment separating the B/BЈ-and the C-sites. As previously reported with in vitro data (35), wild-type ppSKI-1 effectively inhibits the intracellular processing of pro-PDGF-A* (Fig. 2B,  top panels). Retention of the prosegment in the ER through a C-terminal KDEL sequence abolished its inhibitory effect (data not shown), revealing that pro-PDGF-A* is processed at a post cis-Golgi compartment (57). Conversely, shortening of the prosegment to end at the C terminus of the B-site, i.e. at RSLK 137 (STOP), resulted in an inactive protein. Of the BЈ/B mutants tested, only R130A and R134E were found to be inhibitory. Following quadruplicate independent experiments with similar results, we can state that the order of inhibitory potency of the prosegments of pro-PDGF-A* processing by SKI-1 was found to be: WT-ppSKI-1 Ϸ R134E Ͼ R130A (Fig. 2B).

Inhibition of Endogenous SREBP-2 Processing by SKI-1 in
CHO-K1 Cells-We next investigated the intracellular inhibitory properties of the same series of ␣ 1 -AT and ppSKI-1 mutants (Fig. 1, C and D) on the processing of endogenous proS-REBP-2 (SREBP-2), a SKI-1/S1P substrate that is reported to be cleaved in the cis/medial Golgi (37). As shown in Fig. 3A, the membrane-bound SREBP-2 is not significantly processed into nSREBP-2 in cells incubated in medium containing cholesterol (ϩ), whereas its characteristic doublet nuclear N-terminal fragment (nSREBP-2) is observed in the absence of exogenous sterols (Ϫ), as previously reported (56). The latter is generated by the successive cleavage of SREBP-2 by SKI-1/S1P and S2P, and requires the exit of the SREBP-2-SCAP complex from the ER to reach the Golgi, a regulated escort mechanism requiring the removal of cholesterol (49). Thus, all inhibitors were tested in the absence of exogenous sterols. Under these conditions, and in transient transfections, the ␣ 1 -AT variants that were found to be significant intracellular inhibitors (56 -61% inhibition) were the RRYL 358 , RRVL 358 , and RRIL 358 mutants (Fig.   3A). In the same vein, only the R134E mutant of ppSKI-1 was inhibitory, exhibiting ϳ77% inhibition (Fig. 3B). In all cases equal protein levels were loaded onto the gels, as evidenced by the similar amounts of immunoreactive ␣ 1 -AT (Fig. 3A) or ␤-actin (Fig. 3B). Thus, different from the PDGF-A* processing (Fig. 2), the SREBP-2 processing by SKI-1 is not sensitive to ␣ 1 -AT RRLL 358 , RRLE 358 nor to WT ppSKI-1 and its R130A mutant (Fig. 3).
Because these assays relied on the transient transfection of CHO-K1 cells with cDNAs coding for potential inhibitors of the processing of endogenous SREBP-2, and as the transfection efficacy never exceeded 70 -80%, we sought to further test CHO-K1 stably expressing candidate inhibitors. In stable transfectants SREBP-2 processing was significantly inhibited by ppSKI-1 R134E (ϳ79%) and by ␣ 1 -AT RRYL 358 and RRVL 358 (ϳ65-68%) (not shown).
Inhibition of ATF6 Processing by SKI-1 in CHO-K1 Cells-We next tested the inhibitory properties of our mutants on ATF6 cleavage into its nuclear form nATF6, a processing that The precursor of SREBP-2 (SREBP-2) and the nuclear processed form (nSREBP-2) migrated as doublet (56) with apparent molecular masses of ϳ120 kDa and ϳ64 kDa, respectively. Stars indicate significant reduction in nSREBP-2 detection, emphasizing inhibition of SKI-1 activity. Lower panels show immunoblot analysis of the same membrane subsequently rehybridized with the ␣ 1 -AT antibody to confirm the level of expression (A) and of protein loaded based on ␤-actin immunoreactivity (B). The estimated % inhibition are shown at the bottom of each panel.

FIG. 2. Processing of pro-PDGF-A in CHO-K1 cells.
Western blot comparing the processing of pro-PDGF-A* and pro-PDGF-A (A) by either vector (pIRES2-EGFP), ␣ 1 -PDX, WT-␣ 1 -AT, various ␣ 1 -AT variants in 48-h post-transfection media, resolved by SDS-PAGE and immunoblotted using the V5 mAb. B, immunoblot analysis of the processing of pro-PDGF-A* and pro-PDGF-A co-transfected with ppSKI-1 mutants. The % inhibition of pro-PDGF-A or pro-PDGF-A* processing by the various ␣ 1 -AT mutants was estimated by quantitative analysis (Image Quant).
does not require the participation of SCAP, but rather the incubation of cells under stressful conditions resulting in the activation of the UPR, such as the treatment with the Ca 2ϩ ionophore thapsigargin or by the N-glycosylation inhibitor tunicamycin (40,52). Under normal conditions (Fig. 4A, upper panels), in the absence or presence of the chosen inhibitors the majority of ATF6 remains in the ER as an uncleaved ϳ90 kDa precursor (40). However, incubation of cells with either thapsigargin or tunicamycin (Fig. 4B, lower panels) results in the generation of a ϳ50 kDa nATF6 (40). From the co-transfection of ATF6 and inhibitor cDNAs, it is clear that, whereas ␣ 1 -AT and ␣ 1 -PDX had no effect, the ␣ 1 -AT variants RRVL 358 , RRYL 358 , RRIL 358 were relatively good inhibitors (ϳ53-70%), and ppSKI-1 R134E (ϳ74%) in a similar fashion to the processing of SREBP-2.
SDS-stable Complex Formation of SKI-1 with ␣ 1 -AT Variants-Serpins present a RSL that binds to the active site cleft of proteases in a substrate-like manner. The RSL is extremely flexible and assumes several different conformations, many of which are not immediately suitable for binding to the active site of the protease (58,59). An interesting property of inhibitory serpins is their propensity to form a complex with the target proteinase (EI*) that is resistant to heat and SDS treatment (60). The nature of the complex between the selected inhibitory ␣ 1 -AT variants and SKI-1 was examined ex vivo, in SKI (Ϫ) cells transiently overexpressing either active SKI-1 (Fig.  5A) or its inactive H249A mutant (Fig. 5B) (5), and selected ␣ 1 -AT variants. Western blot analysis of cell lysates following SDS-PAGE using a monoclonal antibody that recognizes the V5-epitope at the C terminus of active SKI-1 revealed that only the RSL mutants RRIL 358 , RRVL 358 , and RRYL 358 form SDSand heat-stable complexes with SKI-1 migrating at an apparent molecular mass of ϳ230 kDa (Fig. 5A). A minor form migrating at ϳ180 kDa was also observed. In contrast, neither WT-␣ 1 -AT nor ␣ 1 -PDX form such complexes with SKI-1 (Fig.  5A), and none of the serpins tested can form an SDS-stable complex with inactive SKI-1 (H249A) (Fig. 5B). We also noted that the propeptide mutant R134E and that the RSL mutants RRVL 358 and RRYL 358 inhibited the formation of both the B/BЈ and the active SKI-1 C-forms, suggesting that these are the best SKI-1 inhibitors (Fig. 5, A and C).
It should be noted that ␣ 1 -PDX was reported to form similar complexes with furin (22,61), PC5, and PACE4 (27). Similar analyses were performed in HK293 cells transiently overexpressing both SKI-1 and the serpins (Fig. 6). Again, we could detect the ϳ230 kDa intracellular complex between SKI-1 and either ␣ 1 -AT RRYL 358 , RRVL 358 , and RRIL 358 , but not with ␣ 1 -PDX or WT-␣ 1 -AT (Fig. 6A). Because membrane-bound SKI-1 autocatalytically sheds itself releasing a soluble secreted form (5,7), we also tested the media using an ␣ 1 -AT antibody and detected similar complexes (Fig. 6B). As a further control, we also tested whether these inhibitors can form complexes with the soluble basic-amino acid-specific PC5A (62). Although not shown, only ␣ 1 -PDX was shown to form a ϳ230-kDa com-  4. Effect of ␣ 1 -AT and proSKI-1 mutants on the processing of ER stress response factor ATF6. Immunoblots with anti-FLAG M2 antibody showing the proteolytic processing of 3xFLAG-ATF6 in CHO-K1 cells transfected with ATF6 (CTL), ␣ 1 -AT (WT), ␣ 1 -PDX, as well as ␣ 1 -AT and proSKI-1 mutant cDNAs. The cells were either (A) untreated (Ϫ) or (B) treated (ϩ) with 300 nM thapsigargin for 1 h or 2 g/ml tunicamycin for 4 h. Cells were lysed and total cell extract were resolved on 6% SDS-PAGE. The estimated % inhibition is shown. The * indicates the continued presence of N-glycosylated ER forms after 4 h of tunicamycin inhibition (51). plex with PC5A, both in cells and medium.
Inhibition of the SKI-1-mediated Processing of the CCHF Virus Glycoprotein-We have previously shown that the mature Gn (ϳ37 kDa) of CCHF virus is processed from Pre-Gn (ϳ140 kDa) by SKI-1 following the RRLL2 tetrapeptide sequence (amino acids 516 -519). Furthermore, a CHO-K1-derived cell line expressing a mutant form of SKI-1 (7) was defective in Pre-Gn processing (48). After screening many SKI-1 inhibitors and identifying a few potent ones effective for cellular proteins (Figs. 1-6), we wanted to analyze if those inhibitors were able to reduce or block the processing of a virus glycoprotein processed by SKI-1. For that purpose, CHO-K1 cells stably expressing ␣ 1 -AT, RRVL 358 , RRYL 358 , or R134E were infected by vaccinia virus expressing T7 RNA polymerase and transfected with a T7 driven plasmid expressing CCHF virus glycoprotein and analyzed for the inhibition in Pre-Gn processing. In cells expressing the CCHF virus glycoprotein alone or in combination with inhibitors, Pre-Gn was synthesized as a ϳ140 kDa protein, which was processed to Gn during a 3-h chase period (Fig. 7). Comparison of the levels of CCHF virus Pre-Gn at the 0 and 3 h chase time points shows that ϳ35% of Pre-Gn is proteolytically cleaved during this time period in the control CHO-␣ 1 -AT cells (labeled as WT). As expected, mature Gn is undetectable at the 0 time point, but accumulated in significant amounts following 3-h chase. Comparison of the levels of mature Gn accumulating 3-h postchase in the RRVL 358 , RRYL 358 , and R134E cells relative to the control cells, indicated levels of inhibition of ϳ55%, ϳ38 and ϳ70%, respectively (Fig. 7). Interestingly, these SKI-1 protease inhibitors also appeared to affect the stability of the CCHF virus Pre-Gn precursor protein. Although similar amounts of Pre-Gn are detectable in the various cells during the pulse label period, less of the protein remained detectable after 3-h chase in the RRVL 358 , RRYL 358 , and R134E cells relative to the control WT cells. This instability appears specific to Pre-Gn, as a panel of nonspecific proteins (indicated by asterisk) and cellular calreticulin remained unaffected (Fig. 7).
As CHO-K1 cells are refractory to CCHF virus infection, the experiments using the CHO derived cell lines could only be performed using transfected plasmids encoding the CCHF vi-rus glycoproteins, which does not allow analysis of the effects of these inhibitors on virus glycoprotein processing in the context of actual virus infection. To address this issue, we attempted to develop similar stably expressing cell lines but derived from SW13 cells that are the common cell line used for CCHF virus growth and experimentation. SW13 cell lines stably expressing ␣ 1 -PDX and ␣ 1 -AT-RRVL 358 were successfully obtained. As ␣ 1 -PDX (furin inhibitor) does not affect CCHF virus Pre-Gn processing to mature Gn (48), the ␣ 1 -PDX stably expressing cells may serve as a control for assessment of processing inhibition in ␣ 1 -AT-RRVL 358 -expressing cells. As expected, transient expression of CCHF virus glycoprotein expressed from transfected plasmids in the ␣ 1 -PDX-expressing SW13 cells resulted in the virus Pre-Gn being processed to mature Gn (Fig.  8A). Comparison of the level of accumulation of mature Gn in the ␣ 1 -PDX versus ␣ 1 -AT-RRVL 358 expressing cells in two independent experiments demonstrated ϳ65-70% inhibition of Pre-Gn processing (Fig. 8A). This level of inhibition is somewhat higher than that observed in the corresponding CHOderived cells (Fig. 7).
To determine the level of inhibition of virus Pre-Gn processing seen in the context of a CCHF virus infection, the ␣ 1 -PDX and RRVL 358 stably expressing SW13 cells were infected and the level of processing assessed (Fig. 8B). In CCHF virusinfected SW13-␣ 1 -PDX cells, the virus' Pre-Gn, Pre-Gc, Gc, and nucleocapsid proteins were visualized during the 30-min pulse and Gn appeared during the 5-h chase. Similar amounts of the virus proteins were observed in the ␣ 1 -AT-RRVL 358 expressing virus-infected cells with the exception of mature Gn, which appeared to be reduced (relative to the ␣ 1 -PDX expressing cells) by ϳ50 -55%, when normalized to nucleoprotein content. Taken together, these data and those from the CHO-derived stably expressing cell lines strongly suggest that SKI-1 protease inhibitors may represent promising compounds for further study in the development of effective antiviral strategies for control of CCHF virus infections.
In this work we demonstrated by four different ex vivo assays using pro-PDGF-A*, SREBP-2, ATF6, and Pre-Gn that the best SKI-1 inhibitors are consistently ppSKI-1 R134E and the ␣ 1 -AT variants RRVL 358 , RRYL 358 , and RRIL 358 . Notably, ppSKI-1 R134A was not inhibitory, whereas the R134E mutant was the most potent SKI-1 inhibitor identified in this study. Because the BЈ/B sequence cleaved is RKVF2 134 RSLK2 (7), the R134E mutation places a Glu at the P1Ј position of the BЈ site (RKVF2E 134 ) and at the P4 position of the B-site ( 134 ESLK-) (Fig. 1). The absence of a P4 Arg should block the B-site (7, 37) but could still allow the BЈ to interact with the SKI-1 catalytic pocket. It should be noted that the R134E mutant of the full-length SKI-1 is ϳ50% less processed into the BЈ/B and C forms than the wild-type sequence (7). The importance of a P1Ј acidic residue for SKI-1 activity was not studied before. Interestingly, the recently described crystal structure of furin clearly suggested that a P1Ј acidic residue would greatly favor the insertion of a substrate within the catalytic pocket (64). Indeed, a number of good furin/PC5 substrates such their own prosegments, ␣-integrins, and the Ebola virus glycoprotein do exhibit an acidic P1Ј residue (1). Although the structure of SKI-1 is yet to be unraveled, it is tempting to speculate that similar structural constraints to those reported for furin may be applicable for SKI-1.
Our studies with the RSL mutants of ␣ 1 -AT also unraveled the selectivity of the type of P2 hydrophobic residues acceptable by SKI-1, when P1 is Leu. Thus, whereas a P2 Val, Ile, and Tyr are conducive to an inhibitory RSL, P2 Pro, Leu, Phe are not (Figs. 2-4). It may well be that the RRLL sequence that is best recognized by SKI-1 in a number of substrates (7,44,48) may actually be cleaved when present in the RSL of ␣ 1 -AT. Thus, whereas only an RSL with a P1 Leu in combination with either P2 Val, Ile, and Tyr resulted in an inhibitory serpin, we cannot eliminate the possibility that a P1 Val or Ile with a P2 Val, Ile, or Tyr may not also be inhibitory. This is especially true as site-directed mutagenesis studies of SREBP-2 revealed that whereas SKI-1 can cleave the WT-sequence RSVL2 it does not process an RSVV mutant (65). However, the latter sequence may well be inhibitory.
The ability of ␣ 1 -PDX to form SDS-stable complexes with furin, PC5/6B (22,66) and PACE4 (27), but not with PC2 (22), led to test the possibility that variants of ␣ 1 -AT that better fit the SKI-1 recognition motif may also form stable complexes with this enzyme. Indeed, our data revealed that the best inhibitors of SKI-1 activity, namely ␣ 1 -AT RRVL 358 , RRYL 358 , and RRIL 358 do form such complexes with SKI-1 seen both in cells and media, whereas the non-inhibitory ␣ 1 -AT or ␣ 1 -PDX do not (Figs. 5 and 6). This observation is similar to what has already been published for ␣ 1 -PDX and furin in vitro (22) and in HK293 cells (27), and in all cases only a small portion of the total enzyme can be seen as a complex with the serpin on SDS-PAGE gels, suggesting that only a small amount of enzyme-serpin complex is truly resistant to SDS and heat. Thus, it is conceivable that the reactive site loop of the inhibitory ␣ 1 -AT variants exhibit typical serpin-like properties with respect to their inhibition of SKI-1.
Because SKI-1 silencing is lethal, conditional knockout experiments in liver revealed that silencing of its mRNA by ϳ80% results in ϳ50% decrease in circulating cholesterol and fatty acids (53). Therefore, the proposed R134E prosegment could conceivably represent a potentially useful inhibitor to reduce the activity of SKI-1 in vivo. Because the latter enzyme is also critical for the activation of some hemorrhagic fever viral surface glycoproteins (44 -46, 48), we also tested our best inhibitors on the processing of Pre-Gn into Gn of CCHF virus. Unlike the other cellular proteins analyzed in this report, expression of FIG. 7. Inhibition of CCHF viral glycoprotein A. CHO-KI-derived cells expressing ␣ 1 -AT, RRVL, RRYL, or R143E were infected with T7 expressing vaccinia virus and then transfected with a CCHF virus glycoprotein expressing plasmid. Labeled proteins were immunoprecipitated using an antipeptide antibody against CCHF virus Gn and resolved using 3-7% NuPAGE. 0 and 3 indicate the 30-min pulse and 3-h chase periods. Pre-Gn represents the precursor from which Gn (mature) is processed. The numbers at the left hand side denote the molecular weight size markers. In the bottom, immunoprecipitated calreticulin is provided as a loading control. The asterisk (*) also denotes the loading of equivalent proteins in the gel.
FIG. 8. Inhibition of CCHF virus glycoprotein processing by ␣ 1 -PDX and ␣ 1 -AT RRVL variant. SW13 cells expressing ␣ 1 -PDX and RRVL were tested for their inhibitory properties. A, cells were infected with vaccinia virus (10 MOI) and transfected with CCHF virus glycoprotein-expressing plasmid. CCHF Gn-specific proteins were immunoprecipitated by an antipeptide antibody and resolved in 7% NuPAGE. B, SW13 cells stably expressing ␣ 1-PDX or RRVL were infected with CCHF virus (CHF virus-specific) with virus adsorption for 1 h and labeling performed 18-h postinfection. Immunoprecipitation was performed using Virus HMAF and the proteins were resolved using 3-8% NuPAGE gels. Virus Pre-Gn and Pre-Gc represent the precursors of Gn and Gc, Gn and Gc are the processed forms and NP indicates the nucleocapsid protein.
the glycoprotein required the need for vaccinia virus to drive the T7 promoter. Our attempts to express glycoprotein using the CMV promoter or by providing T7 RNA polymerase from a plasmid were not successful. When using vaccinia virus, we do see inhibition of Pre-Gn processing in cells expressing ␣ 1 -AT-RRVL 358 and ppSKI-1 R134E. The fact that the Pre-Gn becomes unstable during the chase periods makes it harder to estimate the exact level of inhibition using the vaccinia system. Most of the experiments described for the inhibition of cellular proteins were performed with 1:4 substrate and inhibitor plasmid ratio. When we used that ratio in CCHF virus glycoprotein experiments, we observed a complete blockade of virus synthesis and glycoprotein stability even during the 30-min pulse period. We suggest that ␣ 1 -AT-RRVL 358 and ppSKI-1 R134E might be good inhibitors if we could manage to get the ideal expression with improved protein stability. On the other hand, SW13-derived ␣ 1 -AT-RRVL 358 cells showed promising inhibition (ϳ65-70% in transfections and ϳ50 -55% in CCHF virus infection assays) compared with ␣ 1 -PDX cells. In CCHF virus infected cells, there were no significant differences in the levels of expression of Pre-Gn, Pre-Gc, Gc, or nucleocapsid. We are in the process of developing cells stably expressing other promising inhibitors in the context of SW13 cells. Upon successful expression of all the inhibitors and necessary control, we intend to investigate the role of processing inhibition of Pre-Gn on virus assembly, release and infectivity with two target viruses (CCHF and arena viruses) which are causative agents of severe hemorrhagic manifestations in humans and whose glycoproteins are known to be activated by SKI-1. In this context, it was recently shown that lack of SKI-1mediated processing of the glycoprotein of lymphocytic choriomeningitis virus (LCMV) leads to non-infectious virus (47). If the lack of Pre-Gn processing negatively affects the above-mentioned parameters, then the inhibitors responsible for reducing or abrogating the processing would offer a good therapeutic potential.
In conclusion, the proposed inhibitors should be very useful in defining novel functions and/or substrates of SKI-1 in cell lines of choice as well as in vivo in various species. They could be used as prototypes for the development highly potent small molecule SKI-1 inhibitors that could find important clinical and pharmacological applications.