Molecular Cell, Vol. 2, 505–514, October, 1998, Copyright 1998 by Cell Press

Molecular Identification of the Sterol-RegulatedLuminal Protease that Cleaves SREBPsand Controls Lipid Composition of Animal Cells

S1P is the target of feedback regulation: its activity isextinguished in sterol-overloaded cells. So far, nothingis known about the structure or properties of S1P.

SREBPs are bound to the ER membrane and nuclearenvelope in a hairpin orientation. The NH2-terminal and

Juro Sakai, Robert B. Rawson,Peter J. Espenshade, Dong Cheng,Adam C. Seegmiller, Joseph L. Goldstein,*and Michael S. Brown*Department of Molecular GeneticsUniversity of Texas COOH-terminal domains, each about 500 amino acids

in length, project into the cytosol. They are linked by aSouthwestern Medical CenterDallas, Texas 75235 pair of membrane-spanning sequences that flank a short

31-amino-acid hydrophilic loop that projects into thelumen of the ER and nuclear envelope (Brown andGoldstein, 1997). When cells are depleted of sterols,SummaryS1P cleaves the SREBPs at a leucine-serine bond in theluminal loop, thereby separating the proteins into halves,The lipid composition of animal cells is controlled byeach with a single membrane-spanning sequence (Dun-SREBPs, transcription factors released from mem-can et al., 1997). The NH2-terminal half is called thebranes by sterol-regulated proteolysis. Release is initi-“intermediate” form of SREBP. Next, a second protease,ated by Site-1 protease (S1P), which cleaves SREBPsdesignated Site-2 protease (S2P), cleaves the NH2-ter-in the ER luminal loop between two membrane-span-minal intermediate at a leucine-cysteine bond that isning regions. To clone S1P, we prepared pCMV-PLAP-located just within the first membrane-spanning seg-BP2, which encodes a fusion protein that containsment (Duncan et al., 1998). This liberates the NH2-termi-placental alkaline phosphatase (PLAP) in the ER lumennal fragment, which dissociates from the membrane withflanked by cleavage sites for signal peptidase and S1P.three hydrophobic residues at its COOH terminus. ThisIn sterol-deprived cells, cleavage by both proteasesfragment, designated nuclear SREBP (nSREBP), entersleads to PLAP secretion. PLAP is not secreted by SRD-the nucleus and activates gene transcription.12B cells, cholesterol auxotrophs that lack S1P. We

The Site-1 cleavage reaction requires the participationtransfected SRD-12B cells with pCMV-PLAP-BP2 plusof a membrane-bound regulatory protein designatedpools of CHO cDNAs and identified a cDNA that re-SREBP cleavage-activating protein (SCAP) (Hua et al.,stores Site-1 cleavage and PLAP secretion. The cDNA1996a). SCAP has two domains: a hydrophobic NH2-encodes S1P, an intraluminal 1052-amino-acid mem-terminal membrane domain consisting of eight mem-brane-bound subtilisin-like protease. We propose that

S1P is the sterol-regulated protease that controls lipid brane-spanning sequences and a hydrophilic COOH-metabolism in animal cells. terminal domain containing five “WD” repeats that

projects into the cytosol (Nohturfft et al., 1998). TheCOOH-terminal domain of SCAP forms a tight complex

Introduction with the COOH-terminal domain of the SREBPs (Sakaiet al., 1997, 1998). Disruption of this complex by overex-

Cholesterol and fatty acids are the hydrophobic building pression of truncated dominant negative versions ofblocks of cell membranes. Their synthesis and uptake SCAP or SREBP blocks Site-1 cleavage of SREBPs,must be coordinated so as to supply sufficient amounts indicating that the SCAP/SREBP complex is absolutelyfor new membrane synthesis while avoiding overaccu- required for cleavage. Moreover, truncated versions ofmulation. Coordination is achieved by a family of tran- SREBPs, which lack the COOH-terminal domain, fail toscription factors designated sterol regulatory element form complexes with SCAP and fail to undergo Site-1binding proteins (SREBPs) that are bound to membranes cleavage (Sakai et al., 1998).of the endoplasmic reticulum (ER) and nuclear envelope Although the SCAP/SREBP complex is created by(Brown and Goldstein, 1997). When cells are deprived interactions on the cytosolic side of the membrane, theof sterols, a two-step proteolytic process releases the complex activates S1P, which cuts the SREBPs on theactive portions of the SREBPs from cell membranes, opposite (luminal) side (Sakai et al., 1998). The proteaseallowing them to translocate to the nucleus, where they cuts between the leucine and serine of the sequenceactivate transcription of more than a dozen genes en- RSVLS. Recognition requires only the arginine and leu-coding enzymes required for biosynthesis and uptake cine: the other residues can be replaced with alaninesof cholesterol and unsaturated fatty acids. When sterols without reducing cleavage (Duncan et al., 1997).build up in cells, the proteolytic release process is The Site-1 processing reaction is the target for feed-blocked, the SREBPs remain membrane-bound, and back regulation of lipid biosynthesis and uptake in ani-transcription of the target genes declines. A crucial com- mal cells. When sterols accumulate in cells, the Site-1ponent in this regulatory pathway is the Site-1 protease cleavage reaction is blocked (Brown and Goldstein,(S1P), which makes the first cut in the SREBPs, thereby 1997). The Site-2 cleavage reaction is blocked second-initiating release (Sakai et al., 1996; Duncan et al., 1997). arily since it requires prior cleavage by S1P. The sterol

effect appears to be mediated by five of the eight mem-brane-spanning sequences of SCAP, which are desig-* To whom correspondence should be addressed (e-mail: jgolds@

mednet.swmed.edu [J. G.], mbrow1@mednet.swmed.edu [M. B.]). nated as the sterol sensor (Hua et al., 1996a). Point

Molecular Cell506

mutations at two positions within the sterol sensor ren-der SCAP constitutively active and prevent sterol-medi-ated suppression of Site-1 cleavage (Nohturfft et al.,1996). Sequences that resemble the sterol-sensing do-main are found in three other proteins that are postulatedto interact with sterols (Loftus et al., 1997; Nohturfft etal., 1998).

Our laboratory recently cloned the human gene forS2P by complementation of the growth defect in a mu-tant line of Chinese hamster ovary (CHO) cells that failsto synthesize cholesterol owing to a deletion of the S2Pgene (Rawson et al., 1997). This gene encodes a uniquehydrophobic zinc metalloprotease that cleaves the inter-mediate forms of SREBPs within their transmembranesequences.

A similar approach to complementation cloning of S1Phas been unavailable up to now because we were unableto isolate a mutant cell line that fails to carry out Site-1cleavage. We attributed this to the fact that S2P is en-coded by a single-copy gene on the X chromosome ofCHO cells (Rawson et al., 1997). Whenever we mutagen-ized CHO cells and selected for cholesterol auxotrophyand deficiency of low density lipoprotein (LDL) recep-tors, we always obtained mutants of S2P. Recently, wecircumvented this problem by transfecting CHO cells Figure 1. Proteolytic Processing and Secretion of the PLAP-

BP2(513-1141) Fusion Proteinwith a cDNA encoding S2P, thereby assuring multiplecopies of the gene (Rawson et al., 1998). We mutagen- The plasmid was generated by fusing the sequence encoding the

signal peptide and soluble catalytic domain of human placentalized the cells and selected for the very rare cell thatalkaline phosphatase (amino acids 1–506) with the sequence encod-failed to synthesize cholesterol or take up LDL. Thising amino acids 513–1141 of human SREBP-2. The one novel aminoprocedure yielded two cell lines, designated SRD-12Aacid (Y) between the two proteins was generated by blunt ligationand SRD-12B, which failed to carry out Site-1 cleavage.during the construction of the plasmid. Secretion of the catalytic

Cell fusion studies indicated that the defects in both domain of PLAP requires cleavage by signal peptidase and Site-1cell lines were recessive and noncomplementing. We protease.interpreted these findings to indicate that the SRD-12Aand SRD-12B cells had undergone mutations in bothcopies of the gene encoding S1P or another protein

of amino acids 1–506 of human placental alkaline phos-required for S1P function (Rawson et al., 1998).

phatase (PLAP) beginning with the signal peptide. TheIn the current studies, we used the SRD-12B cellsprotein terminates at residue 506, which precedes theas recipients for complementation cloning of the genehydrophobic sequence that normally serves as the sig-encoding S1P. For this purpose, we created an expres-nal for attachment of a glycophospholipid anchor to thesion vector that encodes a fusion protein in which thealkaline phosphatase (Cullen and Malim, 1992). If thissignal peptide and catalytic domain of placental alkalineprotein were ever to be released into the lumen of thephosphatase (PLAP) are joined to the luminal loop ofER, it is predicted to be soluble. The PLAP is fused toSREBP-2 on the NH2-terminal side of Site-1 (referred toSREBP-2 at amino acid position 513, which is six resi-as PLAP-BP2). The catalytic domain of PLAP is thereforedues to the NH2-terminal side of the RSVLS sequencelocated in the ER lumen, where it is flanked by the sitesthat is cleaved by S1P. The SREBP-2 sequence includesfor signal peptidase and for S1P. When transfected withthe second transmembrane domain and the entirethis plasmid, wild-type CHO cells cleave the protein atCOOH-terminal regulatory domain that projects into thethe two flanking sites, and PLAP is secreted into thecytosol. PLAP-BP2 is predicted to be cleaved cotransla-medium where its activity can be readily assayed. SRD-tionally by signal peptidase. The PLAP should remain12B cells, lacking S1P, fail to secrete PLAP. We tran-membrane-bound unless it is cleaved from the SREBP-2siently transfected the SRD-12B cells with the cDNAby S1P, whereupon it should be secreted into the cultureencoding the PLAP-BP2 fusion protein, a cDNA encod-medium. The bottom of Figure 1 shows the amino aciding SCAP, and pools of cDNAs from a CHO cell expres-sequences surrounding the two cleavage sites.sion library. We identified pools of cDNAs that restored

To demonstrate that PLAP-BP2 would be cleaved insecretion of PLAP and used the assay to purify a cDNAthe predicted fashion, we introduced the PLAP-BP2that encodes the responsible protein. The protein is acDNA into human kidney 293 cells by transfection undernovel ER-localized serine protease of the subtilisin fam-control of the CMV promoter (Figure 2A). We alsoily, which we name S1P.transfected increasing amounts of a plasmid encodingwild-type SCAP, also under the control of the CMV pro-Resultsmoter. We measured alkaline phosphatase activity inthe medium by a sensitive chemiluminescence assayFigure 1 shows the fusion protein that we developed to

monitor the activity of S1P. The NH2 terminus consists after inactivating nonplacental alkaline phosphatase by

cDNA Cloning of Site-1 Protease for SREBPs507

Figure 2. SCAP-Stimulated Secretion of Al-kaline Phosphatase in Cells Transfected withpCMV-PLAP-BP2(513-1141)

(A) Duplicate 60 mm dishes of 293 cells weretransfected with 2 mg of pCMV-PLAP-BP2(513-1141), 0.01 mg of pCMVb-gal, and theindicated amount of pCMV-SCAP. The totalamount of DNA was adjusted to 6 mg/dishby addition of pcDNA 3 empty vector. Aftertransfection, the cells were incubated for 16hr in medium D in the absence (closed circle)or presence (open circle) of sterols as de-scribed in Experimental Procedures.(B) Duplicate 60 mm dishes of 293 cells weretransfected with 2 mg of the wild-type (closedcircle) or the indicated mutant (triangle, open

circle) version of pCMV-PLAP-BP2(513-1141), 0.01 mg of pCMVb-gal, and the indicated amount of pCMV-SCAP. After transfection, the cellswere incubated for 16 hr in medium D in the absence of sterols.(C) Duplicate 22 mm wells of CHO-7 cells (open circle) or SRD-12B cells (closed circle) were transfected with 0.1 mg of pCMV-PLAP-BP2(513-1141), 0.01 mg of pCMVb-gal, and the indicated amount of pCMV-SCAP. The total amount of DNA was adjusted to 1.26 mg/well by additionof pcDNA 3 empty vector. After transfection, the cells were cultured in medium A supplemented with fetal calf lipoprotein-deficient serum inthe absence of sterols.(A–C) After incubation for 16 hr, aliquots of medium were removed and assayed for placental alkaline phosphatase activity. The data werenormalized to cellular b-galactosidase activity as described in Experimental Procedures. Each value is the average of duplicate transfections.

heat treatment (Cullen and Malim, 1992). In the absence described in Experimental Procedures. Four of thesepools gave positive results. We then assayed individualof cotransfected SCAP, virtually no PLAP was secreted

into the medium (Figure 2A). The amount of secreted clones from the wells at the intersections of the positiverows and columns. As expected, two of these were posi-PLAP activity rose linearly as the amount of transfected

pCMV-SCAP was increased. When the 293 cells were tive. Restriction maps suggested that these clones hadidentical inserts, and we chose cDNA clone p38 for fur-incubated in the presence of a mixture of cholesterol

plus 25-hydroxycholesterol, PLAP secretion was reduced. ther studies.cDNA clone p38 contains a large open reading frameThe requirement for SCAP and the suppression by ste-

rols indicates that PLAP secretion depends upon S1P. that encodes a protein of 1052 amino acids with a calcu-lated molecular mass of 117.5 kDa (Figure 3). The initialTo confirm the role of signal peptidase and S1P in

PLAP secretion, we prepared constructs encoding two 17 amino acids are hydrophobic and appear to representa signal peptide. The hydrophobic sequence terminatesmutant versions of the fusion protein. In one mutant,

the arginine of the RSVLS sequence at Site-1 was at a glycine, which is an ideal substrate for signal pepti-dase (von Heijne, 1985). Hereafter, we refer to the proteinchanged to alanine (R519A). Previous studies have

shown that this mutation decreases the susceptibility encoded by p38 as S1P.Database searches revealed that S1P has featuresof SREBP-2 to cleavage by S1P (Duncan et al., 1997).

In the second mutant, two amino acids near the signal characteristic of a superfamily of serine proteases, broadlyclassified as subtilisins, which are found in all livingpeptidase cleavage site were changed (S15F/G17I).

These substitutions are predicted to block cleavage by organisms from bacteria to humans (Siezen and Leunis-sen, 1997). The human counterpart of hamster S1P wassignal peptidase (van Heijne, 1985). The secretion of

PLAP was markedly reduced when the fusion protein sequenced previously from a random library of se-quences expressed in KG1 cells, an immature myeloidbore either the signal peptidase mutations or the Site-1

cleavage mutation (Figure 2B). cell line (Nagase et al., 1995). The cDNA sequence wasdesignated KIAA0091 (GenBank Accession No. D42053).As a final test for the requirement for S1P in the secre-

tion of PLAP, we introduced the construct into wild-type Although the function of the KIAA0091 gene productwas not known, it was classified as a member of theCHO cells and into SRD-12B cells that lack S1P activity

(Rawson et al., 1998) (Figure 2C). Whereas the wild-type subtilisin family, and its sequence has been includedin a comparative analysis of this family by Siezen andcells secreted PLAP into the medium, the SRD-12B cells

were almost totally deficient in this secretion. Leunissen (1997), who referred to the KIAA0091 se-quence as “hskiaa.” Overall, the amino acid sequenceThe secretion of PLAP was then used as an assay for

the cloning of S1P by transient complementation of the of hamster S1P is 97% identical to the sequence ofhuman hskiaa. Nearly all of the substitutions are conser-defect in SRD-12B cells as described in Experimental

Procedures. We studied a total of 300 pools of 1000 vative (Figure 3).Subtilisins, like other serine proteases, contain a cata-cDNAs each. Two pools stimulated PLAP secretion to

levels that were significantly higher than background. lytic triad consisting of an aspartic acid, a histidine, anda serine residue. Based on the resemblance to otherPool 116 was divided into subpools of 100 colonies each.

Two of these subpools gave positive results. We then members of the subtilisin family, the sequence of S1Ppredicts that its catalytic triad should consist of Asp218,plated individual colonies from subpool 116–8 into indi-

vidual wells of an 11 3 11 matrix. We prepared cDNAs His249, and Ser414 (Siezen and Leunissen, 1997).Other notable features of the S1P sequence includefrom the pooled rows and columns of the matrix as

Molecular Cell508

nonpolar residues, which is consistent with a mem-brane-spanning sequence (thick overline in Figure 3).This is followed by a sequence of 30 amino acids thatis strikingly rich in prolines and basic residues (6 prolinesand 11 basic residues without a single acidic residue).

Figure 4A shows a hydropathy plot of S1P, whichindicates that the protein is generally hydrophilic withthe exception of the signal sequence at the NH2 terminusand the apparent transmembrane segment near theCOOH terminus. To determine whether the protein ismembrane-bound, we prepared an expression vectorencoding S1P with three copies of a c-Myc tag insertedimmediately following the signal sequence (denoted bythe arrow in Figure 3). The c-Myc tag is expected toremain associated with S1P following cleavage by signalpeptidase. We transfected this cDNA into 293 cells andfractionated the cells into a nuclear extract, a membranefraction, and cytosol, all of which were subjected to SDSpolyacrylamide gel electrophoresis. All of the immuno-detectable S1P was found in the membrane fraction(Figure 4B, lane 5). The antibody also reacted with aprotein of z62 kDa that was present in the nuclear ex-tract of the mock-transfected cells. We believe that thisis the endogenous c-Myc protein of these cells. Figure4C shows that S1P is sensitive to treatment with endo-glycosidase H, confirming that the protein containsN-linked sugars and is located in the lumen of the ER(Kornfeld and Kornfeld, 1985).

To determine whether the SRD-12B cells produce anmRNA encoding S1P, we performed a Northern blotanalysis (Figure 5A). The S1P probe hybridized to a sin-gle mRNA of 4.4 kb in wild-type CHO-7 cells. This mRNAwas also present in the mutant M19 cells, which lackS2P. The mRNA was not detectable in either the SRD-12A or SRD-12B cells. When expression of S1P mRNAwas studied in Multiple Tissue Northern Blots (Clontech),a single band of z4.4 kb was detected in all 15 humantissues that were studied, including liver, brain, and ad-renal cortex.

Figure 5B shows a series of Southern blots performedwith genomic DNA from the parental CHO/pS2P cellsand the two S1P-deficient mutant cell lines. The DNAwas digested with either of two restriction enzymes andprobed with radiolabeled probes corresponding to the59 half of the S1P cDNA or the 39 half. The EcoRI digestof the SRD-12A DNA revealed shifted bands when ana-lyzed with the 59 probe (lane 2) and the 39 probe (laneFigure 3. Amino Acid Sequence of Hamster and Human S1P8). In addition, the intensity of all of the bands wasAmino acid residue numbers are shown on the right. Identical resi-reduced by 50% when compared with the DNA from thedues are highlighted in black. The putative signal sequence is de-other cell lines as revealed by densitometric scanning.noted by the dotted overline. The putative membrane-spanning se-

quence is denoted by the solid overbar. Potential N-linked We interpret this finding to indicate that the SRD-12Aglycosylation sites are denoted by three dots. The sequence corre- cells have a deletion of one copy of the S1P gene andsponding to the catalytic triad for subtilisin-like serine proteases a rearrangement of the second copy. The DNA from(Siezen and Leunissen, 1997) is denoted by asterisks. The vertical the SRD-12B cells showed abnormalities in the BamHIarrow denotes the site of insertion of three tandem copies of the

digest (lanes 6 and 12). The intensities of the other bandsc-Myc epitope in the expression vector pCMV-Myc-S1P. GenBankwere similar to those in wild-type cells. We concludeaccession numbers for hamster and human S1P are AF078105 andthat the SRD-12B cells have undergone a rearrangementD42053, respectively.in at least one copy of the S1P gene. The other copymay be rearranged, or it may have a more subtle muta-

six potential sites of N-linked glycosylation (Asn-X-Ser/ tion that abolishes the production of mRNA.Thr) (X can be any amino acid other than proline; see To demonstrate directly that S1P restores cleavageMarshall, 1972) (designated by 3 dots in Figure 3). Near of SREBPs in SRD-12B cells, we transfected the cells

with pCMV-S1P, a vector in which the expression ofthe COOH terminus, there is an unbroken stretch of 25

cDNA Cloning of Site-1 Protease for SREBPs509

Figure 4. Hydropathy Plot and Cellular Localization of Hamster S1P

(A) The residue-specific hydropathy index was calculated over a window of 20 residues by the method of Kyte and Doolittle (1982), using theGenetics Computer Group Sequence Analysis Software Package, Version 8.1 (Devereux et al., 1984). Arrows denote the three amino acidsthat correspond to the catalytic triad for subtilisin-like serine proteases. SS, signal sequence; TM, transmembrane region.(B) 293 cells were transfected with 1 mg/dish of either pcDNA3 empty vector (lanes 1–3) or pCMV-Myc-S1P (lanes 4–6) as described inExperimental Procedures. The total amount of transfected DNA was adjusted to 5 mg/dish by addition of pcDNA3 empty vector. Transfectedcells were incubated in medium C supplemented with 10% fetal calf serum for 16 hr and fractionated into nuclear extract, membrane, andcytosol fractions as described (Sakai et al., 1996). Aliquots of protein from transfected cells (40 mg, lanes 1 and 4; 60 mg, lanes 2, 3, 5, and6) were subjected to SDS-PAGE and immunoblot analysis with 2 mg/ml of monoclonal anti-c-Myc (clone 9E10) as described in ExperimentalProcedures. The filter was exposed to film for 2 sec. The asterisk denotes the endogenous c-Myc protein.(C) Aliquots of the 105 g membrane fraction from 293 cells transfected with pCMV-Myc-S1P, prepared as in (B), were treated with the followingglycosidases: lane 1, none; lane 2, 0.063 IU/ml peptide N-glycosidase F (PNGaseF); lane 3, 0.9 IU/ml endoglycosidase Hf; lane 4, none.Enzymatic digestion was performed as described by Sakai et al. (1998). Aliquots of treated membrane fractions (5 mg protein) were subjectedto SDS-PAGE and immunoblotted with 2.5 mg/ml of monoclonal anti-c-Myc (clone 9E10). The filter was exposed to film for 1 sec.

S1P is driven by the CMV promoter (Figure 6). As a sterols were added (lane 6). pCMV-S1P did not producea nuclear band when we transfected it together withreporter, we used pTK-HSV-BP2, which produces an

HSV-tagged version of SREBP-2 under control of the a cDNA encoding a mutant version of HSV-BP2 thatcontains the R519A substitution that blocks Site-1thymidine kinase promoter, which gives a near-phy-

siologic level of expression (Hua et al., 1996b). Cell proteolysis (lane 7). We also transfected a cDNA en-coding a mutant version of HSV-BP2 which has themembranes and nuclear extracts were subjected to

electrophoresis and blotted with an antibody against D478RSR→AS substitution that blocks cleavage by S2P.When this was transfected into the SRD-12B cells, wethe HSV epitope-tag. In the absence of transfected S1P,

the SRD-12B cells failed to cleave HSV-BP2, as indi- observed no intermediate form in the membranes andno nuclear form (lane 11), a finding that is consistentcated by the lack of a band in the nuclear extract (lane

3). Cotransfection of pCMV-S1P led to the appearance with the absence of Site-1 cleavage. When S1P wascotransfected together with the D478RSR→AS mutant,of the nuclear form (lane 5), which disappeared when

Figure 5. Molecular Characterization of S1PmRNA and Genomic DNA in Wild-Type andMutant Hamster Cells

(A) Northern blot anlaysis. On day 0, the indi-cated cell line was set up at a density of 5 3

105 cells/100 mm dish in medium B and wererefed the same medium on day 2. On day 3,total RNA was prepared, and poly(A)1 RNAwas purified. A 32P-labeled 4.2 kb SalI/NotIfragment derived from hamster S1P cDNA(1.6 3 106 cpm/ml) was hybridized to 3 mg ofpoly(A)1 RNA from the indicated cells for 3hr at 658C in Rapid-hyb buffer (Amersham),washed twice with 23 SSC and 0.1% SDS atroom temperature for 30 min each, followedby two washes with 0.13 SSC and 0.1% SDSat 708C for 30 min each. The nylon membrane

was exposed to Kodak X-Omat Blue XB-1 film with an intensifying screen at 2808C for 10 hr. The same membrane was subsequently hybridizedwith a 1.1 kb rat glyceraldehyde-3-phosphate dehydrogenase (GAPDH) probe (1 3 106 cpm/ml) (Chen et al., 1991) and exposed to film for 2hr at 2808C with an intensifying screen.(B) Southern blot analysis. Aliquots of genomic DNA (10 mg/lane) from the indicated cell line were digested with the indicated restrictionenzyme, subjected to electrophoresis on 0.7% agarose gels, transferred to filters, and hybridized with the indicated 32P-labeled probe (2 3

106 cpm/ml) as described in Experimental Procedures. The filter in lanes 1–6 was boiled in 0.5% SDS and rehybridized with a probe for thehamster ACAT cDNA (lanes 13–18). Filters were exposed to film with an intensifying screen at 2808C for 17 hr.

Molecular Cell510

Figure 6. Site-1 Protease cDNA Restores Sterol-Regulated Cleav-age of Epitope-Tagged SREBP-2 in Transfected SRD-12B Cells

SRD-12B cells were set up for experiments on day 0 and transfectedon day 1 as described in Experimental Procedures. The indicatedplasmid encoding wild-type or mutant HSV-tagged SREBP-2 (3.5mg/dish; lanes 3–14) was cotransfected into SRD-12B cells with orwithout pCMV-S1P (0.5 mg/dish) as indicated. The total amount ofDNA was adjusted to 4 mg/dish by addition of either pTK emptyvector (Hua et al., 1995) or pcDNA3 empty vector. Transfected cellswere incubated in either the absence (-) or presence (1) of sterols,and the cells were harvested and fractionated as described in Exper-imental Procedures. Aliquots of the nuclear extract (30 mg protein)and membrane (60 mg) fractions were subjected to SDS-PAGE andimmunoblotted with 0.5 mg/ml of mouse IgG-HSV-Tag as the firstantibody and 0.16 mg/ml of donkey anti-mouse IgG as the secondantibody. Filters were exposed to film for 20 sec (nuclear extracts),or 5 sec (membranes). P, I, and N denote the precursor, intermediate,and nuclear forms of HSV-tagged SREBP-2, respectively. Figure 7. Site-1 Cleavage of SREBP-2 in SRD-12B Cells Trans-

fected with Wild-Type or Mutant S1P cDNAs

(A) Secretion of alkaline phosphatase. SRD-12B cells (4 3 104 cells/22 mm well) were transfected with an expression plasmid containingwe noted a new band corresponding to the intermediateno cDNA insert or the indicated c-Myc tagged S1P cDNA (0.03form in the membrane fraction (lane 13). The amountmg/well) together with pCMV-PLAP-BP2 (513–1141) (0.1 mg/well),

of this band was reduced in the presence of sterols, pCMV-SCAP (0.56 mg/well), and pCMVb-gal (0.01 mg/well) as de-confirming that it was the product of Site-1 cleavage scribed in Experimental Procedures. After transfection, the cells(lane 14) (Sakai et al., 1996). were cultured in medium A supplemented with fetal calf lipoprotein-

deficient serum. After incubation for 16 hr, the medium was col-To determine whether the proposed catalytic triad islected, and placental alkaline phosphatase activity was measuredessential for S1P activity, we prepared versions of pCMV-and normalized to cellular b-galactosidase activity as described inMyc-S1P in which the codons for Asp218, His249, andExperimental Procedures. Each value is the average of duplicate

Ser414 were individually mutated to encode different transfections (range of values denoted by error bars).amino acids (D218N, H249F, and S414A). As controls, (B) Immunoblot analysis of epitope-tagged SREBP-2. SRD-12B cells

were set up on day 0 and transfected on day 1 as described inwe mutated two nearby serines that are not proposedExperimental Procedures. Plasmid pCMV-Myc-S1P encoding theto be part of the catalytic triad (S411A and S417A).indicated wild-type or mutant Myc-tagged S1P (0.5 mg/dish) wasEach plasmid was transfected into SRD-12B cells to-cotransfected with pTK-HSV-BP2 (3 mg/dish). The total amount of

gether with pCMV-PLAP-BP2(513-1141), pCMV-SCAP, DNA was adjusted to 3.5 mg/dish as described in Figure 6.and pCMVb-gal (Figure 7A). Whereas the cDNA encod- Transfected cells were incubated for 21 hr in either the absence (-)ing wild-type S1P increased PLAP activity in the culture or presence (1) of sterols as indicated and then harvested and

fractionated as described in Experimental Procedures. Aliquots ofmedium, each of the catalytic triad mutants failed. Thenuclear extract (30 mg protein) or membranes (60 mg) were subjectedcontrol substitutions (S411A and S417A) had activitiesto SDS-PAGE and immunoblotted with 0.5 mg/ml of IgG-HSV-Tagthat were 50%–70% of wild-type levels. or 2 mg/ml of monoclonal anti-c-Myc (clone 9E10) as described in

In a more direct test for S1P catalytic activity, the Experimental Procedures. Filters were exposed to film for 1 min (topmutant plasmids were transfected into the SRD-12B panel), 20 sec (middle panel), or 4 sec (bottom panel). P and N

denote the precursor and nuclear forms of HSV-tagged SREBP-2,cells together with pTK-HSV-BP2. Membranes and nu-respectively.clear extracts were subjected to SDS-PAGE and blotted

with an antibody against the HSV-tagged SREBP-2 (Fig-ure 7B). Wild-type S1P restored cleavage of SREBP-2, sterols (lanes 7–12). In contrast, the control S417A mu-and a band of the appropriate size was detected in the tant had sterol-regulated activity that was similar to wild-nuclear extract (Figure 7B, lane 5). This band disap- type (lanes 13 and 14). The membrane fractions were alsopeared when the cells were treated with sterols (lane subjected to immunoblotting, which revealed that all of6). The three catalytic triad mutants did not produce the transfected cells produced equivalent amounts of the

precursor form of HSV-BP2 and Myc-S1P (Figure 7B).nuclear SREBP-2 either in the absence or presence of

cDNA Cloning of Site-1 Protease for SREBPs511

to a failure to cleave SREBPs at Site-1 (Figures 5A and5B). Second, expression of S1P restores Site-1 cleavageof SREBP-1 and SREBP-2 in SRD-12B cells (Figures 6and 7 for SREBP-2; data not shown for SREBP-1), andthis is associated with a restoration of cholesterol-inde-pendent growth (Figure 8). Third, substitutions in anyone of the three residues that form the postulated cata-lytic triad of S1P abolish its ability to restore Site-1 cleav-age (Figure 7).

In cloning the cDNA for S1P, we employed a novelstrategy that took advantage of a mutant cell line witha deficiency of S1P activity. We also exploited the abilityof S1P to cleave its substrate in the lumen of the ER,thereby allowing the design of a fusion protein whosesecretion is dependent on Site-1 cleavage. We chosePLAP as a reporter for these experiments because itnormally functions in the extracellular environment andbecause a sensitive chemiluminescence assay is avail-able. Other reporter proteins should work equally well.Indeed, in preliminary experiments we employed bovineprolactin in a similar fusion construct with SREBP-2,and its secretion was also shown to be dependent onS1P (data not shown). These fusion proteins provide aconvenient way in which to monitor the activity of S1Pwithout having to disrupt the cells. These reportersshould be useful for studies of S1P regulation in cellculture and also in the bodies of living animals.

Figure 8. S1P cDNA Restores Growth of SRD-12A and SRD-12B Studies of the cleavage of PLAP-BP2 provide informa-Cells in Absence of Cholesterol tion about the general properties of the Site-1 cleavageOn day 0, cells were set up at 3.5 3 105 cells/60 mm dish in medium reaction. Secretion of PLAP requires two cleavages. TheB. On day 1, the cells were transfected with 5 mg/dish of either first occurs at the normal site of cleavage by signalpcDNA 3.1 empty vector or pCMV-S1P using the MBS Transfection peptidase, and the second occurs at Site-1 (Figure 1).Kit as previously described (Rawson et al., 1997). On day 2, cells

Cleavage by signal peptidase normally occurs cotrans-were washed with PBS and refed medium A supplemented with 5%lationally (Blobel, 1980), and it is therefore highly likely(v/v) fetal calf lipoprotein-deficient serum. Cells were refed every 2that this cleavage precedes cleavage at Site-1. Thus,days thereafter. On day 15, cells were washed with PBS, fixed in

95% ethanol, and stained with crystal violet. S1P is able to cleave a protein that has a free NH2

terminus in the lumen of the ER: it does not require thatthe NH2 terminus be anchored to the membrane, as it

Figure 8 shows a growth experiment that tested theis in native SREBP.

ability of different cell lines to grow in the absence of The conserved sequence motifs surrounding the cata-cholesterol, either with or without transfection of pCMV- lytic triad allow S1P to be classified unambiguously asS1P. The parental CHO/pS2P cells grew in the absence a member of the subtilisin superfamily (Siezen and Leu-of cholesterol, and there was no change when they were nissen, 1997). Outside of the catalytic domain, the se-transfected with the cDNA encoding pCMV-S1P. On the quence of S1P bears little relation to the other membersother hand, the SRD-12A and SRD-12B cells failed to of this superfamily. From a functional viewpoint, thegrow in the absence of cholesterol. Growth was restored action of S1P most resembles that of the furins, whichwhen these cells were permanently transfected with are subtilases of the Kex2p-like subfamily that processpCMV-S1P. This result formally demonstrates that res- proteins such as the insulin pro-receptor and pro-endo-toration of S1P activity can restore growth of the mutant thelin-1 prior to secretion in animal cells (Nakayama,cells in the absence of cholesterol. 1997). S1P differs substantially from the furins in two

respects: (1) substrate recognition—furins always cleaveDiscussion after dibasic sequences, most often with the consensus

sequence RX(K/R)R. S1P cleaves after RSVL (DuncanThe current results add a central piece to the puzzle of et al., 1997); and (2) cellular location—furins act in post-sterol-regulated processing of SREBPs. The data reveal Golgi secretory vesicles, whereas S1P acts in a pre-that Site-1 cleavage is catalyzed by a novel membrane- Golgi compartment, most likely the ER. All membersbound glycoprotein that belongs to the subtilisin super- of the Kex2p-like family and most other subtilases arefamily of serine proteases that are sometimes referred synthesized as pre-proteins that must be cleaved pro-to as subtilases (Siezen and Leunissen, 1997). The fol- teolytically to be active. Cleavage removes an NH2-ter-lowing three lines of evidence indicate that the cloned minal prepro-peptide that inhibits activity. As of yet, weS1P is the enzyme that normally cleaves SREBPs at do not know whether S1P contains a prepro-peptide.Site-1: first, rearrangements in the S1P gene that abolish The deduced amino acid sequence of S1P and itsS1P mRNA expression are found in SRD-12A and SRD- sensitivity to endoglycosidase H suggests that S1P to-

pologically is a Type I integral membrane protein with12B cells, which are auxotrophic for cholesterol owing

Molecular Cell512

Construction of pCMV-PLAP-BP2(513-1141)a large protease domain of z1020 amino acids thatThe expression vector pCMV-PLAP-BP2(513-1141) encodes anresides in the lumen of the ER, followed by a single1136-amino acid fusion protein consisting of an initiator methioninemembrane-spanning domain and a short cytoplasmicfollowed by the secreted form of human placental alkaline phospha-

COOH-terminal tail. If S1P does function in the ER, its tase (amino acids 2–506), one novel amino acid (Y) generated byactivity must be tightly regulated so as to prevent it from blunt ligation, and the COOH-terminal half of human SREBP-2

(amino acids 513–1141) (Figure 1). pCMV-PLAP-BP2(513-1141) wasdegrading nascent polypeptides nonspecifically. This isconstructed as follows. First, pCMV/SEAP (purchased from Tropix)presumably the function of SCAP, which directs S1P toencoding the secreted form of human placental alkaline phospha-SREBPs (Sakai et al., 1998).tase (Cullen and Malim, 1992) was cleaved with HpaI and XbaI toWe do not yet know the function of the unusual COOH-remove the stop codon and the 39 untranslated sequence. We iso-

terminal cytoplasmic domain of S1P. This 30-residue lated the 7.5 kb fragment of the vector containing the CMV promoter/sequence has a highly unusual composition that is rich enhancer region and the coding region of human placental alkaline

phosphatase (amino acids 1–506). Second, pTK-HSV-BP2 (Hua etin basic residues and prolines. Its sequence is sugges-al., 1996b) was mutagenized by oligonucleotide site-directed muta-tive of a b-sheet that has a high density of positivegenesis (Kunkel et al., 1987; Hua et al., 1996b) to generate an EcoRVcharges on one surface. This cytoplasmic tail mightrestriction site at the codons encoding amino acids 511 and 512 offunction in determining the subcellular localization ofSREBP-2. This intermediate construct was digested with EcoRV

S1P. It might also play a role in the interaction with the and XbaI to isolate a 2.6 kb fragment encoding amino acids 513 toWD-repeat domain of SCAP, which is required for S1P 1141 of SREBP-2. The 2.6 kb EcoRV-XbaI fragment was blunt-

ligated inframe with the HpaI/XbaI-digested pCMV/SEAP vector,activity (Sakai et al., 1998). Further studies are requiredintroducing a single tyrosine residue at the boundary between theto define the mechanism by which SCAP activates S1Palkaline phosphatase and SREBP-2 fusion protein to yield pCMV-and renders it sensitive to inhibitions by sterols.PLAP-BP2(513-1141). The S15F/G17I and R519A mutant versionsof this plasmid were constructed by site-directed mutagenesis (Kun-

Experimental Procedures kel et al., 1987; Hua et al., 1996b). The final structures of all plasmidswere confirmed by sequencing all ligation junctions.

General Methods and MaterialsThe Phospha-Light Reporter Gene Assay System and Luminescent

cDNA Transfectionb-Galactosidase Genetic Reporter System II were purchased fromOn day 0, 293 cells were set up at a density of 4 3 105 cells/60 mmTropix and Clontech, respectively. Total cellular RNA was obtaineddish in medium C supplemented with 10% fetal calf serum. Onwith RNA-DNA STAT 60 (Tel-Test, Inc.), and poly(A)1 RNA was iso-day 2, duplicate dishes of cells were transfected with the indicatedlated with an mRNA Purification Kit (Pharmacia Biotech). cDNAplasmids by the calcium phosphate method using the MBS Trans-probes were radiolabeled by random priming with [a-32P]dCTP usingfection Kit (Stratagene) as described (Hua et al., 1995). After incuba-the Megaprime DNA Labeling System (Amersham). Plasmid pCMVb-tion for 3 hr, the cells were washed once with PBS and fed with 5gal, encoding a b-galactosidase reporter driven by the CMV pro-ml of medium D (medium C supplemented with 10% newborn calfmoter/enhancer, was obtained from Stratagene. Plasmids pcDNAlipoprotein-deficient serum, 50 mM compactin, and 50 mM sodium3 and 3.1, which are empty vectors containing the CMV promoter/mevalonate) in the absence or presence of sterols (1 mg/ml of 25-enhancer, were obtained from Invitrogen. We obtained monoclonalhydroxycholesterol plus 10 mg/ml of cholesterol added in ethanolantibody HSV-Tag (IgG1) from Novagen, monoclonal antibody anti-at a final concentration of 0.2%). After incubation for 16 hr, aliquotsc-Myc (clone 9E10) (IgG1) from Boehringer Mannheim, a polyclonalof 1 ml of medium were removed for assay of alkaline phosphataseaffinity-purified donkey anti-mouse IgG from Jackson Immunore-as described below. CHO-7 and SRD-12B cells were transfected assearch Laboratories, and glycosidase from New England Biolabs.follows. On day 0, cells were set up at a density of 4 3 104 cells/Lipoprotein-deficient serum (d .1.215 g/ml) was prepared as de-22 mm well in medium B. On day 1, duplicate wells of cells werescribed (Goldstein et al., 1983).transfected with the indicated plasmids using the LipofectAMINEPLUS reagent (Life Technologies) according to the manufacturer’s

Cell Culture instructions with modifications as follows. We used a total of 1.26Cells were maintained in monolayer culture at 378C in 8%–9% CO2. mg of plasmid DNA, 4.5 ml of LipofectAMINE reagent, and 4 ml ofCHO-7 cells are a clone of CHO-K1 cells selected for growth in PLUS reagent per well in a final volume of 0.4 ml. After incubationnewborn calf lipoprotein-deficient serum (Metherall et al., 1989). for 3 hr, the medium was removed, and the cells were fed with 2M19 cells are a clone of amphotericin-resistant CHO-K1 cells auxo- ml of medium A supplemented with 5% fetal calf serum withouttrophic for cholesterol and unsaturated fatty acids (Hasan et al., washing. After incubation for 16 hr, aliquots of 1 ml of medium were1994), owing to a deletion in the S2P gene (Rawson et al., 1997). removed for assay of alkaline phosphatase.CHO/pS2P cells (Rawson et al., 1998) are a clone of CHO-7 cellsstably transfected with pCMV-HSV-S2P, a CMV-driven plasmid thatencodes the human Site-2 protease (Rawson et al., 1998). SRD- Assay for Secreted Alkaline Phosphatase

Cells were transfected with pCMV-PLAP-BP2(513-1141) as de-12A and SRD-12B cells are previously described cholesterol andunsaturated fatty acid auxotrophs derived from g-irradiated CHO/ scribed above. After the indicated interval, aliquots of medium were

removed and spun at top speed (14,000 rpm) in an Eppendorf micro-pS2P cells and selected using an amphotericin B resistance protocol(Rawson et al., 1998). Stock cultures of CHO-7 cells were maintained centrifuge (model 5417C) for 20 min at 48C. The supernatant was

treated at 658C for 30 min to inactivate non-placental alkaline phos-in medium A (1:1 mixture of Ham’s F12 medium and Dulbecco’smodified Eagle medium containing 100 U/ml penicillin and 100 mg/ phatase, after which an aliquot (33 ml) was assayed for placental

alkaline phosphatase activity with the Tropix Phospha-Light assayml streptomycin sulfate) supplemented with 5% (v/v) newborn orfetal calf lipoprotein-deficient serum. Stock-cultures of CHO/pS2P using the substrate and reaction conditions recommended by the

manufacturer. The total volume of the assay was 300 ml. After 10cells were maintained in medium A supplemented with 5% fetal calflipoprotein-deficient serum, 2 mM compactin, and 500 mg/ml G418. to 20 min, chemiluminescence was quantified on an Optima II lumi-

nometer (MGM Instruments).Stock cultures of SRD-12A, SRD-12B, and M19 cells were main-tained in medium B (medium A supplemented with 5% fetal calf After removal of the medium, the cells were lysed with 0.2 ml of

13 Reporter Lysis Buffer (Promega), and aliquots (5 ml) were usedserum, 5 mg/ml cholesterol, 1 mM sodium mevalonate, and 20 mMsodium oleate). Human embryonic kidney 293 cells were maintained for measurement of b-galactosidase activity. The b-galactosidase

assay was carried out with the Genetic Reporter System II kit (Clon-in medium C (Dulbecco’s modified Eagle medium containing 100U/ml penicillin and 100 mg/ml streptomycin sulfate) supplemented tech). This method generated a chemiluminescent product that was

quantified by luminometry. To account for differences in transfectionwith 10% fetal calf serum.

cDNA Cloning of Site-1 Protease for SREBPs513

efficiency, the amount of alkaline phosphatase activity in the me- amino acid 23 of hamster S1P (which is the COOH-terminal aminoacid of the putative signal sequence of S1P), generating intermedi-dium was corrected for the amount of b-galactosidase activity in

the cells. ate construct-2. Third, a 120 bp fragment flanked by NotI sitesencoding three tandem copies of the c-Myc epitope tag (describedabove, kind gift of Dr. C. Kaiser, Massachusetts Institute of Technol-Expression Cloning of Hamster Site-1 Proteaseogy) was inserted into the NotI site of intermediate construct-2,We used pools of hamster cDNAs from a previously described ex-yielding a construct expressing epitope-tagged hamster S1P, desig-pression library (Hua et al., 1996a). Expression was driven by thenated pCMV-Myc-S1P. Point mutations in the S1P coding regionCMV promoter/enhancer. Pools of cDNAs from the library wereof this plasmid were constructed by site-directed mutagenesis astransfected into SRD-12B cells together with three supporting plas-described above. The final structures of all plasmids were confirmedmids as described below. On day 0, replicate wells of SRD-12Bby sequencing all ligation junctions.cells (1 3 105 cells/22 mm well) were plated in medium B. On day

1, duplicate wells of cells were transfected with the following plas-mids: 1 mg of a cDNA pool, 0.21 mg of pCMV-SCAP (Sakai et al.,

Immunoblot Analysis of SREBP Processing in SRD-12B Cells1997), 0.04 mg of pCMV-PLAP-BP2(513-1141), and 0.01 mg of

Transfected with pCMV-S1P and pCMV-Myc-S1PpCMVb-gal using the LipofectAMINE PLUS transfection method as

On day 0, SRD-12B cells were set up at a density of 6 3 105 cells/described above. After incubation for 3 hr, the medium was re-

60 mm dish in medium B. On day 1, cells were transfected with 4moved, and the cells were fed with 2 ml of medium A supplemented

mg of the indicated plasmid DNA plus 12 ml of LipofectAMINE re-with 5% fetal calf serum without washing. After 16 hr, the medium

agent (Life Technologies) according to the manufacturer’s instruc-was removed, heat-treated, and analyzed for alkaline phosphatase

tions, with modifications as follows. LipofectAMINE/DNA complexesactivity. The cells were harvested and assayed for b-galactosidase

were formed in 1 ml of serum-free medium for 15 min at roomactivity as described above.

temperature, and 0.5 ml were added to 1 ml of serum-free mediumThe average value for alkaline phosphatase activity in cells

per dish. After incubation for 3 hr at 378C in a 8%-9% CO2 incubator,transfected with the plasmid cDNAs pools was 0.2 relative light unit

the medium was removed, and the cells were cultured in the mediumof alkaline phosphatase per relative light unit of b-galactosidase. In

A supplemented with 5% fetal calf lipoprotein-deficient serum, 50a screen of 300 pools of 1000 cDNAs per pool, two positive pools

mM compactin, and 50 mM sodium mevalonate in the absence (-)gave a normalized value of 4- to 9-fold higher than the background

or presence (1) of sterols (1 mg/ml of 25-hydroxycholesterol plus 10value of 0.2 and were considered positive. The DNA from one of

mg/ml of cholesterol added in a final concentration of 0.2% ethanol).these pools (No.116) was transformed into E. coli DH5a cells to

After incubation at 378C for 20 hr, the cells received N-acetyl-leuci-generate multiple pools of z100 independent transformants per

nal-leucinal norleucinal (ALLN) at a final concentration of 25 mg/ml.pool. These plasmid pools were transiently transfected into SRD-

After incubation for 1 hr, the cells were harvested and fractionated12B cells along with the three supporting plasmids as described

into nuclear extracts and 105 g membrane pellets as previouslyabove. After 16 hr, we assayed for secreted alkaline phosphatase

described (Sakai et al., 1996). Samples from the nuclear extract andand cellular b-galactosidase activities as described above. Plasmid

membrane fractions were mixed in a ratio of 5:1 with 53 SDS loadingDNA from one positive pool of 100 cDNAs was retransformed into

buffer (13 loading buffer contains 30 mM Tris-HCl at pH 7.4, 3%E. coli, and 121 colonies from this transformation were randomly

SDS, 5% (v/v) glycerol, 0.004% bromphenol blue, and 2.5% b-mer-picked and plated onto an 11 3 11 matrix. Bacterial cultures were

captoethanol). After boiling for 5 min, the proteins were subjected toprepared from pooled colonies from each row and column of the

SDS-PAGE and transferred to Hybond-C Extra nitrocellulose filtersmatrix. Plasmids were isolated from each of these pooled cultures

(Amersham). The filters were incubated with the antibodies de-and transfected into SRD-12B cells together with the three support-

scribed in the figure legends. Bound antibodies were visualizeding plasmids. We then assayed for secreted alkaline phosphatase

with peroxidase-conjugated affinity-purified donkey anti-mouse IgGand cellular b-galactosidase activities. Two positive rows and col-

using the SuperSignal CL-HRP substrate system (Pierce) accordingumns were identified. Plasmids were isolated from the individual

to the manufacturer’s instructions, except that nitrocellulose filtersbacterial colonies at the intersection of the positive rows and col-

were blocked by incubation for 60 min at room temperature withumns. These pure plasmids were transfected into SRD-12B cells as

PBS containing 0.05% (v/v) Tween 20, 5% (v/v) nonfat dry milk,described above, and the medium was assayed for secreted alkaline

and 5% newborn calf serum. Gels were calibrated with prestainedphosphatase activity. As expected, two of the four plasmids gave

molecular weight markers (New England Biolabs). Filters were ex-positive results. These two cDNA clones, designated p38 and p80,

posed to X-Omat Blue XB-1 film (Kodak) with intensifying screenswere identical by restriction mapping and partial DNA sequencing.

at room temperature for the indicated time.Subsequent studies were carried out with clone p38, which is hereaf-ter designated pCMV-S1P. Both DNA strands of pCMV-S1P weresequenced with vector-specific or insert-specific primers by the Blot Hybridization of Genomic DNAdideoxy chain termination method (Sanger et al., 1980). Sequencing Genomic DNA (10 mg) was isolated from cultured cells using thereactions were performed on Applied Biosystems Model 373A and DNA Extraction Kit (Stratagene). The DNA was digested with the377 DNA sequencers. indicated restriction enzymes, subjected to electrophoresis in a

0.7% agarose gel, and transferred to Hybond N1 filters (Amersham)by capillary blotting. After transfer, the filters were cross-linked atConstruction of Epitope-Tagged pCMV-Myc-S1P

pCMV-Myc-S1P encodes an epitope-tagged version of pCMV-S1P 120 mJ in a Stratalinker (Stratagene). Filters were then prehybridizedwith Rapid-hyb buffer (Amersham) at 658C for 1 hr and blotted within which the c-Myc epitope was placed at the NH2 terminus of S1P.

The resulting 1094-amino-acid fusion protein consists of an initiator 32P-labeled probes overnight at 658C. Probes 1 and 2 (see below)were generated by PCR using pCMV-S1P as a template. Probe 1methionine, amino acids 2–23 of hamster S1P (the putative signal

sequence), three novel amino acids (GGR) encoded by the sequence (2.1 kb), corresponding to the 59 end of the S1P cDNA, was amplifiedby PCR using the 59 primer 59-GTGGTGCAAATGGAGTCTAGG-39of the NotI restriction site, three tandem copies of the 9E10 epitope

derived from the human c-Myc protein (SEQKLISEEDLNGEQKLI and the 39 primer 59-CACAGAAATGGAGATGGCCAG-39; probe 2 (2.1kb), corresponding to the 39 end of the S1P cDNA, was amplifiedSEEDLNGEQKLISEEDLNSSG), and amino acids 24–1052 of ham-

ster S1P. pCMV-Myc-S1P was constructed as follows. First, pCMV- using the 59 primer 59-ACCAAGAAGGCAGCTTCCTGG-39 and the 39

primer 59TTCCCAGATCTGTGCACATGC-39; probe 3 (3.0 kb), corre-S1P (described above) was digested with SalI and NotI, and theresulting 4.2 kb insert containing hamster S1P was subsequently sponding to the entire cDNA of hamster ACAT, was isolated from

the vector pRC-CMV7SB by digestion with SalI and NotI (Cao et al.,cloned into the EcoRV–NotI sites of pcDNA3. To destroy the NotIrestriction site, this plasmid was digested with NotI, filled in with 1996). Probes were gel-purified and then labeled with [a-32P]dCTP

by random priming using the Megaprime DNA Labeling System.Klenow fragment, and religated, generating intermediate construct-1.Second, intermediate construct-1 was subjected to in vitro site- Hybridized filters were washed at room temperature for 30 min with

23 SSC and 0.1% SDS, followed by a 1 hr incubation in 0.13 SSCdirected mutagenesis (Kunkel et al., 1987; Hua et al., 1996b) tointroduce a NotI restriction site at the position that corresponds to and 0.1% SDS at 658C. Washed filters were exposed to Kodak

Molecular Cell514

X-Omat Blue XB-1 film at 2808C with intensifying screens and devel- Kyte, J., and Doolittle, R.F. (1982). A simple method for displayingthe hydropathic character of a protein. J. Mol. Biol. 157, 105–132.oped using an automated processor.

Loftus, S.K., Morris, J.A., Carstea, E.D., Gu, J.Z., Cummings, C.,Brown, A., Ellison, J., Ohno, K., Rosenfeld, M.A., Tagle, D.A., et al.Acknowledgments(1997). Murine model of Niemann-Pick C disease: mutation in acholesterol homeostasis gene. Science 277, 232–235.We thank our colleague Kip Guy for helpful discussions, Lisa Beatty

and Vinnie Choudhry for invaluable help with tissue culture, Tammy Marshall, R.D. (1972). Glycoproteins. Annu. Rev. Biochem. 41,Dinh and Susan Burke for excellent technical assistance, and Jeff 673–702.Cormier and Michelle Laremore for DNA sequencing. This work was Metherall, J.E., Goldstein, J.L., Luskey, K.L., and Brown, M.S. (1989).supported by grants from the National Institutes of Health (NIH) (HL- Loss of transcriptional repression of three sterol-regulated genes20948) and the Perot Family Foundation. P. J. E. is the recipient of in mutant hamster cells. J. Biol. Chem. 264, 15634–15641.a NIH Research Science Fellowship Award (HL09993). A. C. S. is

Nagase, T., Miyajima, N., Tanaka, A., Sazuka, T., Seki, N., Sato, S.,supported by NIH Medical Scientist Training Grant GM08014.

Tabata, S., Ishikawa, K.-i., Kawarabayasi, Y., Kotani, H., et al. (1995).Prediction of the coding sequences of unidentified human genes.

Received August 7, 1998; revised September 8, 1998. III. The coding sequences of 40 new genes (KIAA0081-KIAA0120)deduced by analysis of cDNA clones from human cell line KG-1.DNA Res. 2, 37–43.ReferencesNakayama, K. (1997). Furin: a mammalian subtilisin/Kex2p-like en-doprotease involved in processing of a wide variety of precursorBlobel, G. (1980). Intracellular protein topogenesis. Proc. Natl. Acad.proteins. Biochem. J. 327, 625–635.Sci. USA 77, 1496–1500.Nohturfft, A., Hua, X., Brown, M.S., and Goldstein, J.L. (1996). Recur-Brown, M.S., and Goldstein, J.L. (1997). The SREBP pathway: regu-rent G-to-A substitution in a single codon of SREBP cleavage-acti-lation of cholesterol metabolism by proteolysis of a membrane-vating protein causes sterol resistance in three mutant CHO cellbound transcription factor. Cell 89, 331–340.lines. Proc. Natl. Acad. Sci. USA 93, 13709–13714.Cao, G., Goldstein, J.L., and Brown, M.S. (1996). ComplementationNohturfft, A., Brown, M.S., and Goldstein, J.L. (1998). Topology ofof mutation in acyl-CoA: cholesterol acyltransferase (ACAT) fails toSREBP cleavage-activating protein, a polytopic membrane proteinrestore sterol regulation in ACAT-defective sterol-resistant hamsterwith a sterol-sensing domain. J. Biol. Chem. 273, 17243–17250.cells. J. Biol. Chem. 271, 14642–14648.Rawson, R.B., Zelenski, N.G., Nijhawan, D., Ye, J., Sakai, J., Hasan,Chen, W.-J., Andres, D.A., Goldstein, J.L., Russell, D.W., and Brown,M.T., Chang, T.-Y., Brown, M.S., and Goldstein, J.L. (1997). Comple-M.S. (1991). cDNA cloning and expression of the peptide-binding bmentation cloning of S2P, a gene encoding a putative metallopro-subunit of rat p21ras farnesyltransferase, the counterpart of yeasttease required for intramembrane cleavage of SREBPs. Mol. Cell 1,DPR1/RAM1. Cell 66, 327–334.47–57.Cullen, B., and Malim, M. (1992). Secreted placental alkaline phos-Rawson, R.B., Cheng, D., Brown, M.S., and Goldstein, J.L. (1998).phatase as a eukaryotic reporter gene. Methods Enzymol. 216,Isolation of cholesterol-requiring mutant CHO cells with defects in362–368.cleavage of sterol regulatory element binding proteins at Site-1. J.Devereux, J., Haeberli, P., and Smithies, O. (1984). A comprehensiveBiol. Chem 273, 28261–28269.set of sequence analysis programs for the VAX. Nucleic Acids Res.Sakai, J., Duncan, E.A., Rawson, R.B., Hua, X., Brown, M.S., and12, 387–395.Goldstein, J.L. (1996). Sterol-regulated release of SREBP-2 fromDuncan, E.A., Brown, M.S., Goldstein, J.L., and Sakai, J. (1997).cell membranes requires two sequential cleavages, one within aCleavage site for sterol-regulated protease localized to a Leu-Sertransmembrane segment. Cell 85, 1037–1046.bond in lumenal loop of sterol regulatory element binding protein-2.Sakai, J., Nohturfft, A., Cheng, D., Ho, Y.K., Brown, M.S., andJ. Biol. Chem. 272, 12778–12785.Goldstein, J.L. (1997). Identification of complexes between theDuncan, E.A., Dave, U.P., Sakai, J., Goldstein, J.L., and Brown, M.S.COOH-terminal domains of sterol regulatory element binding pro-(1998). Second-site cleavage in sterol regulatory element-bindingteins (SREBPs) and SREBP Cleavage-Activating Protein (SCAP). J.protein occurs at transmembrane junction as determined by cyste-Biol. Chem. 272, 20213–20221.ine panning. J. Biol. Chem. 273, 17801–17809.Sakai, J., Nohturfft, A., Goldstein, J.L., and Brown, M.S. (1998).Goldstein, J.L., Basu, S.K., and Brown, M.S. (1983). Receptor-medi-Cleavage of sterol regulatory element binding proteins (SREBPs) atated endocytosis of LDL in cultured cells. Methods Enzymol. 98,site-1 requires interaction with SREBP cleavage-activating protein.241–260.Evidence from in vivo competition studies. J. Biol. Chem. 273, 5785–

Hasan, M.T., Chang, C.C.Y., and Chang, T.Y. (1994). Somatic cell 5793.genetic biochemical characterization of cell lines resulting from hu-

Sanger, F., Coulson, A.R., Barrell, B.G., Smith, A.J.H., and Roe, B.A.man genomic DNA transfections of Chinese hamster ovary cell mu-(1980). Cloning in single-stranded bacteriophage as an aid to rapidtants defective in sterol-dependent activation of sterol synthesisDNA sequencing. J. Mol. Biol. 143, 161–178.and LDL receptor expression. Somat. Cell Mol. Genet. 20, 183–194.Siezen, R.J., and Leunissen, J.A.M. (1997). Subtilases: the superfam-Hua, X., Sakai, J., Ho, Y.K., Goldstein, J.L., and Brown, M.S. (1995).ily of subtilisin-like serine proteases. Prot. Sci. 6, 501–523.Hairpin orientation of sterol regulatory element binding protein-2 invon Heijne, G. (1985). Signal sequences. The limits of variation. J.cell membranes as determined by protease protection. J. Biol.Mol. Biol. 184, 99–105.Chem. 270, 29422–29427.

Hua, X., Nohturfft, A., Goldstein, J.L., and Brown, M.S. (1996a). SterolGenBank Accession Numbersresistance in CHO cells traced to point mutation in SREBP cleavage

activating protein (SCAP). Cell 87, 415–426.The accession number for the sequence reported in this paper is

Hua, X., Sakai, J., Brown, M.S., and Goldstein, J.L. (1996b). Regu- AF078105.lated cleavage of sterol regulatory element binding proteins(SREBPs) requires sequences on both sides of the endoplasmicreticulum membrane. J. Biol. Chem. 271, 10379–10384.

Kornfeld, R., and Kornfeld, S. (1985). Assembly of asparagine-linkedoligosaccharides. Annu. Rev. Biochem. 54, 631–664.

Kunkel, T.A., Roberts, J.D., and Zakour, R.A. (1987). Rapid and effi-cient site-specific mutagenesis without phenotypic selection. Meth-ods Enzymol. 154, 367–382.