ER Stress Induces Cleavage of Membrane-Bound ATF6 by the Same Proteases that Process SREBPs

Transcription

1 Molecular Cell, Vol. 6, , December, 2000, Copyright 2000 by Cell Press ER Stress Induces Cleavage of Membrane-Bound ATF6 by the Same Proteases that Process SREBPs Jin Ye,* Robert B. Rawson,* Ryutaro Komuro,* Xi Chen, Utpal P. Davé,* Ron Prywes, Michael S. Brown,* and Joseph L. Goldstein* * Department of Molecular Genetics University of Texas Southwestern Medical Center Dallas, Texas Department of Biological Sciences Columbia University New York, New York Summary ATF6 is a membrane-bound transcription factor that activates genes in the endoplasmic reticulum (ER) stress response. When unfolded proteins accumulate in the ER, ATF6 is cleaved to release its cytoplasmic domain, which enters the nucleus. Here, we show that ATF6 is processed by Site-1 protease (S1P) and Site-2 protease (S2P), the enzymes that process SREBPs in response to cholesterol deprivation. ATF6 processing was blocked completely in cells lacking S2P and partially in cells lacking S1P. ATF6 processing required the RxxL and asparagine/proline motifs, known requirements for S1P and S2P processing, respectively. Cells lacking S2P failed to induce GRP78, an ATF6 target, in response to ER stress. ATF6 processing did not require SCAP, which is essential for SREBP processing. We conclude that S1P and S2P are required for the ER stress response as well as for lipid synthesis. Introduction Regulated intramembrane proteolysis (Rip) is the pro- cess by which transmembrane proteins are cleaved to release cytoplasmic domains that often enter the nucleus to regulate gene transcription (Brown et al., 2000). In the examples identified to date, proteolysis proceeds in two steps. First, the extracytoplasmic domain is cleaved to shorten the juxtamembrane sequence to no more than 25 amino acids. Then a second protease cleaves within the transmembrane segment to liberate the cytoplasmic fragment. One of the proteins postulated to undergo Rip in animal cells is ATF6, a transcription factor that plays a central role in the unfolded protein response, also called the endoplasmic reticulum (ER) stress response (Yoshida et al., 1998). As described by Haze et al. (1999), ATF6 is a 670-amino acid, single-pass Type 2 transmembrane protein, i.e., one in which the NH 2 -terminal domain is cytosolic. The NH 2 -terminal domain of amino acids is a transcription factor of the basic-leucine zipper (bzip) family. This is followed by a 21-amino acid transmembrane domain and a 270-amino acid extracyto- To whom correspondence should be addressed ( jgolds@ mednet.swmed.edu and mbrow1@mednet.swmed.edu). plasmic domain that projects into the ER lumen. When unfolded proteins accumulate in the ER, ATF6 is cleaved proteolytically to release the NH 2 -terminal domain, which enters the nucleus (Haze et al., 1999). There it activates transcription of at least three genes (GRP78, GRP94, and calnexin) encoding chaperone proteins that restore the folding of proteins in the ER lumen. This unfolded protein response can be triggered by treatment of cells with tunicamycin, which blocks an enzyme essential for the synthesis of asparagine-linked carbohydrates; with thapsigargin, which depletes the ER of calcium; and with other agents that disrupt the folding of ER proteins (Pahl, 1999). The proteases that process ATF6, the sites of cleavage within the protein, and the cellular locus of this cleavage have not yet been elucidated. A well-understood model of Rip involves the pro- cessing of sterol regulatory element binding proteins (SREBPs), transcription factors that activate the syn- thesis of cholesterol and fatty acids and their uptake from plasma lipoproteins in animal cells (Brown and Goldstein, 1997). The SREBPs are tripartite proteins whose NH 2 -terminal domains of 480 amino acids are basic-helix-loop-helix-leucine zipper (bhlh-zip) transcription factors. This domain is followed by a mem- brane attachment domain of 90 amino acids consisting of two transmembrane helices separated by a short hy- drophilic loop. The COOH-terminal domain of 590 amino acids plays a regulatory role. The SREBPs are oriented in a hairpin fashion with their NH 2 -terminal and COOH-terminal segments facing the cytosol and the hydrophilic loop projecting into the ER lumen. The SREBPs form complexes with another ER protein, SREBP cleavage activating-protein (SCAP), which escorts the SREBPs to the Golgi complex, where the SREBPs undergo Rip (Sakai et al., 1998a; DeBose-Boyd et al., 1999; Nohturfft et al., 2000). This process has been studied most extensively for SREBP-2, but the general features apply to the other two SREBP isoforms. Rip is initiated by Site-1 protease (S1P), a membrane- anchored serine protease whose active site faces the Golgi lumen (Sakai et al., 1998b; Espenshade et al., 1999). S1P cleaves SREBP-2 after the leucine of the sequence Arg-Ser-Val-Leu (RSVL), which lies in the luminal loop (Duncan et al., 1997). The enzyme shows a strong prefer- ence for arginine at the P4 position and leucine or lysine at the P1 position. The intervening residues can be replaced without inhibiting cleavage. Thus, the S1P recog- nition sequence can be stated as RxxL/K, where x is any amino acid (Espenshade et al., 1999). Cleavage by S1P separates the two transmembrane domains of the SREBPs. The NH 2 -terminal product, designated as the intermediate fragment, is a Type 2 trans- membrane protein with a short luminal sequence of only 19 amino acids. This intermediate is cleaved by Site-2 protease (S2P), an unusually hydrophobic zinc metalloprotease that is tightly embedded in the membrane (Rawson et al., 1997; Zelenski et al., 1999). S2P cleaves the intermediate at the leucine-cysteine bond in the sequence Asp-Arg-Ser-Arg-Ile-Leu-Leu-Cys (DRSRILLC), where the isoleucine is the first hydrophobic residue of

2 Molecular Cell 1356 Figure 1. Comparison of Amino Acid Sequences of Human SREBP-2 and Human ATF6 in Region of Regulated Intramembrane Cleavage The sequences are aligned without the introduction of gaps. The box denotes the positionof the first transmembrane domain of SREBP-2. Amino acids that are necessary for Site-1 and Site-2 cleavage of SREBP-2 are highlighted in orange and green, respectively. Amino acids that potentially play the same role in the proteolytic cleavage of ATF6 are numbered and highlighted in the same way. Figure 1 compares the sequences of human ATF6 and SREBP-2 in the transmembrane domains and the imme- diately surrounding regions. In the luminal segment, ATF6 contains two overlapping sequences (RRHL and RHLL) that fit the RxxL consensus for an S1P recognition site as defined in the context of SREBP-2 (Duncan et al., 1997). The transmembrane domain of ATF6 contains an asparagine and a proline, as does SREBP-2 (Ye et al., 2000). However, in contrast to the SREBPs where the asparagine and proline are always adjacent, in ATF6 these two amino acids are separated by two residues (tyrosine-glycine). Notably, the asparagine of the transmembrane sequence and the first arginine of the RRHLL sequence are in the same positions as the corresponding residues of SREBPs relative to the predicted boundaries of the transmembrane domains. To study the proteolytic processing of ATF6, we constructed a plasmid (ptk-hsv-atf6) encoding human ATF6 tagged with two copies of an 11-amino acid NH 2 - terminal epitope derived from an HSV glycoprotein plus a 10-residue spacer. The epitope-tagged ATF6 is ex- pressed under control of the HSV thymidine kinase (TK) promoter, which gives a low but detectable level of ex- pression. ptk-hsv-atf6 was introduced into HEK-293 cells by transfection (Figure 2A). To induce the unfolded protein response, we incubated the HEK-293 cells for 4 hr prior to harvest in the presence of tunicamycin. We also incubated the cells in the absence or presence of N-acetyl-leucine-leucine-norleucinal (ALLN), an inhibi- tor of protein degradation by the proteasome. We included ALLN because previous studies have shown that ALLN blocks the rapid degradation of nuclear fragments of SREBPs, allowing them to be visualized in nuclear extracts (Wang et al., 1994), and we thought the same might apply to ATF6. After these treatments, the cells were harvested, and crude membrane pellets and nu- clear extracts were subjected to SDS PAGE and blotted with an antibody against the NH 2 -terminal HSV tag on ATF6. The ptk-hsv-atf6 plasmid produced a 95 kda band in the membrane fraction that corresponded to the precursor form (denoted as P in Figure 2A). When the cells were incubated with tunicamycin, a 65 kda band appeared in nuclear extracts (designated N in Fig- ure 2A). This nuclear protein was barely visible in the absence of ALLN (lane 3), and it was increased substan- tially in the presence of this inhibitor (lane 5). In the membrane fraction, tunicamycin induced a band designated as P* that migrated just below the precursor. We the membrane-spanning segment (Duncan et al., 1998). The NH 2 -terminal fragment then leaves the membrane with three hydrophobic amino acids (ILL) at its COOH terminus. S2P does not show a stringent requirement for particular amino acids at the cleavage site. Cleavage requires the DRSR sequence just outside the membrane (Hua et al., 1996; Duncan et al., 1998), but recent studies suggest that the SR of DRSR is sufficient (U. P. D., J. Y., J. L. G., and M. S. B., unpublished data). Cleavage by S2P requires the sequence asparagine- proline (NP), which is located near the middle of the membrane-spanning helix and is 11 residues to the COOH-terminal side of the cleaved bond (Ye et al., 2000). Single replacement of either the asparagine or the pro- line reduces cleavage only slightly, but replacement of both residues abolishes cleavage. Asparagine and proline often form the first and second residues, respec- tively, of helices, thereby forming an NH 2 -terminal cap (Richardson and Richardson, 1988). This led to the hy- pothesis that cleavage by S2P requires the unwinding of the portion of the transmembrane helix to the NH 2 -terminal side of the Asn-Pro cap, thereby exposing the Leu-Cys bond to proteolytic attack (Ye et al., 2000). A precedent for partial unfolding of a transmembrane helix, induced by proline, has been observed (Monné et al., 1999). A possible relation between the processing of ATF6 and SREBP was suggested by the observation of an RxxL sequence in the lumen of ATF6 at a site that is 17 residues from the transmembrane domain (Wang et al., 2000) (J. Y., R. B. R., J. L. G., and M. S. B., unpublished observations). This is the same relative location as the RSVL sequence that is cleaved by S1P in SREBP-2 (Dun- can et al., 1997). In addition, ATF6 has an asparagine and a proline in its transmembrane domain, although these are separated by two amino acids and are not contiguous as they are in SREBP-2 (see Figure 1). The current studies were designed to test the hypothesis that S1P and S2P participate in the processing of ATF6. For this purpose, we used lines of mutant CHO cells that are known to be deficient in S1P (Rawson et al., 1998) or S2P (Rawson et al., 1997). These cells were selected as cholesterol auxotrophs because they fail to process SREBPs. Here we show that the S2P-deficient cells have a severe defect in processing of ATF6 and that the S1P-deficient cells have a partial defect. Furthermore, we show that efficient processing of ATF6 in wild-type cells requires the RxxL recognition motif that is required by S1P and the asparagine/proline combination required by S2P. Results

3 Cleavage of ATF6 by SREBP Proteases 1357 Figure 2. Proteolytic Processing and Transcriptional Activity of ATF6 Are Stimulated by Tunicamycin but Not by Sterol Depletion in Transfected HEK-23 Cells believe that this represents nonglycosylated ATF6 that was synthesized in the presence of tunicamycin. In all future experiments, we included ALLN to prevent the degradation of the nuclear form of ATF6. To determine whether the cleavage of ATF6 is regulated by sterols in the same fashion as SREBPs, we measured the appearance of the nuclear form of ATF6 in cells that were incubated either in fetal calf serum (FCS) or in lipoprotein-deficient serum (LPDS) with or without added sterols (Figure 2B). As observed above, tunicamycin induced the appearance of nuclear ATF6 when cells were incubated in FCS (lane 3). A similar induction was seen in lipoprotein-deficient serum in the absence or presence of sterols (lanes 5 and 7, respectively). These data indicate that the cleavage of ATF6 is regulated independently of the cleavage of SREBPs. In Figure 2B, as in Figure 2A, we noted that tunicamycin also induced the appearance of a fragment in the membrane fraction that was of a similar size as the nuclear fragment (designated I in Figure 2B). In part, this band may represent a contamination of the membrane fraction with nuclei. In part, it may also represent an intermediate in the cleavage of ATF6. The evidence for the latter hypothesis is discussed below. To further test for the independence of regulation of ATF6 and SREBPs, we transfected HEK-293 cells with reporter cdnas encoding luciferase under control of a promoter containing a polymerized ATF6 response element (Figure 2C) or a polymerized SREBP response element (Figure 2D). The cells were assayed for luciferase activity after incubation for 16 hr in lipoprotein-deficient serum in the absence or presence of sterols and in the absence or presence of tunicamycin. The data were normalized for transfection efficiency by measurement of -galactosidase activity generated by a cotransfected plasmid encoding -galactosidase under control of the unregulated CMV promoter. A similar assay was shown previously to reflect tunicamycin-induced activation of ATF6 (Wang et al., 2000) and sterol-regulated activation of SREBPs (Rawson et al., 1998). The assay demonstrated that the ATF6 luciferase reporter construct was activated by tunicamycin, and this response was not affected by sterols (Figure 2C). The SREBP reporter was active in the absence of sterols and mark- In (A), HEK-293 cells were set up for experiments on day 0 and transfected on day 2 as described in Experimental Procedures with scribed in Experimental Procedures. Filters were exposed for 7 s. 2 g/dish of ptk-hsv-atf6 or empty-vector pcdna 3.1 (Invitrogen) P, P*, I, and N denote the precursor, precursor with no glycosylation, as indicated. Three hours after transfection, cells were switched to intermediate, and nuclear forms of HSV-tagged ATF6, respectively. medium containing 5% fetal calf serum. On day 3, cells received a In (C) and (D), HEK-293 cells were set up on day 0 at a density of direct addition of 25 g/ml ALLN in the absence or presence of cells per 22 mm well. On day 2, the cells were transfected g/ml tunicamycin as indicated. Four hours later, cells were har- with 1 g/dish of either p5xatf6gl3 (C) or psre-luciferase (D) vested and fractionated as described in Experimental Procedures. together with 0.05 g/dish of pcmv -Gal using an MBS kit as In (B), HEK-293 cells were transfected on day 2 with 2 g/dish described in Experimental Procedures. Three hours after transfecof ptk-hsv-atf6 or empty vector pcdna3.1 as indicated. After tion, cells were switched to medium containing 5% newborn calf transfection, cells were switched to medium containing either 5% lipoprotein-deficient serum containing 50 M compactin and 50 fetal calf serum (FCS) or 5% newborn calf lipoprotein-deficient se- M sodium mevalonate in the absence or presence of 2 g/ml of rum containing 50 M compactin and 50 M sodium mevalonate tunicamycin and/or 1 g/ml of 25-hydroxycholesterol plus 10 g/ (LPDS) in the absence or presence of 1 g/ml of 25-hydroxycholes- ml cholesterol (sterols) as indicated. After incubation for 16 hr, cells terol plus 10 g/ml cholesterol (sterols) as indicated. On day 3, cells were harvested, and luciferase activity was measured and normalreceived a direct addition of 25 g/ml ALLN in the absence or pres- ized to -galactosidase activity as described in Experimental Proceence of 2 g/ml of tunicamycin as indicated. After incubation for 4 dures. A value of 1 represents the normalized luciferase activity hr, the cells were harvested and fractionated. In (A )and (B), aliquots in cells not treated with either tunicamycin or sterols. Each value of nuclear extracts (30 g protein) and membranes (10 g protein) represents the average (and range) of duplicate incubations. Similar were subjected to SDS PAGE and immunoblot analysis as de- results were obtained in one other independent experiment.

4 Molecular Cell 1358 Figure 3. Proteolytic Processing and Transcriptional Activity of ATF6 in Wild-Type and Mutant CHO Cells Lacking S1P, S2P, or SCAP (A) Proteolytic processing. The indicated cells were set up and transfected with 2 g/dish of ptk-hsv-atf6 or empty-vector pcdna3.1 as described in Experimental Procedures. After 20 hr, cells were treated with or without 2 g/ ml of tunicamycin as indicated. Four hours later, the cells were harvested and fractionated, and aliquots of nuclear extracts (30 g protein) and membranes (10 g protein) were subjected to SDS PAGE and immunoblot analysis. Filters were exposed for 15 s. P, P*, X, I, and N denote the precursor, precursor with no glycosylation, band X, intermediate, and nuclear forms of HSV-tagged ATF6, respectively. (B) Transcriptional activity. On day 0, cells were set up at a density of cells per 22 mm well. On day 1, the cells were transfected with p5xatf6gl3 (1 g/dish) and pcmv -Gal (0.05 g/dish) using the Fugene-6 reagent as described in Experimental Procedures. After 10 hr, the cells were treated with or without 2 g/ml of tunicamycin as indicated. Sixteen hours later, the cells were harvested, and luciferase activity was measured and normalized to -galactosidase activity as described in Experimental Procedures. A value of 1 represents the normalized luciferase activity in cells not treated with tunicamycin. Each value represents the average (and range) of duplicate incubations. Similar results were obtained in two other independent experiments. edly reduced in the presence of sterols (Figure 2D). Tuni- The membrane fraction of the SRD-12B cells showed a camycin did not activate the SREBP reporter under either band that migrated slightly slower than the I form. We condition. designated this as band X, and we discuss it in more We next performed a series of experiments in wildtype detail below. SRD-13A cells, which lack SCAP, proprocessing and mutant CHO cells to determine whether ATF6 duced nuclear ATF6. Production was high in the ab- requires the same proteins that are required sence of tunicamycin, and there was no further induction for SREBP processing. When we transfected the ptk- when tunicamycin was added (lanes 11 and 12). Again, HSV-ATF6 construct into wild-type CHO cells, we noted we believe that the high basal level of processing of one difference from the behavior of HEK-293 cells. ATF6 in the absence of tunicamycin was due to overexpression Whereas HEK-293 cells exhibited little cleavage of the of ATF6 in CHO cells (see Discussion). transfected ATF6 in the absence of tunicamycin (Figures To determine whether the protease-deficient cell lines 2A and 2B), the CHO cells showed appreciable cleavage, could mount a normal regulatory response to unfolded and the cleaved fragment appeared in nuclear extracts proteins, we transfected the cells with the plasmid encoding (Figure 3A, lane 2). We believe that this cleavage is luciferase driven by the promoter containing the attributable to a partial stimulation of the unfolded pro- polymerized ATF6 binding site. After transfection, the tein response by the overproduction of HSV-tagged cells were incubated in the presence or absence of tuni- ATF6. For unknown reasons, CHO cells are more sensi- camycin, and luciferase activity was measured (Figure tive to this overproduction than are the HEK-293 cells. 3B). In wild-type CHO cells, luciferase activity increased Despite this high level of background cleavage, the CHO 6-fold in response to tunicamycin. This relative dependence cells showed an increase in the nuclear fragment when on tunicamycin was greater than the tunicamycincells tunicamycin was added (Figure 3A, lane 3). M19 cells, dependence of ATF6 cleavage (Figure 3A, lanes 2 and 3). which lack S2P, did not produce any nuclear form of We attribute this difference to the fact that the luciferase ATF6 with or without tunicamycin. Instead, we observed reporter experiments do not require overproduction of a membrane-bound intermediate form (I) (lanes 5 and ATF6 but depend on endogenous ATF6 (see Discus- 6), which we believe is the fragment produced by Site-1 sion). In S2P-deficient M19 cells, luciferase activity was cleavage (see below). SRD-12B cells, which lack S1P, low in the absence of tunicamycin and rose only slightly cleaved transfected ATF6, but the efficiency of pro- after tunicamycin treatment. In S1P-deficient SRD-12B cessing was decreased as judged by the reduced ratio cells, luciferase activity was also low, but it was partially of nuclear fragments to precursor (lanes 8 and 9) com- stimulated by tunicamycin, reaching maximal levels that pared with that in wild-type CHO cells (lanes 2 and 3). were one third of the induced level in the wild-type CHO

5 Cleavage of ATF6 by SREBP Proteases 1359 were blotted with an antibody against the HSV tag (Figure 4). When ptk-hsv-atf6 was transfected alone, we detected a small amount of cleaved nuclear ATF6 and a small amount of band X in the membrane fraction (lane 2). There was no change when tunicamycin was added (lane 3). When we transfected the cdna encoding wildtype S1P and incubated the cells with tunicamycin, the nuclear fragment of ATF6 increased substantially (lane 4). In the membrane fraction, a band appeared that migrated slightly faster than band X. We believe that this lower band represents the intermediate that is produced by Site-1 cleavage. The control plasmid encoding inactive S1P did not produce an increase in the nuclear fragment, nor did band I appear (lane 5). If some of the ATF6 processing is mediated by S1P, then the processing should be reduced when the amino acids at the S1P recognition site are replaced. We therefore produced a series of cdnas encoding HSV-tagged ATF6 with mutations in the RRHLL sequence that contains two overlapping RxxL motifs that are potential Figure 4. Transfected S1P Stimulates Proteolytic Processing of recognition sites for S1P (see Figure 1). When we trans- ATF6 in Mutant CHO Cells Lacking S1P fected a cdna encoding wild-type ATF6 into HEK-293 S1P-deficient SRD-12B cells were cotransfected with a plasmid cells, we observed a tunicamycin-induced cleavage that encoding HSV-tagged ATF6 (2 g/dish) together with 0.5 g/dish generated the cleaved form (N) in the nucleus and the of either wild-type S1P (WT) or mutant S1P (S414A) as indicated. intermediate form (I) in the membranes (Figure 5, lanes The total amount of DNA was adjusted to 2.5 g/dish by addition of pcdna3.1 empty vector. After 20 hr, cells were treated with or 3 and 4). There was no reduction in the N or I bands without 2 g/ml tunicamycin as indicated. Four hours later, the cells when we replaced the first arginine of the RRHLL se- were harvested and fractionated, and aliquots of nuclear extracts quence with alanine (R415A) (lanes 5 and 6). However, (30 g protein) and membranes (10 g protein) were subjected to when we replaced the second arginine, the nuclear form SDS PAGE and immunoblot analysis as described in Experimental and the I band no longer appeared, and instead we Procedures. Filters were exposed for 15 s. P, P*, X, I, and N denote observed an increase in the band that migrates slightly the precursor, precursor with no glycosylation, band X, intermediate, and nuclear forms of HSV-tagged ATF6, respectively. slower than the I form (designated X in Figure 5). When both arginine residues were changed to alanine, the nuclear and intermediate forms were again undetectable cells. The SCAP-deficient SRD-13A cells responded to (lanes 9 and 10). tunicamycin like the wild-type cells. Considered to- Up to this point, the data of Figure 5 indicate that the gether, the results in Figure 3A and 3B indicate that S2P S1P recognition site begins with the second arginine of is required for ATF6 cleavage but SCAP is not required. the RRHLL sequence, and, therefore, the recognition The requirement for S1P is only partial. sequence is RHLL. If this is the case, cleavage should To further test the hypothesis that S1P participates occur after the second leucine (L419; see Figure 1). A in ATF6 processing, we corrected the defect in the S1P- stringent test of this hypothesis lies in the replacement deficient SRD-12B cells by transient transfection with a of L419 with valine. This highly conservative substitution cdna encoding wild-type S1P. As a control, we used a removes only one methyl group, yet in SREBP-2 the cdna that encodes an inactive version of S1P with the corresponding mutation severely reduces S1P cleavage active-site serine replaced with an alanine (S414A) (Sa- (Duncan et al., 1997). As shown in Figure 5, the L419V kai et al., 1998b). The cells were also transfected with mutation in ATF6 severely reduced the amount of the ptk-hsv-atf6, and membrane and nuclear extracts nuclear fragment (lanes 11 and 12). Moreover, it led to Figure 5. Mutation Altering S1P Recognition Site in ATF6 Reduces Its Proteolytic Processing in Transfected HEK-293 Cells HEK-293 cells were set up and transfected with a plasmid encoding either wild-type or mutant HSV-tagged ATF6 (2 g/dish) as indicated. After 20 hr, cells were treated with or without 2 g/ml of tunicamycin as indicated. Four hours later, the cells were harvested and fractionated, and aliquots of nuclear extracts (30 g protein) and membranes (10 g protein) were subjected to SDS PAGE and immunoblot analysis. Filters were exposed for 7 s. P, P*, X, I, and N denote the precursor, precursor with no glycosylation, band X, intermediate, and nuclear forms of HSV-tagged ATF6, respectively.

6 Molecular Cell 1360 ence of tunicamycin, but the intermediate form (I) was observed in the membrane fraction (lanes 2 and 3). When wild-type S2P was coexpressed in the presence of tunicamycin, the nuclear form appeared (lane 4). In contrast, when catalytically inactive S2P was expressed, no nuclear form was observed (lane 5). To assay the effect of S2P on endogenous ATF6- dependent transcription, we used the luciferase assay that was described in Figure 3. As shown in Figure 6B, in wild-type CHO cells tunicamycin induced high levels of expression of luciferase that was produced from a plasmid containing five copies of the ATF6 binding site. The induction was much less in the S2P-deficient M19 cells. The induction was restored to normal in a stable line of M19 cells in which S2P expression had been restored by transfection with wild-type S2P. Luciferase induction was not restored when a stable line of M19 cells expressed the inactive mutant version of S2P (H171F). To determine whether the asparagine and proline of the ATF6 transmembrane sequence (Figure 1) are required for cleavage by S2P, we produced ATF6 cdnas in which these residues were substituted individually or together with amino acids that are known to abolish cleavage when substituted into the NP sequence of SREBP-2 (Figure 7). Upon transfection into HEK-293 cells, wild-type ATF6 was cleaved to generate a nuclear fragment in a tunicamycin-dependent fashion (lanes 3 and 4). Neither the N391F mutation nor the P394L mutation abolished cleavage (lanes 5 8). However, when both residues were replaced, cleavage was abolished (lanes 9 and 10). These results parallel the results with SREBP-2, in which substitution of both the asparagine and proline of the NP sequence are necessary in order to abolish cleavage by S2P (Ye et al., 2000). In the membrane fraction, all of the ATF6 constructs gave rise to some of the I band in the presence of tunicamycin, but this band accumulated to much higher levels when S2P cleavage was blocked by the combined asparagine/pro- Figure 6. Transfected S2P Restores Proteolytic Processing and Transcriptional Activity of ATF6 in Mutant CHO Cells Lacking S2P (A) Proteolytic processing. M19 cells were cotransfected with a plasmid encoding HSV-tagged ATF6 (2 g/dish) together with 0.5 g/ dish of either wild-type (WT) or mutant S2P (H171F) as indicated. The total amount of DNA was adjusted to 2.5 g/dish by addition of pcdna3.1 empty vector. After 20 hr, cells were treated with or without 2 g/ml of tunicamycin as indicated. Four hours later, the cells were harvested and fractionated, and aliquots of nuclear extracts (30 g protein) and membranes (10 g protein) were subjected line mutation (lane 10). to SDS PAGE and immunoblot analysis. Filters were exposed to Inasmuch as S2P appears to be strongly required for film for 15 s. P, P*, I, and N denote the precursor, precursor with ATF6 processing, the induction of ATF6 target genes no glycosylation, intermediate, and nuclear forms of HSV-tagged during the unfolded protein response should be severely ATF6, respectively. (B) Transcriptional activity. Cells were set up, transfected, incupothesis, we used an antibody against GRP78, an ER decreased in S2P-deficient M19 cells. To test this hybated, and harvested for luciferase assay as described in the legend to Figure 3B. Each value represents the average (and range) of chaperone that is known to be induced directly by ATF6 duplicate incubations. Similar results were obtained in two other (Haze et al., 1999). When wild-type CHO cells were independent experiments. treated either with tunicamycin or thapsigargin, the amount of immunodetectable GRP78 rose (Figure 8, a major increase in band X, thereby reproducing both lanes 2 and 3). This induction did not occur in S2P- deficient M19 cells (lanes 5 and 6). Induction occurred in the stable line of M19 cells that had been transfected with the cdna encoding wild-type S2P (lanes 8 and 9) but not in the cells expressing the catalytically inactive H171F mutant (lanes 11 and 12). Figure 8B shows that the level of expression of the wild-type S2P and the inactive H171F mutant of S2P were similar in the two stably transfected cell lines, as determined by immu- noblotting. effects that were generated by the R416A mutation. The finding of a requirement for an arginine at the potential P4 position and a leucine at the presumed P1 position provides strong support for the hypothesis that S1P participates in ATF6 processing. To document the role of S2P in ATF6 processing, we introduced ptk-hsv-atf6 into S2P-deficient M19 cells by transfection together with a cdna encoding either wild-type S2P or an inactive mutant with a replacement of a crucial histidine at the zinc binding site (H171F) (Rawson et al., 1997). Membrane and nuclear extracts were immunoblotted with the anti-hsv antibody (Figure 6A). In the absence of S2P, no cleaved form of ATF6 was observed in nuclear extracts in the absence or pres- Discussion The current data expand the roles of the Site-1 and Site-2 proteases beyond the regulation of lipid metabolism,

7 Cleavage of ATF6 by SREBP Proteases 1361 Figure 7. Asparagine and Proline within Membrane-Spanning Segment of HSV-ATF6 Are Required for Its Proteolytic Processing in Transfected HEK-293 Cells HEK-293 cells were set up and transfected with a plasmid encoding either wild-type or mutant HSV-tagged ATF6 (2 g/dish) as indicated. Cells were then incubated and harvested for immunoblot analysis as described in the legend to Figure 5. Filters were exposed to film for 15 s. P, P*, I, and N denote the precursor, precursor with no glycosylation, intermediate, and nuclear forms of HSVtagged ATF6, respectively. and they have direct implications for the mechanism by S1P (Duncan et al., 1997; Espenshade et al., 1999). The which the ER stress response is triggered in animal cells. processing also required an asparagine and a proline By processing ATF6, S1P and S2P are central to the in the transmembrane sequence, which mimics the requirement ER stress response. The requirement for S2P in ATF6 for cleavage by S2P. As with SREBP cleavage processing was essentially absolute. After transfection (Ye et al., 2000), processing was not blocked when either with a cdna encoding epitope-tagged ATF6, mutant of these residues was replaced singly, but processing CHO cells lacking S2P showed no detectable nuclear was completely blocked when the two residues were ATF6 fragment after treatment with tunicamycin (Figures replaced simultaneously (Figure 7). 3 and 6), and this was restored by expression of wild- In order to obtain physiologically relevant data in type S2P but not a catalytically inactive mutant (Figure transfection experiments, we found it necessary to limit 6). The requirement for S1P was less complete. Thus, the amount of ATF6 that was produced. In experiments mutant CHO cells lacking S1P generated a reduced but not shown, when we used the strong CMV promoter to detectable amount of nuclear ATF6 (Figures 3 and 4), drive expression of epitope-tagged ATF6, we observed and this defect was erased by transfection of a cdna constitutive cleavage in the absence of ER stress. This encoding wild-type S1P but not a catalytically inactive cleavage occurred in the absence of S1P and S2P, and mutant (Figure 4). it generated fragments whose sizes differed from those The conclusions reached from the studies with mutant generated by the S1P- and S2P-dependent reactions. CHO cells are supported by experiments with wild-type We believe that these bands arise from adventitious HEK-293 cells. In these cells, processing of ATF6 re- cleavage of the highly expressed ATF6 by proteases quired arginine and leucine at the P4 and P1 positions of that do not process this protein when it is present at the putative S1P recognition site (Figure 5), which is the physiologic levels. A similar type of problem was observed same specificity that applies to cleavage of SREBP-2 by by Haze et al. (1999). We partially overcame this Figure 8. Treatment with Tunicamycin or Thapsigargin Fails to Induce GRP78 (BIP) in Mutant CHO Cells Deficient in S2P (A) Immunoblot analysis of GRP78. On day 0, CHO-7 cells (lanes 1 3), M19 cells (lanes 4 6), and M19 cells stably transfected with either HSV-tagged wild-type S2P (lanes 7 9) or HSV-tagged mutant S2P (H171F) (lanes 10 12) were set up at a density of per 60 mm dish. On day 1, the cells received either tunicamycin (2 g/ml) or thapsigargin (300 nm) as indicated. After incubation for 16 hr, the cells were harvested and fractionated, and aliquots of membranes (20 g protein) were subjected to SDS PAGE and immunoblot analysis with anti-kdel antibody as described in Experimental Procedures. Filters were exposed to film for 2 s. (B) Immunoblot analysis of S2P. Aliquots of membranes (20 g protein) from lanes 4, 7, and 10 in (A) were subjected to SDS PAGE and immunoblot analysis with anti-hsv-tag antibody to visualize wild-type and mutant S2P. Filters were exposed to film for 15 s.

8 Molecular Cell 1362 problem through use of the weak TK promoter to drive to the processing site, and how this transport may be ATF6 expression. Even in this case, however, in experiments regulated. Previous studies of SREBP processing have with transfected CHO cells we observed relatively revealed that active S1P resides in a post-er compart- high constitutive cleavage of ATF6 in the absence of ment that is in or near the Golgi complex (Nohturfft et ER stress (see Figure 3). We believe that this cleavage al., 1999). In order to be processed by S1P, SREBPs represented the physiologic process because it retained must be transported to this site by an escort protein, its complete requirement for S2P and at least a partial SCAP. Sterols block the exit of the SCAP/SREBP complex requirement for S1P. Constitutive cleavage was much from the ER, thereby blocking SREBP processing less of a problem in transfected HEK-293 cells, which and repressing their own synthesis (DeBose-Boyd et al., showed a strong and consistent requirement for ER 1999; Nohturfft et al., 2000). By analogy with SREBPs, stress in order to exhibit appreciable amounts of nuclear it seems likely that ATF6 must leave the ER in order to ATF6 (Figure 5). be processed and that this exit may be triggered by the In light of the potential concerns arising from experi- accumulation of unfolded proteins in the ER. Clearly ments in which ATF6 is overproduced by transfection, SCAP is not required for this transport, since ATF6 is we paid special attention to the experiments with the processed normally in mutant cells that lack SCAP (Figure ATF6-driven luciferase reporter plasmid, which relied on 3, lanes 11 and 12). The mechanism for the proposed endogenous ATF6 and did not involve overexpression ER retention of ATF6 and its postulated regulated exit of ATF6. This experiment and others like it (not shown) from this organelle are now open to study. confirmed the requirement for S1P and S2P in order to achieve a strong induction of expression by tunicamycin Experimental Procedures (Figure 3B). The requirement for S2P was further indi- Materials cated by the experiment of Figure 8, in which tuni- We obtained tunicamycin and thapsigargin from Sigma, monoclonal camycin and thapsigargin failed to induce GRP78 in anti-hsv-tag (IgG1) from Novagen, anti-kdel antibody from Stress- CHO cells lacking S2P, whereas they accomplished this Gen, and horseradish peroxidase-conjugated donkey anti-mouse induction in a permanent cell line whose S2P expression IgG (affinity purified) from Jackson ImmunoResearch Laboratories. Other reagents were obtained from sources as described (Sakai et had been restored. This experiment did not involve any al., 1996; Espenshade et al., 1999; Rawson et al., 1999). Lipoproteintransient transfection at all. deficient serum (density g/ml) was prepared by ultracentrif- In addition to the nuclear fragment of ATF6, we also ugation of newborn calf serum (Goldstein et al., 1983). observed two closely spaced membrane-bound fragments that we designated band I (for intermediate) and Plasmid Constructs band X. Although the identities of these fragments are The following plasmids were described in the indicated reference: ptk-hsv-srebp2, encoding Herpes simplex virus (HSV) epitopenot yet established for certain, the data allow some tagged human SREBP-2 driven by the Herpes simplex virus thyprovisional conclusions to be drawn. We believe that midine kinase (TK) promoter (Sakai et al., 1996); pcmv-myc-s2p, band I represents the intermediate fragment that is pro- encoding Myc epitope-tagged human S2P driven by the CMV enhancer/promoter duced by the Site-1 cleavage reaction. This band is (Rawson et al., 1997); pcmv-s1p-myc, encoding barely visible in wild-type cells because the second hamster S1P with c-myc epitope tagged at the COOH terminus followed by six histidine residues (Espenshade et al., 1999); cleavage follows immediately after the first. Band I appcgnatf6, encoding full-length human ATF6 with an HA epitope pears most prominently in S2P-deficient CHO cells (Figtag at the NH 2 terminus driven by the CMV enhancer/promoter (Wang ures 3 and 6) and in HEK-293 cells that express the et al., 2000); p5xatf6gl3, encoding firefly luciferase driven by five asparagine/proline mutant form of ATF6 where Site-2 repeats of the ATF6 binding site (Wang et al., 2000); and psreluciferase, cleavage does not take place (Figure 7, lane 10). This encoding luciferase driven by three repeats of the SRE-1 is similar to the results with SREBP-2, wherein the intera site in the LDL receptor promoter (Rawson et al., 1997). pcmv -Gal, plasmid encoding -galactosidase driven by the CMV enhancer/ mediate formed by Site-1 cleavage is visualized only promoter, was obtained from Stratagene. when Site-2 cleavage is blocked (Sakai et al., 1996). ptk-hsv-atf6 encodes full-length human ATF6 (amino acids The identity of band X is more problematic. This band 1 670) preceded by an initiation methionine, two tandem copies migrates only slightly more slowly than band I on SDS of the HSV epitope (QPELAPEDPED), and ten novel amino acids PAGE, and it sometimes fails to be separated at all. (TRELGKGTCE). It is driven by Herpes simplex virus TK promoter. Band X appears most prominently when Site-1 cleavage It was generated by the following method. First, ptk-hsv-srebp2 was digested with NotI and SpeI, and the 10 kb backbone fragment is blocked. This is seen most clearly in Figure 4 where was isolated. It was then ligated with sense and antisense oligonuthe S1P-deficient cells showed band X (lanes 2 and 3). cleotides encoding an initiation methionine and two tandem copies This band was diminished, and the lower band I ap- of HSV epitope tag along with NotI and SpeI overhangs. The 3 SpeI peared when the cells expressed active S1P (lane 4) but overhang was designed to fit into the open reading frame ACT AGT. not inactive S1P (lane 5). Similarly, band X was promikb The resulting plasmid was digested with SpeI and XbaI, and the 6 nent in the experiment of Figure 5, when HEK-293 cells backbone fragment was ligated to the 2.5 kb fragment that was generated by digestion of pcgnatf6 with XbaI. The resulting plasproduced the R416A and the L419V mutants of ATF6 mid with the correct orientation was selected. Integrity of ptk-hsvthat fail to undergo Site-1 cleavage (lanes 8 and 12). AT6 was confirmed by sequencing the entire coding region and the These data are consistent with the possibility that band ligation joints. X is produced by cleavage of ATF6 by an enzyme other All plasmids expressing ATF6 mutants were generated by a PCRthan S1P, which generates a fragment that cannot be based approach using HSV epitope-tagged human ATF6 as the cleaved by S2P and, therefore, remains membrane template. The entire coding region of each mutant plasmid was bound. This alternate cleavage occurs to an appreciable sequenced in its entirety. extent only when Site-1 cleavage fails to take place. Cell Culture The current findings raise the immediate question of Cells were maintained in monolayer culture at 37 C in 8% 9% CO 2. where in the cell ATF6 is processed, how it is transported CHO-7 cells are a clone of CHO-K1 cells selected for growth in

9 Cleavage of ATF6 by SREBP Proteases 1363 lipoprotein-deficient serum (Metherall et al., 1989). SRD-12B cells Acknowledgments (deficient in S1P) (Rawson et al., 1998) and SRD-13A cells (deficient in SCAP) (Rawson et al., 1999) are mutant cells derived from CHO-7 We thank Jonathan M. Starkey and Clinton Steffey for excellent cells. M19 cells (deficient in S2P) are mutant cells derived from CHO- technical assistance; Lisa Beatty and Jill Abadia for invaluable help K1 cells (Hasan et al., 1994; Rawson et al., 1997). Stock cultures of with tissue culture; and Jeff Cormier and Lorena Avila for DNA sequencing. CHO-7 cells were maintained in medium A (1:1 mixture of Ham s This work was supported by grants from the National F12 medium and Dulbecco s modified Eagle s medium containing Institutes of Health (HL and CA50329) and the Perot Family 100 U/ml penicillin and 100 g/ml streptomycin sulfate) supplemented Foundation. R. K. is the recipient of a 2000 Banyu Fellowship Award with 5% (v/v) newborn calf lipoprotein-deficient serum. M19, in Cardiovascular Medicine from the Merck Company Foundation. SRD-12B, and SRD-13A cells were maintained in medium B (medium U. P. D. is the recipient of a Postdoctoral Fellowship for Physicians A supplemented with 5% (v/v) fetal calf serum, 5 g/ml cholesterol, from the Howard Hughes Medical Institute. 1mM sodium mevalonate, and 20 M sodium oleate). Human embryonic kidney-293 (HEK-293) cells were maintained in medium A sup- Received August 17, 2000; revised September 20, plemented with 5% (v/v) fetal calf serum. References Transfection of Cells and Immunoblot Analysis Brown, M.S., and Goldstein, J.L. (1997). The SREBP pathway: Regu- Stably transfected M19 cells expressing either wild-type human lation of cholesterol metabolism by proteolysis of a membrane- HSV-tagged S2P or mutant S2P(H171F) were isolated as follows. bound transcription factor. Cell 89, The transfected cells were selected as previously described (Raw- Brown, M.S., Ye, J., Rawson, R.B., and Goldstein, J.L. (2000). Reguson et al., 1997), and surviving cells were cloned by limiting dilution. lated intramembrane proteolysis: a control mechanism conserved The resulting clonal lines were screened by immunoblot analysis from bacteria to humans. Cell 100, for expression of the HSV-tagged S2P. Cell lines expressing similar DeBose-Boyd, R.A., Brown, M.S., Li, W.-P., Nohturfft, A., Goldstein, levels of wild-type and mutant S2P were used in this study. J.L., and Espenshade, P.J. (1999). Transport-dependent proteolysis CHO-7, M19, SRD-12B, and SRD-13A cells were transiently transof SREBP: relocation of Site-1 protease from Golgi to ER obviates fected with Fugene-6 reagent (Roche Molecular Biochemicals). M19 the need for SREBP transport to Golgi. Cell 99, cells were set up at a density of per 60 mm dish and transfected two days later. CHO-7, SRD-12B, and SRD-13A cells Duncan, E.A., Brown, M.S., Goldstein, J.L., and Sakai, J. (1997). were set up at a density of per 60 mm dish and transfected Cleavage site for sterol-regulated protease localized to a Leu-Ser one day later. For each transfection, the indicated amount of DNA bond in lumenal loop of sterol regulatory element binding protein-2. J. Biol. Chem. 272, was mixed with Fugene-6 (3 l per g of DNA) in serum-free medium (0.2 ml) according to the manufacturer s protocol. The DNA/Fugene-6 Duncan, E.A., Davé, U.P., Sakai, J., Goldstein, J.L., and Brown, M.S. mixture was then added to cells, after which each monolayer was (1998). Second-site cleavage in sterol regulatory element-binding cultured in 4 ml medium A supplemented with 5% (v/v) fetal calf protein occurs at transmembrane junction as determined by cyste- serum. ine panning. J. Biol. Chem. 273, HEK-293 cells were transiently transfected using an MBS kit (Stra- Espenshade, P.J., Cheng, D., Goldstein, J.L., and Brown, M.S. tagene). Cells were set up at a density of per 60 mm dish (1999). Autocatalytic processing of Site-1 protease removes propepand transfected two days later as previously described (Sakai et al., tide and permits cleavage of sterol regulatory element-binding pro- 1996). teins. J. Biol. Chem. 274, After transient transfection, cells were incubated at 37 C for 20 hr, Goldstein, J.L., Basu, S.K., and Brown, M.S. (1983). Receptor-mediafter which each monolayer received a direct addition of N-acetyl- ated endocytosis of LDL in cultured cells. Meth. Enzymol. 98, leucinal-leucinal-norleucinal (ALLN) at a final concentration of g/ml. In addition, other drugs (tunicamycin and sterols) were added Hasan, M.T., Chang, C.C.Y., and Chang, T.Y. (1994). Somatic cell as indicated in the figure legend. ALLN and tunicamycin were disgenetic biochemical characterization of cell lines resulting from husolved in DMSO (final concentration of DMSO in each dish was man genomic DNA transfections of Chinese hamster ovary cell mu- 0.2%). Sterols (mixture of 25-hydroxycholesterol plus cholesterol) tants defective in sterol-dependent activation of sterol synthesis were dissolved in ethanol (final concentration of ethanol was 0.2%). and LDL receptor expression. Somat. Cell Mol. Genet. 20, Nontreated cells received 0.2% DMSO and/or 0.2% ethanol. After Haze, K., Yoshida, H., Yanagi, H., Yura, T., and Mori, K. (1999). incubation for 4 hr at 37 C, the cells were harvested and fractionated Mammalian transcription factor ATF6 is synthesized as a transmeminto membrane and nuclear extract fractions as previously debrane protein and activated by proteolysis in response to endoplasscribed (Sakai et al., 1996) with one modification: after 1 hr incubamic reticulum stress. Mol. Biol. Cell 10, tion with buffer C (20 mm HEPES [ph 7.6], 2.5% (v/v) glycerol, 0.42 M NaCl, 1.5 mm MgCl 2 ), the nuclear suspension was centrifuged at Hua, X., Sakai, J., Brown, M.S., and Goldstein, J.L. (1996). Regulated g instead of g for 0.5 hr to collect the supernatant cleavage of sterol regulatory element binding proteins (SREBPs) as nuclear extract. requires sequences on both sides of the endoplasmic reticulum Immunoblot analysis was carried out after 8% SDS PAGE as de- membrane. J. Biol. Chem. 271, scribed (Sakai et al., 1996) by using the SuperSignal CL-HRP Sub- Metherall, J.E., Goldstein, J.L., Luskey, K.L., and Brown, M.S. (1989). strate System (Pierce). HSV-ATF6 was detected with 0.2 g/ml Loss of transcriptional repression of three sterol-regulated genes monoclonal IgG-HSV-Tag, and GRP78 (Bip) was detected with 0.55 in mutant hamster cells. J. Biol. Chem. 264, g/ml anti KDEL antibody (StressGen). They were both visualized Monné, M., Hermansson, M., and von Heijne, G. (1999). A turn prowith a 1:5000 dilution of peroxidase-conjugated, affinity-purified pensity scale for transmembrane helices. J. Mol. Biol. 288, donkey anti-mouse IgG (Jackson ImmunoResearch). Gels were ex- Nohturfft, A., DeBose-Boyd, R.A., Scheek, S., Goldstein, J.L., and posed at room temperature to X-Omat Blue XB-1 film (Kodak) for Brown, M.S. (1999). Sterols regulate cycling of SREBP cleavagethe indicated time. activating protein (SCAP) between endoplasmic reticulum and Golgi. Proc. Natl. Acad. Sci. USA 96, Nohturfft, A., Yabe, D., Goldstein, J.L., Brown, M.S., and Espen- Luciferase Assay shade, P.J. (2000). Regulated step in cholesterol feedback localized Transfected cells were lysed and assayed for luciferase and to budding of SCAP from ER membranes. Cell 102, galactosidase activity as described previously (Rawson et al., 1997). Photons were detected in an Optima II Luminator (MGM Inlum Pahl, H.L. (1999). Signal transduction from the endoplasmic reticustruments). The amount of luciferase activity was normalized to to the cell nucleus. Physiol. Rev. 79, the amount of -galactosidase activity to correct for transfection Rawson, R.B., Zelenski, N.G., Nijhawan, D., Ye, J., Sakai, J., Hasan, efficiency in each experiment. M.T., Chang, T.-Y., Brown, M.S., and Goldstein, J.L. (1997). Comple-