The Proprotein Convertase SKI-1/S1PALTERNATE TRANSLATION AND SUBCELLULAR LOCALIZATION*

Received for publication, April 16, 2007, and in revised form, July 2, 2007 Published, JBC Papers in Press, July 10, 2007, DOI 10.1074/jbc.M703200200

Philomena Pullikotil‡, Suzanne Benjannet‡, Janice Mayne§, and Nabil G. Seidah‡1

From the ‡Laboratory of Biochemical Neuroendocrinology, Clinical Research Institute of Montreal,Montreal, Quebec H2W 1R7, Canada and §Hormones, Growth, and Development, Ottawa HealthResearch Institute, The Ottawa Hospital, University of Ottawa, Ottawa, Ontario K1Y 4E9, Canada

Subtilisin kexin isozyme-1 (SKI-1) represents the firstmammalian member of secretory subtilisin-like processingenzymes that cleaves after nonbasic residues. It is synthesizedas an inactive precursor that undergoes three sequentialautocatalytic processing steps of its N-terminal prosegmentand an ectodomain shedding at a site near the transmem-brane domain. The various cellular functions of SKI-1emphasize the need to understand the sites of its activationand shedding. We have previously shown that SKI-1 under-goes autocatalytic shedding at the sequence KHQKLL9532,resulting in a membrane-bound stump called St-1 (aminoacids 954–1052). However, little is known about the cellularlocalization of SKI-1 or its shed forms. In the present study,we have further identified a smaller C-terminal fragment St-2generated closer to the transmembrane domain. By sequenc-ing and mass spectrometric analysis, the start site and themolecular mass of St-2 were determined. Site-directedmutagenesis revealed the critical amino acid involved in thisnovel process. Mutation of Met990 to M990A, M990I, andM990L failed to generate St-2, suggesting an internal alter-nate translation event at Met990, as confirmed by an in vitrotranscription/translation assay. Confocal microscopy de-fined the subcellular localization of SKI-1 and its fragments.The data show that most of membrane-bound SKI-1 and itsstumps St-1 and St-2 localize to the Golgi and can enter theendosomal/lysosomal compartments but do not sort to thecell surface. Deletion studies showed that the transmembranedomain of SKI-1 determines its trafficking. Finally, rSt-1 andrSt-2 seem to affect the processing of ATF6 by SKI-1, butcellular stress does not regulate the production of St-2.

Several secretory proteins are synthesized as inactive precur-sors, which when converted to their mature forms by proteo-lytic enzymes generate a large diversity of bioactive proteinsand peptides. The nine-member family of the proprotein con-

vertases (PCs)2 participates actively in the generation of suchmolecular diversity (1–3). There are seven basic amino acid-specific kexin-like mammalian proprotein convertases thatcleave various precursors at the general consensus motif(K/R)Xn(K/R)2, where n � 0, 2, 4, or 6, and X represents anyamino acid. The eighth member is the pyrolysin-like subtilisinkexin isozyme-1 (SKI-1) (4), also known as site-1 protease (S1P)(5). It cleaves substrates at the consensus motif (R/K)X-(hydro-phobic)-X2, where X is variable (6). The last member PCSK9cleaves the sequence VFAQ1522 within its prosegment (7).

SKI-1 represents the first mammalian member of secretorysubtilisin-like processing enzymes that cleaves after nonbasicresidues (2, 3). The ubiquitously expressed convertase SKI-1 (4)regulates the synthesis of cholesterol and fatty acids and theirmetabolism, through the processing of the membrane-boundtranscription factors sterol regulatory element-binding pro-teins (8, 9). It also regulates endoplasmic reticulum (ER)-stressresponse through cleavage of ATF6 (10) and is involved in theprocessing of probrain-derived neurotrophic factor (4).Recently, SKI-1 was reported to process Luman, a basic leucinezipper transcription factor, as well as other CREB-like bZIPfactors (11). The enzyme also plays a major role in the process-ing of surface glycoproteins of infectious viruses, such as Lassa(12), lymphocytic choriomeningitis (13), and Crimean Congohemorrhagic fever viruses (14).SKI-1 is synthesized as an inactive precursor of 1052 amino

acids (aa) that is activated following three sequential autocata-lytic processing events within its prosegment (sites B/B� and C;Fig. 1A) (15). The membrane-bound SKI-1 is rendered solublevia an autocatalytic ectodomain shedding event at aKHQKLL9532 site near the transmembrane domain (Fig. 1A)(15), similar to many cell surface proteins (e.g. growth factors,cytokine, growth factor receptors, and �-amyloid precursorproteins) (16). Shedding of SKI-1 releases a soluble protein sol-uble SKI-1 (aa 188–953), leaving behind a membrane-boundstub (St-1; aa 954–1052) (Fig. 1A). The biological importance of

* This work was supported by Canadian Institutes of Health ResearchGrant MOP-36496, by a Canada Chair, and by a private donation fromthe Strauss Foundation. The costs of publication of this article weredefrayed in part by the payment of page charges. This article musttherefore be hereby marked “advertisement” in accordance with 18U.S.C. Section 1734 solely to indicate this fact.

1 To whom correspondence should be addressed: Laboratory of BiochemicalNeuroendocrinology, Clinical Research Institute of Montreal, 110 Pine Ave.W., Montreal, Quebec H2W 1R7, Canada. Tel.: 514-987-5609; Fax: 514-987-5542; E-mail: seidahn@ircm.qc.ca.

2 The abbreviations used are: PC, proprotein convertase; SKI-1, subtilisin kexinisozyme-1; S1P, site-1-protease; ER, endoplasmic reticulum; St-1, stump-1;St-2, stump-2; rSt-1 and rSt-2, recombinant St-1 and -2, respectively; BTMD,before transmembrane domain; rBTMD, recombinant BTMD; CT, cytosolictail; rCT, recombinant CT; WT, wild type; �1-PDX, �1-antitrypsin Portland;CHO, Chinese hamster ovary; Ab, antibody; mAb, monoclonal antibody;Ab/N, anti-SKI-1 N terminus antibody; Ab/CT, anti-SKI-1 C-terminal anti-body; BFA, brefeldin A; CI-MPR, cation-independent mannose 6-phos-phate receptor; AEBSF, 4-(2-aminoethyl)benzenesulfonyl fluoride hydro-chloride; ALLN, N-acetyl-leucinal-leucinal-norleucinal.

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 282, NO. 37, pp. 27402–27413, September 14, 2007© 2007 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

27402 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 282 • NUMBER 37 • SEPTEMBER 14, 2007

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from

the shedding and the role, if any, of the membrane-bound stubSt-1 is unknown.In the present study, we have identified a new mechanism

generating a smaller membrane-bound C-terminal fragment,denoted as St-2 (Fig. 1A). Microsequencing and mass spectro-metric data revealed that St-2 starts at Met990. Site-directedmutagenesis and in vitro transcription/translation indicated aunique alternate translation event starting at Met990. Immuno-cytochemistry performed on endogenous and overexpressedSKI-1 and various constructs, including those coding for St-1and St-2, revealed that SKI-1 can sort to endosome/lysosomesbut not to the cell surface. These stumps seem to interfere withthe ability of SKI-1 to process pro-ATF6 into its nuclear form inresponse to cellular stress.

EXPERIMENTAL PROCEDURES

Cell Culture—Human embryonic kidney (HEK293) cellswere grown in Dulbecco’s modified Eagle’s medium with 10%heat-inactivated fetal bovine serum. Chinese hamster ovary(CHO)-K1, human hepatocytes (HuH7), M19 cells lackingsite-2 protease S2P (17), and Neuro 2A were grown in Ham’sDulbecco’s modified Eagle’s medium/F-12 medium containing10% fetal bovine serum. SRD-12B cells lacking SKI-1/S1Pexpression (18) were cultured in a 1:1 mixture of Ham’s F-12medium and Dulbecco’s modified Eagle’s medium containing100 �g/ml streptomycin sulfate supplemented with 5% fetalbovine serum, 5 �g/ml cholesterol, 1 mM sodium mevalonate,and 20 �M sodium oleate.Recombinant Stump-1 (rSt-1) and Its Mutants—All oligonu-

cleotides used in the various constructions are listed in Table 1.Toobtain enoughprotein for sequencing andmass spectromet-ric analysis, the C-terminal fragment of human SKI-1 (aa 954–1052) rSt-1was amplified by PCRusing pIRES-SKI-1-V5 cDNA(4) as template, using primers S1/AS1 with XhoI/BamHIrestriction sites. The PCR product obtained was digested withXhoI/BamHI and cloned into the vector pIRES-EGFP (Invitro-gen) to which the FLAG (DYKDDDK) epitope was fused inframe just after the signal peptide cleavage site and the V5(GKPIPNPLLGLDST) epitope placed at the C terminus of themolecule. All other point mutants of SKI-1 and rSt-1 were gen-erated with the QuikChange II site-directed mutagenesis kit(Stratagene) using the above constructs as template and appro-priate pairs of oligonucleotides (Table 1): I985A (S2/AS2),M990A (S3/AS3), M990L (S4/AS4), M990I (S5/AS5), I989L(S6/AS6), Y994A (S7/AS7), N995A (S8/AS8), Y/N994/995A

(S9/AS9). All recombinant cDNAs constructs were confirmedby DNA sequencing.Recombinant Stump-2 (rSt-2) and Cytosolic Tail (rCT)—The

smaller C-terminal fragments starting atMet990 (aa 990–1052)and rCT (aa 1023–1052) were PCR-amplified using SKI-1-V5as template and the oligonucleotide pairs S10/AS10 and S11/AS11, respectively. The PCR products obtained were digestedwith either XhoI/BamHI or PstI/AgeI and cloned into pIRES-EGFP, and their DNA was sequenced in both directions. Theother two constructs, r�CT and rBTMD, were previouslydescribed (15).Biosynthetic Analysis—HEK293 cells (2–4 � 105) in 60-mm

dishes were transiently transfected with 1.2 �g of SKI-1-V5cDNA using Effectene (Qiagen). Two days post-transfection,the cells were washed and pulse-labeled with 400 �Ci/ml[3H]Arg (Amersham Biosciences) for 4 h (19). The cell lysateswere immunoprecipitated with mAb/V5 (1:500) in buffer con-taining 150 mM NaCl, 50 mM Tris-HCl (pH 6.8), 0.5% NonidetP-40, 0.5% sodium deoxycholate, and a mixture of proteaseinhibitors (Roche Applied Science). The immunoprecipitateswere resolved by SDS-PAGE on 8% Tricine gels, dried, andautoradiographed as described (20).Western Blot Analysis—To detect the presence of St-2,

HEK293 cells were transiently transfected with either the wild-type or active site mutant H249A cDNA constructs, as men-tioned previously. CHO-K1 and SRD-12B cells were transientlytransfected with 4 �g of SKI-1 or SKI-1 mutant cDNAs usingLipofectamine 2000 (Invitrogen). Two days post-transfection,cells were washed twice with phosphate-buffered saline andlysed in SDS buffer (10 mM Tris-HCl (pH 7.5), 100 mM NaCl,and 1% SDS) containing a mixture of protease inhibitors. Thesamples were incubated on ice for 30 min, and the lysates wererun on 8% Tris-Tricine gel. Following SDS-PAGE, proteinswere transferred to nitrocellulose membrane (Amersham Bio-sciences) that was subsequently analyzed by immunoblottingusing mAb/V5 (1:5000), and the protein bands were visualizedby an ECL chemiluminescence kit (Amersham Biosciences),used according to the manufacturer’s instructions.Effect of Various Classes of Protease Inhibitors on the Gener-

ation of rSt-2—4 � 105 CHO-K1 cells were transiently trans-fected with 4 �g of SKI-1-V5 cDNA and rSt-1. 24 h post-trans-fection, the cells were washed and incubated for 6 h in mediumcontaining different protease inhibitors (21). Western blotanalyses were carried out following 8% SDS-PAGE usingmAb/V5 (1:5000). Inhibitors were tested at various concentra-

TABLE 1Oligonucleotides used for site-directed mutagenesis

Mutants Sense (S) Antisense (AS)S1/AS1 GATAAGCTCGAGTCCATTGACCTGGACAAGGTG CTTCGGCCAGTAACGTTAGGGGS2/AS2 GGCGCCTGGGACGCTCCTGGAGGGATC GATCCCTCCAGGAGCGTCCCAGGCGCCS3/AS3 CCTGGAGGGATCGCGCCTGGCCGCTAC GTAGCGGCCAGGCGCGATCCCTCCAGGS4/AS4 CCTGGAGGGATCCTGCCTGGCCGCTAC GTAGCGGCCAGGCAGGATCCCTCCAGGS5/AS5 CCTGGAGGGATCATACCTGGCCGCTAC GTAGCGGCCAGGTATGATCCCTCCAGGS6/AS6 ATTCCTGGAGGGCTCATGCCTGGCCGC GCGGCCAGGCATGAGCCCTCCAGGAATS7/AS7 ATGCCTGGCCGCGCCAACCAGGAGGTG CACCTCCTGGTTGGCGCGGCCAGGCATS8/AS8 CCTGGCCGCTACGCCCAGGAGGTGGGC GCCCACCTCCTGGGCGTAGCGGCCAGGS9/AS9 ATCATGCCTGGCCGCGCCGCCCAGGAGGTGGGCCAG CTGGCCCACCTCCTGGGCGGCGCGGCCAGGCATGATS10/AS10 GGGCGGTAGGCGTGTACGGTGG CGGGATCCTCACACCGAAGGGGTCTTTGS11/AS11 CAAACTGCAGATGAACAAGGCCAAGAGCAGGC GGCTTACCGGTCACCGAAGGGGTCTTTGG

Alternate Transcript of SKI-1/S1P

SEPTEMBER 14, 2007 • VOLUME 282 • NUMBER 37 JOURNAL OF BIOLOGICAL CHEMISTRY 27403

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from

tions (Table 2). Thus, the inhibitors were directed against (i)metalloproteases and matrix metalloproteases (EDTA (200�M), Captopril (100 �M) (Sigma), tissue inhibitors of metallo-proteinase TIMP1 (150 �M) and TIMP2 (180 �M), phosphor-amidon (10 �M), tumor necrosis factor-� protease inhibitorTAPI (100 �M) (Peptides International Inc.), and GM6001 (25�M) (Chemicon International Inc.)); (ii) serine/cysteine prote-ase inhibitors (4-(2-aminoethyl)benzenesulfonyl fluoridehydrochloride (AEBSF) (300�M), polyarginine (2�M), zinc sul-fate (ZnSO4) (50 �M) (Sigma), soybean trypsin inhibitor (15�M), leupeptin (100 �M) (Roche Applied Science), N�-p-tosyl-L-lysine chloromethyl ketone (100 �M), aprotinin (10 �M), phe-nylmethanesulfonyl fluoride (10 �M), and chymostatin (100�M) (Roche Applied Science)); (iii) proprotein convertaseinhibitor decanoyl-Arg-Val-Lys-Arg-choloromethylketone (10�M) (Bachem); (iv) aspartic protease pepstatin (1�M); (v) signalpeptide peptidase inhibitor (Z-LL)2-ketone (100 �M) (Calbio-chem); (vi) proteasome inhibitor N-acetyl-leucinal-leucinal-norleucinal (ALLN) (2 �M) (Sigma); (vii) cAMP-dependentprotein kinase inhibitor phorbol 12-myristate 13-acetate (10�M); (viii) �-secretase inhibitors �-XVIII (10 �M), �-IX (10 �M)(Calbiochem), and �-secretase inhibitor (100�M) (Enzyme Sys-tems ProductDFK-167); and (ix) brefeldinA (BFA) (5�M) (Cal-biochem). Some of the protease inhibitors were analyzed fol-lowing co-transfection (e.g. in CHO-K1 and Neuro 2A cells)with rSt-1 cDNA and protease nexin-1, Spn4.1A wild type(WT), Spn4.1RRLLmutant,�1-antitrypsin,�1-antichymotryp-sin, �1-PDX cDNAs at a ratio of 1:4 (substrate/inhibitor). 48 hpost-transfection, cells were washed and resolved on an 8%Tris-Tricine gel. Detection byWestern blotting was done usingmAb/V5 (1:5000).N-terminal Microsequencing Analysis of rSt-2—Transiently

transfected HEK293 cells were pulse-labeled for 4 h with 250�Ci/ml [35S]Met or 350 �Ci/ml [3H]Tyr (Amersham Bio-sciences). Lysates were immunoprecipitated with mAb/V5(1:500), and proteins were resolved by SDS-PAGE on 8% Tris-Tricine gel, immobilized to nitrocellulose PSQmembrane (Mil-lipore). The radiolabeled rSt-2 protein was excised and micro-sequenced as previously described on an Applied BiosystemsProcise cLC protein sequencer (20).Mass Spectrometric Analyses—CHO-K1 and HEK293 cells

were transfected with either pIRES-V5 vector control, SKI-1-V5, or rSt-1-V5 cDNAs. 48 h post-transfection, samples wereimmunoprecipitated using mAb/V5 (1:500). All of the boundantibody-antigen complex was eluted from Protein A beads byincubation in 2 � 150 �l of 0.1 M glycine (pH 2.8) for 10 min atroom temperature with shaking. Supernatants were collected,combined, and neutralized with 30 �l of 1 M Tris-HCl (pH 9.0).The elutants were concentrated 20� with an Amicon UltraYM3 Centricon (Millipore Corp., Bedford, MA) and equili-brated in 0.1% trifluoroacetic acid. For time-of-flight massspectrometric analysis, 10�l of the samplewas applied to anAuChip (CiphergenBiosystems Inc., PaloAlto, CA) and allowed toair-dry at room temperature. 1 �l of saturated sinapinic acid in50% acetonitrile plus 0.5% trifluoroacetic acid was added toeach spot. Mass spectrometric analysis was performed by time-of-flight mass spectrometry on a Ciphergen Protein BiologySystem II (PBS II). Analyses represent an average of 100 shots,

and masses were calibrated internally with All-in-1 ProteinStandards (Ciphergen Biosystems Inc., Palo Alto, CA).In Vitro Transcription/Translation—rSt-1 was excised from

pIRES-rSt-1-V5 using HindIII/BamHI and subcloned into thepCDNA3 vector. The mutants M990A and I989L wereobtained using the Quikclone mutagenesis kit (Stratagene).Recombinant pCDNA3-rSt1-V5 (WT), M990A and I989LcDNAs were transcribed/translated using a TNT coupledreticulocyte system according to the manufacturer’s protocolwith wheat germ extract (Promega) containing [35S]Met in thepresence or absence of canine pancreatic microsomal mem-branes. The reaction was carried out at 30 °C for 90 min, sam-ples were separated on SDS-PAGE, and Western blots wereanalyzed using mAb/V5 (1:5000).Production of Human SKI-CT Antibody—Human SKI-1

polyclonal antibody was generated by immunizing two rab-bits with a peptide containing an N-terminal Cys linked tothe segment aa 1036–1051 of SKI-1 (C-RPQLMQQVHPPK-TPSV), which was conjugated to keyhole limpet hemocyaninusing a standard keyhole limpet hemocyanin conjugation kitprotocol (Sigma). The two rabbit bleeds were analyzed, andthe best one (R2-02) called Ab-CT was used forWestern blotanalysis (1:1000) and subcellular localization (1:500) studies.Immunocytochemistry and Confocal Microscopy—To detect

endogenous SKI-1, human hepatocyte (HuH7) cells wereplated, and 24 h later the cells were fixed with 4% formaldehydefor 30 min at room temperature and washed with phosphate-buffered saline containing 0.1% Triton to permeabilize them.Immunostaining was done as described previously (22). Cellswere incubated with primary Ab/CT antibody (1:500) at 4 °Covernight along with markers for different compartments cis/medial Golgi (Golga1) (1:500), early endosomal marker (EEA1)(1:500), and late endosomal marker (cation-independent man-nose 6-phosphate receptor; CI-MPR) (1:500) (Abcam) at 4 °Covernight. The cells were washed and incubated with amixtureof two secondary fluorescently labeled antibodies (i.e. anti-rab-bit (Alexa Fluor 555) or anti-mouse (Alexa Fluor 647) (Invitro-gen) at room temperature for 1 h). Samples were analyzed on aZeiss LSM-510 confocal microscope. HuH7 cells (4� 105 cells)were transiently transfected with SKI-1-V5, rSt-1, rSt-2, �CT,rBTMD (15), and rCT constructs. 24 h post-transfection, thepermeabilized cells were treated and analyzed as above.Immunocytochemistry of SKI-1 at the Cell Surface—For cell

surface labeling, we used nonpermeabilizing conditions (23).HuH7 cells were transiently transfected with 4 �g of SKI-1cDNA, and 24 h later the cells were labeled with the fluorescentCT-B conjugate as recommended (Vybrant Lipid Raft LabelingKit; Molecular Probes). The cells were then washed three timeswith 1� phosphate-buffered saline fixed with 4% paraformal-dehyde for 1 h at 4 °C. After fixation, the cells were washedagain and incubated overnight with 1:500 Ab/N (15) and thenincubated with a secondary fluorescently labeled Alexa Fluor555-conjugated goat anti-rabbit IgG (Invitrogen) for 1 h atroom temperature.Effects of rSt-1 and rSt-2 on the Processing of ATF6 and Tuni-

camycin on theGeneration of St-2—4� 103CHO-K1 cells wereco-transfected with 1 �g of rATF6 and empty 3 �g of pIRES2

Alternate Transcript of SKI-1/S1P

27404 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 282 • NUMBER 37 • SEPTEMBER 14, 2007

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from

vector (control) or with 3 �g of cDNAs coding for SKI-1, rSt-1,its mutant I989L, or rSt-2. In another experiment, we used tri-ple transfections of 1 �g each of rATF6 and SKI-1, along witheither an empty 2 �g of pIRES2 vector (control) or 2 �g ofcDNAs coding for either rSt-1 or rSt-2. 24 h later, cells weretreated or not with 2 �g/ml tunicamycin for 12 h. Thereafter,the cells were treated with ALLN (Sigma) at a final concentra-tion of 25 �g/ml for 1 h. The lysates were resolved on 6% SDS-PAGE and analyzed byWestern blot using either a 1:5000 dilu-tion of anti-FLAG M2 monoclonal antibody (Stratagene), asreported (6, 24), or mAb/V5 at a 1:5000 dilution.

RESULTS

Analysis of the C-terminal Products of SKI-1—Followingtransient transfection in three cell lines of C-terminallyV5-tagged full-length SKI-1 (SKI-1), we analyzed the process-ing of the membrane-bound zymogen proSKI-1. HEK293 cellswere pulse-labeled with [3H]Arg for 4 h, the cell lysates wereimmunoprecipitated with mAb/V5, and the products wereanalyzed by SDS-PAGE in Tris-Tricine. The reported mem-

brane-bound active �106-kDaSKI-1 (Fig. 1A) autocatalyticallysheds itself to produce a solublesecreted �98-kDa soluble SKI-1(15), leaving behind a membrane-bound �13-kDa stump (St-1) (Fig.1B), as reported (15). However, inthis more resolving separation, wealso observed a smaller �8.5-kDaV5-positive product running at thebottom of the gel, estimated at11–17% of total (St-2; Fig. 1B), sug-gesting that it represents a shorterC-terminal fragment. To furtherprobe the possible cell type depend-ence of the generation of St-2, weanalyzed the SKI-1 products in twoother cell lines, namely SRD-12B(lacking SKI-1 expression) (18) andCHO-K1. These cells were tran-siently transfected with cDNAscoding for WT SKI-1, its active sitemutant (H249A), and its sheddingsite mutants (L952A, KLL/A, andK948A) (15). Cell lysates were ana-lyzed by Western blotting using amAb/V5, revealing the presence ofSt-1 and/or St-2 in all cells (Fig. 1B).In addition, although the autocata-lytic generation of St-1 is abrogatedin cells expressing the inactiveenzyme SKI-1-H249A and in thetriple shedding site mutant KLL/A(15), we observed that in all cases,the St-2 fragmentwas generated at alevel of �15%. Furthermore, unlikethe autocatalytic shedding eventleading to the production of St-1,

the generation of St-2 is an SKI-1 activity-independent process.This is based on the presence of this fragment in both SRD-12Bcells and the three cell lines expressing the inactive SKI-1-H249A (Fig. 1B). Based on our calculations, we can now statethat in CHO-K1 cells St-2 is �2-fold better generated from therSt-1 than from full-length SKI-1 or itsH249Amutant (Fig. 2,Aand B). The estimated percentage generation of St-2 from FL-SKI-1 and its H249A mutant are comparable in CHO-K1 cellsand somewhat lower in HEK293 cells. Since these values arerepresentative of at least five experiments, we trust that thisphenomenon is cell type-dependent. We note that the ratio ofSt-1 to St-2 is much higher in HEK293 cells than in CHO-K1cells. The reason behind this is not clear. The results also dem-onstrated that shedding into St-1 is not a prerequisite for thegeneration of St-2, since its production was observed in all shedsite mutants that result in reduced (L952A and K948A) orabsent (KLL/A) St-1 as well as in SKI-1-H249A (Fig. 1B).Inhibition of St-2 Production—In order to identify whether

the generation of St-2 is protease-dependent, we overexpressedWT SKI-1 (Fig. 2A) or a secretable form of St-1 containing a

FIGURE 1. Schematic representation of human SKI-1/S1P and the generation of St-2. A, schematic repre-sentation of human SKI-1 with respective autocatalytic processing sites. The arrows point to the positions ofthe signal peptide, A, B�/B, and C cleavages, and shedding site. The fragments generated after each event areindicated with numbered amino acids. The recognition sites of N-terminal Ab/N (aa 634 – 651) and C-terminalAb/CT (aa 1036 –1051), mAb/V5 antibodies used for immunoblotting and immunocytochemistry analysis, aredepicted. B, the leftmost panel shows the biosynthetic analysis of HEK293 cells transiently transfected withSKI-1-V5 and labeled with [3H]Arg. Cell lysates were immunoprecipitated and run on 8% Tris-Tricine gels andautoradiographed. The other panels represent Western blot analysis of HEK293, SRD-12B, and CHO-K1 tran-siently transfected with SKI-1, H249A, and shedding site mutants L952A, K948A/L952A/L953A (KLL/A), andK948A. 48 h post-transfection, cell lysates were resolved on SDS-PAGE, and the immunoreactive SKI-1-contain-ing proteins were revealed using mAb/V5. The arrowheads point to the migration positions of SKI-1, St-1, andSt-2. The percentage generation of St-2 is indicated at the bottom of each gel. These values are representativeof similar independent experiments repeated more than five times.

Alternate Transcript of SKI-1/S1P

SEPTEMBER 14, 2007 • VOLUME 282 • NUMBER 37 JOURNAL OF BIOLOGICAL CHEMISTRY 27405

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from

signal peptide followed by a FLAG tag at the N terminus and aV5 tag at the C terminus (rSt-1; Figs. 1A and 2B) in CHO-K1cells. Two days post-transfection, we incubated the cellsexpressing SKI-1 (Fig. 2A) with either various protease inhibi-tors targeting different classes of enzymes (Table 2) (21) or thegeneral serine protease inhibitor AEBSF (21) or brefeldin A(BFA), a fungal metabolite that leads to the dissociation andfusion of most of the Golgi apparatus with the ER (25). In addi-tion, we expressed either WT or SKI-1-H249A in M19 cellslacking the expression of site-2 protease S2P (17). We also co-expressed rSt-1 with general protease inhibitors, namely prote-ase nexin-1 (26), Spn4.1A WT and its RRLL derivativeSpn4.1A-RRLL (27), �1-antitrypsin, �1-antichymotrypsin, andthe furin/PC5 inhibitor �1-PDX (19, 28) in CHO-K1 andNeuro-2a cells (Fig. 2B). Cell lysates were analyzed byWesternblot using mAb/V5. Although all inhibitors did not affect theproduction of St-2, we confirmed the earlier observation thatAEBSF inhibits SKI-1 (21) and hence the autocatalytic genera-tion of St-1 (Fig. 2A, cross). In Fig. 2A, we note that ALLN, aproteasome and an inhibitor of Ca2�-dependent lysosomal cys-teine proteases (29), increases mostly the level of St-1 but notreally St-2 (30% compared with 25% for WT). Since we latershow that St-1 and St-2 as well as SKI-1 sort to endosomes/lysosomes, it is possible that St-1 is degraded by lysosomal pro-

teases, including Cys-proteases, andthe latter are inhibited by ALLN,which is not as specific a protea-some inhibitor as lactacystine. Sim-ilarly, EDTA seems to decrease theproduction of St-1, possibly by itsability to bind calcium,whichwouldreduce the activity of SKI-1 (4, 15).In view of its �8.5-kDa size andlikely membrane attachment, it waspossible that the signal peptidasemay be the cognate-convertingenzyme. However, incubation withthe signal peptidase inhibitor(Z-LL)2-chloromethylketone (30)had no effect on St-2 production(Fig. 2A, right). However, wenoticed that only BFA treatmentresulted in a major reduction orcomplete inhibition of the produc-tion of St-2 either from its inactiveH249A mutant (Fig. 2A, star) orresulting from expression of rSt-1(Fig. 2B, star). Altogether, the avail-able data suggest that St-2 produc-tion can occur in the ER and is abro-gated by BFA treatment. Since BFAwas never reported to be a proteaseinhibitor but could inhibit proteinsynthesis in cultured cells (31), wewere faced with the possibility thateither BFA may bring in from theGolgi an inhibitor or St-2 may notbe generated by proteolysis but

rather may be the result of an alternate mechanism that isinhibited by BFA. However, since the only known Golgi-resi-dent protease inhibitor PN1 (26) does not affect St-2 produc-tion (Fig. 2B), this suggests that the latter hypothesis is morelikely.N-terminal Sequencing and Mass Spectrometric Analysis of

St-2—Since the level of St-2 obtained from full-length SKI-1 israther low (Fig. 1B), and in order to isolate enough protein formicrosequencing, HEK293 cells were transiently transfectedwith theC-terminally V5-tagged rSt-1 (Fig. 1A). Twodays post-transfection, the cells were labeled with either [35S]Met or[3H]Tyr for 4 h, the lysates were immunoprecipitated withmAb/V5, and the proteins were separated by SDS-PAGE.Microsequencing data of the generated St-2 revealed the pres-ence of Met at positions 1 and 22 and Tyr at position 5, dem-onstrating that the St-2 protein starts at Met990 of SKI-1 (seesequence in Fig. 3A). To determine the molecular mass of St-2,we generated enough of this protein from 5 � 107 transientlytransfectedHEK293 andCHO-K1 cells with cDNAs of SKI-1 orrSt-1 and analyzed by time-of-flight mass spectrometry themolecular masses of immunoprecipitated products (Fig. 3B).The data showed the presence of an �8.5-kDa St-2 proteinobtained from both cell lines with an experimental molecularmass of 8472.7–8473.0 Da. Based on a calculated theoretical

FIGURE 2. Inhibition of the generation of St-2. Different classes of protease inhibitors were tested for theirability to inhibit the production of St-2. A, CHO-K1 cells were transiently transfected with SKI-1 or H249A. 24 hpost-transfection, cells were washed and incubated with different protease inhibitors for 6 h as describedunder “Experimental Procedures.” Following treatment, the cell lysates were resolved on 8% Tris-Tricine gels,and the immunoreactive proteins were revealed using mAb/V5. Indicated by a cross is the effect of AEBSF onthe production of St-1; the star points to the BFA treatment, which inhibits the production of St-2. B, co-transfection of rSt-1 cDNA (aa 954 –1052) along with various known serine protease inhibitors in CHO-K1 andNeuro2A cells at a ratio of 1:4 (substrate/inhibitor). 48 h post-transfection, the cell lysates were run on 8%Tris-Tricine gels and analyzed using mAb/V5. The star indicates the inhibition of St-2 in cells treated with BFA.The percentage generation of St-2 is indicated at the bottom of each gel. These values are representative ofsimilar independent experiments repeated more than five times. PMA, phorbol 12-myristate 13-acetate; SBTI,soybean trypsin inhibitor; TLCK, N�-p-tosyl-L-lysine chloromethyl ketone; PMSF, phenylmethanesulfonylfluoride.

Alternate Transcript of SKI-1/S1P

27406 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 282 • NUMBER 37 • SEPTEMBER 14, 2007

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from

mass of 8466.8 Da (lacking the N-terminal Met990), we con-cluded that most of the St-2 represents the segment (aa 991–1052 � 14-aa V5). However, microsequencing data of radiola-beled St-2 indicated that it starts at Met990, probably reflectingthe difference between the de novo synthetic product that stillretains some of theN-terminalMet990 and the steady state pro-tein, which lost the Met990, probably by an aminopeptidase,which recognizes to a lesser extent the Met-Pro991 sequence(32, 33). Accordingly, the generation of St-2 could be due tocleavage at DIPGGI9892M990PGRYN (Fig. 3A) or result fromalternative translation starting at Met990.IdentificationofMet990asaCriticalAminoAcidbyMutagenesis

and in Vitro Transcription/Translation Studies—To define theimportant amino acids involved in the generation of St-2, dif-ferent mutant constructs of rSt-1 were obtained using site-di-rected mutagenesis (Fig. 4A): I985A, I989L, M990L, M990A,M990I, Y994A, N995A, and the double mutant Y994A/N995A(Fig. 4A). These rSt-1 constructs were expressed in CHO-K1cells, and lysates were analyzed byWestern blot using mAb/V5to detect the generation of St-2 (Fig. 4B). Mutants of Met990 toleucine, alanine, and isoleucine failed to generate St-2 (Fig. 4B,stars), suggesting the critical importance of Met990 for the gen-eration of St-2.In order to test whether Met990 represents a start codon for

an alternatively translated transcript, we mutagenized thenucleotide present at positions�3 from theATG (ATCATG toCTCATG), which would create a less favorable Kozaksequence (34) and result in an I989L mutation. The data

clearly show that the latter muta-tion resulted in �90% reduction inthe generation of St-2 (Fig. 4B,star), whereas all other mutantshad no effect (Fig. 4B). An inde-pendent confirmation of this obser-vation was sought using the full-length protein. Accordingly, analysisof the WT and mutated M990ASKI-1 revealed that in both HEK293andCHO-K1 cells, theM990Amuta-tion completely blocked the genera-tion of St-2 from SKI-1 (Fig. 4C). Theabsolute requirement of methio-nine for St-2 generation stronglysuggested that this product was aresult of alternate translation start-ing at Met990. We therefore gener-ated constructs in the pcDNA3vector of WT and the M990A andI989L mutants of rSt-1 (Fig. 4A).These constructs were transcribedand translated in vitro using TNTQuick Coupled Transcription/Translation systems using a wheatgerm extract in the presence orabsence of canine microsomes, asrecommended by the manufac-turer (Promega). Microsomeswere added to enable efficient

FIGURE 3. Sequencing and mass spectrometric analysis. A, HEK293 cells were transiently transfected withrSt-1 cDNA and labeled for 4 h with [35S]Met or [3H]Tyr. Cell lysates were immunoprecipitated with mAb/V5 andresolved on an 8% Tris-Tricine gel. The deduced sequence positions of the radiolabeled residues are shown.B, mass spectrometric analysis of SKI-1 cleavage products. Shown are time-of-flight mass spectrometric anal-yses of the molecular masses of the proteins immunoprecipitated with mAb/V5 from cell lysates of HEK293 andCHO-K1 cells overexpressing either the control empty vector, rSt-1, or SKI-1-V5. The data show a peak at 8473.0Da following the sequence PGGIM9902 (experimental 8472.7– 8473.0 Da versus theoretical 8466.8 Da).

TABLE 2Protease inhibitors used for inhibition studies (21)

Protease class/Inhibitor Final concentration�M

MetalloproteaseEDTA 200TIMP1 150TIMP2 180Phosphoramidon 10GM6001 25TAPI 100Captopril 100

Serine and cysteine proteasesAEBSF 300Soybean trypsin inhibitor 15Aprotinin 10Phenylmethanesulfonyl fluoride 10ZnSO4 50Poly(R) 2Decanoyl-Arg-Val-Lys-Arg-choloromethylketone 10Chymostatin 100Leupeptin 100N�-p-tosyl-L-lysine chloromethyl ketone 100

Aspartic proteasePepstatin A 1ZLL-ketone 100

ProteasomeALLN 2

cAMP-dependent protein kinasePhorbol 12-myristate 13-acetate 10

�-Secretase�-XVIII 10�-IX 10�-Secretase 100

Antibiotic from brefeldianumBFA 5

Alternate Transcript of SKI-1/S1P

SEPTEMBER 14, 2007 • VOLUME 282 • NUMBER 37 JOURNAL OF BIOLOGICAL CHEMISTRY 27407

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from

cleavage of the signal peptide, membrane insertion, andtranslocation ensuring higher translation efficiency. Thereaction products were separated on an 8% Tris-Tricine gel,and the proteins were transferred to a nitrocellulose mem-brane and immunoblotted using mAb/V5 (Fig. 4D). The dataclearly show the presence of St-2 in the WT, which isreduced by�70% in the I989Lmutant and completely absentin the M990A mutant (Fig. 4D). Thus, all available data sup-port the conclusion that St-2 is produced by alternate trans-lation. Since it lacks a signal peptide, rSt-2 is expected tohave an opposite membrane orientation compared withSKI-1 or rSt-1. It was therefore of interest to define the sub-cellular localization of rSt-2 compared with either that ofendogenous SKI-1 or its recombinant WT SKI-1 and rSt-1.Generation and Analysis of a Specific C-terminal Tail SKI-1

Antibody—To study the subcellular localization of SKI-1-WT,rSt-1, rSt-2, and various recombinant constructs (Fig. 1A), wesynthesized the 17-mer peptide C1037RPQLMQQVHPPKT-PSV1052, representing the 16 aa of the C terminus of hSKI-1linked to anN-terminal cysteine residue, whichwas conjugatedto the carrier protein keyhole limpet hemocyanin. This peptideantigen was injected into two rabbits, and the serum from eachbleed was collected. The properties of the best resulting poly-clonal antibody (Ab/CT, R2-02) were compared with those ofmAb/V5 byWestern blot and immunocytochemistry.Western

blot analysis of transiently trans-fected human hepatocytes HuH7and HepG2 cells with the WTcDNA revealed that both antibodiesrecognize the A, B/B�, and C formsof SKI-1 (Fig. 5A). Confocal analysisof the subcellular localization ofSKI-1 by immunocytochemistry(22) was performed 24 h after tran-sient transfection of HuH7 cellswith WT SKI-1 cDNA and its inac-tive H249A mutant (Fig. 5B). WTSKI-1 mostly concentrated inparanuclear structures reminiscentof the Golgi (Fig. 5B, arrows),whereas the active site mutantH249A exhibits a diffused perinu-clear staining indicating its pres-ence in the ER (Fig. 5B,arrowheads).Thus,Western blot and immunocy-tochemical analyses give similarresults using either mAb/V5 orpolyclonal Ab/CT.Subcellular Localization of Endo-

genous SKI-1 and Its RecombinantConstructs—Western blot analysis ofendogenous SKI-1 inHepG2 and/orHuH7 cells revealed similar A, B/B�,and C forms (Fig. 6A). Subcellularlocalization studies in HuH7 cells(Fig. 6B) revealed that SKI-1 co-lo-calizes with markers of differentsubcellular compartments (i.e.

Golga1 (a cis/medial Golgi marker), EEA1 (an early endosomalmarker), and CI-MPR (a late endosomal marker) (22)).Although this agrees with the earlier report on the presence inCHO-K1 cells of endogenous SKI-1/S1P in the cis/medial Golgi(35), our present data further revealed that endogenous SKI-1can also traffic to endosomes/lysosomes. Upon overexpressionof SKI-1 in either HEK293 cells (4) or HuH7 cells (Fig. 6C), weobserved similar punctate localizations reminiscent of endo-somes/lysosomes (Fig. 6D) (4). However, immunocytochemicalanalysis of overexpressed SKI-1 using Ab/N recognizing aa634–651 (15) did not reveal the presence of this enzyme at thecell surface, as compared with the endogenous raft CT-B com-ponent cell surfacemarker (Fig. 6E). These data suggest that theendosomal/lysosomal localization of SKI-1 probably resultsfromadirect transport from theGolgi to endosomes/lysosomesand not via cell surface internalization, as is the case for furin(36).We next analyzed the localization of overexpressed shorter

forms of SKI-1 (Fig. 1A), including rSt-1 (Figs. 6C and 7A) andrSt-2 (Figs. 6C and 7B), r�CT lacking the C-terminal cytosolictail but still keeping the transmembrane domain (Fig. 7C) (15),rBTMD, lacking both transmembrane domain and cytosolictail (Fig. 7D) (15), and the cytosolic tail construct with an addedinitiator methionine rCT (Fig. 7E). Except for the rBTMD andrCT, localization studies show that all other constructs traffic to

FIGURE 4. Mutation analysis identifies the critical amino acid Met990 for the generation of St-2. A, sche-matic representation of rSt-1 (aa 995–1052) with the signal peptide and N- and C-terminal recognition tagshighlighting the different mutants generated. B, Western blot analysis of CHO-K1 cells transiently transfectedwith the rSt-1 cDNA construct revealed with mAb/V5. Mutation of amino acid Met990 into Leu, Ala, and Ilecompletely inhibited the generation of St-2, whereas Ile989 inhibited it by almost 90% as indicated by stars.C, CHO-K1 and HEK293 cells transfected with SKI-1 or its M990A mutant. Analysis by Western blotting usingmAb/V5 demonstrates the failure to generate St-2 in both cell lines. D, in vitro transcription/translation of WTand its mutants M990A and I989L cloned into pcDNA3 vector. These constructs were transcribed/translated invitro with wheat germ extract in the presence and absence of canine microsomes. The products were sepa-rated on a 8% Tris-Tricine gel, transferred onto a nitrocellulose membrane, and revealed with mAb/V5. Thepercentage generation of St-2 is indicated at the bottom of each gel. These values are representative of similarindependent experiments repeated at least three times.

Alternate Transcript of SKI-1/S1P

27408 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 282 • NUMBER 37 • SEPTEMBER 14, 2007

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from

the same subcellular compartments as SKI-1 (Figs. 6D and 7).rBTMD is not detectable intracellularly, since it is rapidlysecreted (15). Interestingly, rCT that exhibits a putative basicnuclear localization signal (KRRK1032) colocalized with thenuclear marker TOPO3, with a major accumulation withinnucleoli (Fig. 7E). The transmembrane domain is common toSKI-1, rSt-1, rSt-2, and r�CT. It seems to be the dominantdomain that determines the localization of these proteins to theGolgi, and subsequent subcellular compartments (i.e. endo-somes/lysosomes), since its absence results in a secreted pro-tein that does not accumulate inside the cell.Overexpressed rSt-1 or rSt-2 Impede ATF6 Processing by

SKI-1—It is well known that treatment of cells with tunicamy-cin induces ER stress, resulting in increased processing of ATF6by SKI-1/S1P followedby S2P into a nuclear form (10, 21, 24). Inan effort to define a functional role for either rSt-1 or rSt-2, weinvestigated the effect of either construct on the processing ofATF6 precursor (pATF6) into its nuclear form (nATF6) understress conditions, such as tunicamycin treatment (10). The datashow that co-expression of either rSt-1, its I989L mutant, orrSt-2withATF6 reduces its processing into the nuclear formby�55, 32, and 55%, respectively, whereas SKI-1 increases thelevel of the latter by �1.7-fold (Fig. 8, left). St-1-induced reduc-tion of ATF6 processing is �40% less by the I989L mutant thatexhibits a reduced production of St-2. This suggests that St-2 isa better inhibitor of ATF6 processing than St-1. The data alsoshow that co-expression of ATF6 together with SKI-1 and witheither rSt-1 or rSt-2, respectively, reduces�60 and�72%of thephenotype triggered by SKI-1 alone (Fig. 8, right). In order to

test whether stress conditions regu-late the alternate translation ofSKI-1, resulting in the production ofSt-2, we analyzed the effect of tuni-camycin treatment on the genera-tion of St-2. Accordingly, CHO-K1cells were transiently transfectedwith either FL-SKI-1, its H249Amutant (Fig. 9A), rSt-1, rSt-2, andM990Amutant (Fig. 9B) were testedin the absence or presence of tuni-camycin. Western blot analysis ofthe cell lysates revealed that stressinduced by tunicamycin does notaffect the percentage production ofSt-2 from either FL-SKI-1, itsH249A mutant (Fig. 9A), rSt-1, orrSt-2 (Fig. 9B).

DISCUSSION

Since the discovery of the propro-tein convertase SKI-1/S1P (4, 18), itwas realized that this convertase canprocess a number of membrane-bound proteins in the cis/medialGolgi, including transcription fac-tors (2, 3, 11, 24, 37–39) and viralsurface glycoproteins (6, 12, 14, 40).Biosynthetic analysis of the zymo-

gen-processing SKI-1 revealed an ordered autocatalytic activa-tion process leading to the formation of intermediate B/B�forms in the ER and then an active C-form in the cis/medialGolgi (4, 15, 41, 42). Indeed, it was soon found that most of theSKI-1 protein localizes in early Golgi compartments in bothHEK293 cells (4) and in CHO-K1 cells (35). However, none ofthese data revealed whether SKI-1 can transiently be localizedat other subcellular compartments, whichmay be implicated inits further processing via autocatalytic shedding into a solublesecreted form (15). It was originally suggested, but not proven,that SKI-1 may be found in punctate structures reminiscent ofendosomes/lysosomes and that its concentration could beincreased upon neutralizing some lysosomal proteases (4). Inthis study, we investigated in more detail the trafficking ofSKI-1, its autocatalytic shedding into a secreted solubleSKI-1 and a membrane-bound stump St-1, and the genera-tion of a novel membrane-bound form called St-2, independ-ent of SKI-1 activity and/or shedding (Fig. 1).In agreement with an earlier report (15), we confirmed that

SKI-1 activity is needed for its shedding, since both the activesite mutant H249A and the SKI-1 inhibitor AEBSF (21, 43) pre-vented shedding (Figs. 1 and 2). Biosynthetic radiolabelinganalysis of either SKI-1 or BTMD-SKI-1-KDEL, which remainsin the ER (15), revealed that treatment with brefeldin A orexpression of the SKI-1-KDEL construct prevented the produc-tion of the St-1 fragment (not shown). This suggests that shed-ding requires the exit of SKI-1 from the ER and/or acidic con-ditions found in the cis/medial Golgi. In contrast to themembrane-bound convertases PC7 (44), PC5B (45, 46), and

FIGURE 5. Comparative analysis of C-terminal polyclonal Ab/CT with mAb/V5. A, schematic representationof human SKI-1 with respective autocatalytic processing sites. The arrows point to the positions of the signalpeptide, A, B�/B, and C cleavages. The recognition sequence positions of Ab/N, Ab/CT, and mAb/V5 are shown.B, HuH7 and HepG2 cells were transiently transfected with SKI-1 cDNA. 48 h post-transfection, cells were lysedand run on an 8% glycine gel and revealed using mAb/V5 or Ab/CT. The data show the processing of SKI-1 intoA, B/B�, and C forms. The pattern of recognition using both antibodies is comparable. C, HuH7 cells transientlytransfected with SKI-1 or its active site mutant H249A. The SKI-1 protein was visualized by immunofluorescenceby confocal microscopy. Both antibodies reveal that H249A localizes to the ER, whereas SKI-1 localizes to theparanuclear structures reminiscent of the cis/medial Golgi. Bar, 10 �m.

Alternate Transcript of SKI-1/S1P

SEPTEMBER 14, 2007 • VOLUME 282 • NUMBER 37 JOURNAL OF BIOLOGICAL CHEMISTRY 27409

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from

furin (36), which cycle from the cell surface to endosomes,SKI-1 does not seem to localize to the cell surface (Fig. 6E) andhence may directly sort to endosomes/lysosomes (Fig. 6) fromthe trans-Golgi network. This is in agreement with our obser-vation that shedding is not affected by inhibitors ofmetallopro-teases (Fig. 2 and Table 2), which usually process membrane-bound proteins at the cell surface (16, 47, 48). We deduce thatSKI-1 autocatalytic shedding may well occur in the Golgi,where most of the active SKI-1 resides (Fig. 6), and not at thecell surface, where the sheddases are usually localized.Microsequencing and mass spectrometric analyses demon-

strated that St-2 starts at Met990 and consists of the segmentcomprising amino acids 990–1052 (Fig. 3). Since the produc-tion of St-2 is inhibited by BFA treatment (Fig. 2) that can pre-vent mRNA translation (31) and is abrogated when Met990 isreplaced by other amino acids in either the full-length proteinor in the rSt-1 construct (Fig. 4), we suspected that St-2 may beproduced by an alternate translation of the primary transcriptat Met990. In vitro transcription/translation analyses of rSt-1and its mutants M990A and I989L (Fig. 4D) confirmed thatproduction of St-2 utilizes an internal translation start site at

Met990. Although viruses com-monly make use of this mechanismto express variant proteins, alter-nate translation of secretory pro-teins is rather rare (49). Initiation atMet990 could either be attributed tointernal entry of the ribosomal com-plex, instead of its attachment to the5� cap, or to leaky ribosomal scan-ning due to less favorable nucleotidesequence surrounding the initiatormethionine. Based on Kozak’s rules(34, 50, 51) the consensus sequencearound the initiator methionine isGCC GCC (A/G)CC ATG (G/A)(the underlined residues are themost critical). We note that thehuman SKI-1 nucleotide sequencesurrounding Met1 is CTT GTGACC ATG A, whereas that aroundMet990 is GGT GGG ATC ATG C.This shows that although bothsequences exhibit optimal nucleo-tides at �3- and �6-positions, thefirst one has additionally an optimal�4 nucleotide, whereas the secondone has a �9 nucleotide. The effi-ciency of translation is clearly infavor of Met1 (Fig. 1), befitting thepredictions of the scanning hypoth-esis (34, 50, 51). Interestingly, theATG coding forMet990 is conservedin most vertebrates known toexpress SKI-1, from Danio rerio toXenopus tropicalis and Homo sapi-ens. Since it is found in the last exonXXIV, which is five nucleotides

away from the 3� acceptor splice site (52), this may suggest thatthe C-terminal segment encoded by exon XXIV may have aspecific function.Interestingly, although the membrane orientation of St-2 is

expected to be reversed from that of SKI-1 or St-1, they all sharethe same transmembrane domain, which seems to be criticalfor their subcellular localization to the Golgi and endosomes/lysosomes (Figs. 6 and 7). A similar critical importance of thetransmembrane domain for subcellular localization in the Golgiand possibly endosomes has been reported for the �-secretasesBACE1 (53) and BACE2 (54). Intriguingly, we observed thatexpression of the highly basic cytosolic tail of SKI-1, lackingacidic residues (4), localizes the polypeptide to nucleoli (Fig.7E), a subnuclear structure involved in ribosomal RNA genetranscription. Mutation of the stretch KRRK1032 to KAAK1032

or the putative phosphorylation sites S1026A, S1052A, andT12049A did not change such nucleolar localization (notshown). However, the addition of a V5 tag at the C terminus ofthe cytosolic tail prevented its nuclear localization (not shown),suggesting that the nuclear localization determinants mayreside at the C terminus of this sequence.

FIGURE 6. Subcellular localization of endogenous and overexpressed SKI-1. A, to detect endogenous SKI-1,cell lysates from the human cell lines HuH7 and HepG2 were resolved by 6% SDS-PAGE and immunoblottedusing Ab/CT. The different forms of endogenous SKI-1, A, B/B�, and C forms are indicated. B, immunofluores-cence analysis by confocal microscopy of endogenous SKI-1. Merged images of SKI-1 (red labeling), cis/medialGolgi (Golga1), early (EEA1), and late (CI-MPR) endosomal markers (blue labeling) show co-localization of SKI-1with each of the above markers (pink). C, Western blot using Ab/CT of the lysates of HuH7 cells overexpressingSKI-1, rSt-1, or rSt-2. D, subcellular localization studies using confocal microscopy of overexpressed SKI-1 (Ab/CT; red) and any of the subcellular markers Golga1, EEA1, or CI-MPR (blue). The merged images representco-localized SKI-1 in the Golgi, early, and late endosomal compartments. Bar, 10 �m. E, cell surface labeling ofSKI-1 under nonpermeabilizing conditions. HuH7 cells transiently transfected with SKI-1 were labeled with thered fluorescent CT-B component lipid rafts as cell surface marker and with Ab/N to look for possible SKI-1localization at the cell surface. Bar, 10 �m.

Alternate Transcript of SKI-1/S1P

27410 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 282 • NUMBER 37 • SEPTEMBER 14, 2007

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from

Finally, in an effort to define a possible functional role for theobserved stumps St-1 and St-2, we checked whether the SKI-1generated processing of the ER stress response factor ATF6was

FIGURE 7. Immunocytochemical localization of rSt-1 and rSt-2. HuH7 cells were transiently transfected with cDNAs coding for rSt-1 (A), rSt-2 (B), r�CT (C),BTMD (D), or rCT (E). The cells were fixed, permeabilized, and analyzed for expression of SKI-1 using Ab/CT or Ab/N (red), along with various markers (blue), suchas the Golgi marker, Golga1, early endosomal marker, EEA1, late endosomal marker, CI-MPR, or nuclear marker, TOPO3. The presence of pink coloring in mergedimages demonstrates co-localization of SKI-1 or its fragments with appropriate markers of different compartments. Bar, 10 �m.

FIGURE 8. Effect of rSt-1 and rSt-2 on ATF6 processing. Western blot anal-ysis of ATF6 using anti-FLAG antibody of CHO-K1 cells transiently co-trans-fected with cDNAs coding for ATF6 and either empty vector (pIRES) or SKI-1,rSt-1, its mutant I989L, or rSt-2 (right) or triply transfected with cDNAs codingfor ATF6, SKI-1, and either control pIRES, rSt-1, or rSt-2. 24 h later, cells weretreated with 2 �g/ml tunicamycin for 12 h. Thereafter, the cells were treatedwith 25 �g/ml ALLN for 1 h. The lysates were resolved on 6% SDS-PAGE. Themigration positions of the ATF6 precursor (pATF6, �90 kDa) and its nuclearform (nATF6, �50 kDa) as well as the percentage cleavages of pATF6 intonATF6 are indicated. These values are representative of similar independentexperiments repeated three times.

FIGURE 9. Effect of tunicamycin-induced stress on the generation of St-2.Shown is Western blot analysis using mAb/V5 of CHO-K1 cells incubated inthe presence or absence of 2 �g/ml tunicamycin (Tm) and transiently trans-fected with cDNAs coding for FL-SKI-1 or its active site mutant H249A (A) andrSt-2, rSt-1, or its mutant M990A (B). The samples were analyzed as in Fig. 2.The percentage generation of St-2 is indicated at the bottom of each gel.These values are representative of similar independent experimentsrepeated three times.

Alternate Transcript of SKI-1/S1P

SEPTEMBER 14, 2007 • VOLUME 282 • NUMBER 37 JOURNAL OF BIOLOGICAL CHEMISTRY 27411

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from

affected by the overexpression of either constructs. A repre-sentative figure of our three independent experiments showsthat both rSt-1 and rSt-2 diminish the processing of pATF6 intonATF6 by an as yet undefinedmechanism (Fig. 8, left). Further-more, we also show that the co-expression of SKI-1 with eitherrSt-1 or rSt-2 reduces the phenotype triggered by SKI-1 alone(Fig. 8, right). Finally, tunicamycin-induced stress does notseem to increase the level of production of St-1 (Fig. 9), suggest-ing that the observed phenomenon of alternate translation isnot regulated by cellular stress. It is a matter of speculationwhether themembrane-bound stumps St-1 and St-2may inter-fere with the ER-stressmachinery either in the ER or cis/medialGolgi. Alternatively, the rSt-1 and rSt-2 may act as dominantnegatives of SKI-1.In conclusion, the major SKI-1 activity is expected to be

mostly in the cis/medial Golgi and may also be found in endo-somes. However, although many candidate membrane-boundtranscription factor substrates of SKI-1 are known (2, 3, 11, 24,37–39), the putative endosomal substrates are yet to be identi-fied. Alternatively, the presence of SKI-1 in endosomes/lyso-somesmay reflect its normalmetabolic fate, as was reported forBACE1 (20). It will be interesting to define the determinants ofthe transmembrane domain that are critical for the subcellularlocalization of SKI-1 and to possibly separate those required forGolgi localization from the ones needed for endosomal traffick-ing. Finally, the detailed mechanism behind the possible func-tion of the St-1 and St-2 proteins on ATF6 processing is yet tobe unraveled.

Acknowledgments—We are indebted to Josee Hamelin, Marie-Claude Asselin, Jadwiga Marcinkiewicz, Ann Chamberland, andAndrew Chen for technical help; Antonella Pasquato and MonicaDettin for peptide synthesis and conjugation; and Nasha Nassoury,GaetanMayer, Eric Bergeron, and Annik Prat for constant and valu-able advice and help. We gratefully acknowledge the valuable gifts ofinhibitors PN-1 (Y. Peng Loh), Spn4.1 and its RRLL mutant (F. Jean),and �1-PDX (G. Thomas). The secretarial assistance of BrigitteMaryis greatly appreciated.

REFERENCES1. Seidah, N. G., and Chretien, M. (1999) Brain Res. 848, 45–622. Seidah, N. G., and Prat, A. (2002) Essays Biochem. 38, 79–943. Seidah, N. G., Khatib, A. M., and Prat, A. (2006) Biol. Chem. 387,

871–8774. Seidah, N. G., Mowla, S. J., Hamelin, J., Mamarbachi, A. M., Benjannet, S.,

Toure, B. B., Basak, A., Munzer, J. S., Marcinkiewicz, J., Zhong,M., Barale,J. C., Lazure, C., Murphy, R. A., Chretien, M., and Marcinkiewicz, M.(1999) Proc. Natl. Acad. Sci. U. S. A. 96, 1321–1326

5. Sakai, J., Rawson, R. B., Espenshade, P. J., Cheng, D., Seegmiller, A. C.,Goldstein, J. L., and Brown, M. S. (1998)Mol. Cell 2, 505–514

6. Pasquato, A., Pullikotil, P., Asselin, M. C., Vacatello, M., Paolillo, L.,Ghezzo, F., Basso, F., Di Bello, C., Dettin, M., and Seidah, N. G. (2006)J. Biol. Chem. 281, 23471–23481

7. Benjannet, S., Rhainds, D., Essalmani, R., Mayne, J., Wickham, L., Jin, W.,Asselin, M. C., Hamelin, J., Varret, M., Allard, D., Trillard, M., Abifadel,M., Tebon, A., Attie, A. D., Rader, D. J., Boileau, C., Brissette, L., Chretien,M., Prat, A., and Seidah, N. G. (2004) J. Biol. Chem. 279, 48865–48875

8. Rawson, R. B. (2003) Biochem. Soc. Symp. 70, 221–2319. Goldstein, J. L., DeBose-Boyd, R. A., and Brown, M. S. (2006) Cell 124,

35–4610. Haze, K., Yoshida, H., Yanagi, H., Yura, T., and Mori, K. (1999)Mol. Biol.

Cell 10, 3787–379911. Stirling, J., and O’Hare, P. (2006)Mol. Biol. Cell 17, 413–42612. Lenz, O., terMeulen, J., Klenk, H. D., Seidah, N. G., and Garten,W. (2001)

Proc. Natl. Acad. Sci. U. S. A. 98, 12701–1270513. Beyer, W. R., Popplau, D., Garten, W., Von Laer, D., and Lenz, O. (2003)

J. Virol. 77, 2866–287214. Vincent, M. J., Sanchez, A. J., Erickson, B. R., Basak, A., Chretien, M.,

Seidah, N. G., and Nichol, S. T. (2003) J. Virol. 77, 8640–864915. Elagoz, A., Benjannet, S., Mammarbassi, A., Wickham, L., and Seidah,

N. G. (2002) J. Biol. Chem. 277, 11265–1127516. Huovila, A. P., Turner, A. J., Pelto-Huikko, M., Karkkainen, I., and Ortiz,

R. M. (2005) Trends Biochem. Sci. 30, 413–42217. Rawson, R. B., Zelenski, N. G., Nijhawan, D., Ye, J., Sakai, J., Hasan,

M. T., Chang, T. Y., Brown, M. S., and Goldstein, J. L. (1997)Mol. Cell1, 47–57

18. Rawson, R. B., Cheng, D., Brown, M. S., and Goldstein, J. L. (1998) J. Biol.Chem. 273, 28261–28269

19. Benjannet, S., Savaria, D., Laslop, A.,Munzer, J. S., Chretien,M.,Marcink-iewicz, M., and Seidah, N. G. (1997) J. Biol. Chem. 272, 26210–26218

20. Benjannet, S., Elagoz, A., Wickham, L., Mamarbachi, M., Munzer, J. S.,Basak, A., Lazure, C., Cromlish, J. A., Sisodia, S., Checler, F., Chretien, M.,and Seidah, N. G. (2001) J. Biol. Chem. 276, 10879–10887

21. Okada, T., Haze, K., Nadanaka, S., Yoshida, H., Seidah, N. G., Hirano, Y.,Sato, R., Negishi,M., andMori, K. (2003) J. Biol. Chem. 278, 31024–31032

22. Nassoury, N., Blasiole, D. A., Tebon, O. A., Benjannet, S., Hamelin, J.,Poupon, V., McPherson, P. S., Attie, A. D., Prat, A., and Seidah, N. G.(2007) Traffic 8, 718–732

23. Nour, N., Mayer, G., Mort, J. S., Salvas, A., Mbikay, M., Morrison, C. J.,Overall, C. M., and Seidah, N. G. (2005)Mol. Biol. Cell 16, 5215–5226

24. Pullikotil, P., Vincent, M., Nichol, S. T., and Seidah, N. G. (2004) J. Biol.Chem. 279, 17338–17347

25. Lippincott-Schwartz, J., Yuan, L., Tipper, C., Amherdt, M., Orci, L., andKlausner, R. D. (1991) Cell 67, 601–616

26. Kim, T., and Loh, Y. P. (2006)Mol. Biol. Cell 17, 789–79827. Richer, M. J., Keays, C. A., Waterhouse, J., Minhas, J., Hashimoto, C., and

Jean, F. (2004) Proc. Natl. Acad. Sci. U. S. A. 101, 10560–1056528. Anderson, E. D., Thomas, L., Hayflick, J. S., and Thomas, G. (1993) J. Biol.

Chem. 268, 24887–2489129. Ye, S. Q., Reardon, C. A., and Getz, G. S. (1993) J. Biol. Chem. 268,

8497–850230. Weihofen, A., Lemberg, M. K., Ploegh, H. L., Bogyo, M., andMartoglio, B.

(2000) J. Biol. Chem. 275, 30951–3095631. Fishman, P. H., and Curran, P. K. (1992) FEBS Lett. 314, 371–37432. Bradshaw, R. A., and Yi, E. (2002) Essays Biochem. 38, 65–7833. Addlagatta, A., Hu, X., Liu, J. O., andMatthews, B.W. (2005)Biochemistry

44, 14741–1474934. Kozak, M. (2006) Gene (Amst.) 382, 1–1135. DeBose-Boyd, R. A., Brown,M. S., Li, W. P., Nohturfft, A., Goldstein, J. L.,

and Espenshade, P. J. (1999) Cell 99, 703–71236. Thomas, G. (2002) Nat. Rev. Mol. Cell Biol. 3, 753–76637. Brown, M. S., Ye, J., Rawson, R. B., and Goldstein, J. L. (2000) Cell 100,

391–39838. Ye, J., Rawson, R. B., Komuro, R., Chen, X., Dave, U. P., Prywes, R., Brown,

M. S., and Goldstein, J. L. (2000)Mol. Cell 6, 1355–136439. Raggo, C., Rapin, N., Stirling, J., Gobeil, P., Smith-Windsor, E., O’Hare, P.,

and Misra, V. (2002)Mol. Cell. Biol. 22, 5639–564940. Sanchez, A. J., Vincent, M. J., Erickson, B. R., and Nichol, S. T. (2006)

J. Virol. 80, 514–52541. Espenshade, P. J., Cheng, D., Goldstein, J. L., and Brown, M. S. (1999)

J. Biol. Chem. 274, 22795–2280442. Toure, B. B., Munzer, J. S., Basak, A., Benjannet, S., Rochemont, J.,

Lazure, C., Chretien, M., and Seidah, N. G. (2000) J. Biol. Chem. 275,2349–2358

43. Basak, A. (2005) J. Mol. Med. 83, 844–85544. Wouters, S., Decroly, E., Vandenbranden,M., Shober,D., Fuchs, R.,Morel,

V., Leruth, M., Seidah, N. G., Courtoy, P. J., and Ruysschaert, J. M. (1999)FEBS Lett. 456, 97–102

45. De Bie, I., Marcinkiewicz, M., Malide, D., Lazure, C., Nakayama, K., Ben-

Alternate Transcript of SKI-1/S1P

27412 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 282 • NUMBER 37 • SEPTEMBER 14, 2007

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from

dayan, M., and Seidah, N. G. (1996) J. Cell Biol. 135, 1261–127546. Xiang, Y., Molloy, S. S., Thomas, L., and Thomas, G. (2000)Mol. Biol. Cell

11, 1257–127347. Blobel, C. P. (2000) Curr. Opin. Cell Biol. 12, 606–61248. Blobel, C. P. (2005) Nat. Rev. Mol. Cell Biol. 6, 32–4349. Sachs, A. B., Sarnow, P., and Hentze, M. W. (1997) Cell 89, 831–83850. Kozak, M. (1997) EMBO J. 16, 2482–2492

51. Kozak, M. (2001) Sci. STKE 2001, E152. Seidah, N. G. (2002) in The Enzymes (Dalbey, R. E., and Sigman, D. S., eds)

Vol. 22, pp. 237–258, Academic Press, Inc., San Diego, CA53. Yan, R., Han, P., Miao, H., Greengard, P., and Xu, H. (2001) J. Biol. Chem.

276, 36788–3679654. Yan, R., Munzner, J. B., Shuck, M. E., and Bienkowski, M. J. (2001) J. Biol.

Chem. 276, 34019–34027

Alternate Transcript of SKI-1/S1P

SEPTEMBER 14, 2007 • VOLUME 282 • NUMBER 37 JOURNAL OF BIOLOGICAL CHEMISTRY 27413

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from

Philomena Pullikotil, Suzanne Benjannet, Janice Mayne and Nabil G. SeidahSUBCELLULAR LOCALIZATION

The Proprotein Convertase SKI-1/S1P: ALTERNATE TRANSLATION AND

doi: 10.1074/jbc.M703200200 originally published online July 10, 20072007, 282:27402-27413.J. Biol. Chem. 

  10.1074/jbc.M703200200Access the most updated version of this article at doi:

 Alerts:

  When a correction for this article is posted• 

When this article is cited• 

to choose from all of JBC's e-mail alertsClick here

  http://www.jbc.org/content/282/37/27402.full.html#ref-list-1

This article cites 52 references, 30 of which can be accessed free at

by guest on April 13, 2016

http://ww

w.jbc.org/

Dow

nloaded from