Index Introduction Experimental Procedures Preparation, culture and transfection of ATF6α+/+ and ATF6α–/– MEFs Microarray analysis Northern blot hybridization Immunological techniques Results Identification of ATF6α-target genes Requirement of ATF6α for transcriptional induction of most ER chaperones Requirement of ATF6α for transcriptional induction of some ERAD components Cross talk among UPR mediators Dispensability of ATF6α in transcriptional induction of proteins involved in translocation into and transport from the ER Discussion Acknowledgements References

Introduction

Newly synthesized secretory and transmembrane proteins are translocated into the endoplasmic reticulum (ER), which contains a number of molecular chaperones and folding enzymes (collectively termed ER chaperones hereafter) and provides an optimal environment for the productive folding of these proteins. Proteins remaining unfolded or misfolded even after the assistance of ER chaperones are retrotranslocated back to the cytosol, where they are ubiquitinated and degraded by the proteasome through a process termed ER-associated degradation (ERAD). These two mechanisms, productive folding and ERAD, ensure the quality of proteins that pass through the ER and allow only correctly folded molecules to move along the secretory pathway (Bukau et al., 2006). However, the ER quality control system is compromised under a variety of conditions, collectively termed ER stress, resulting in the accumulation of unfolded proteins in the ER. Essentially all eukaryotic cells cope with ER stress and maintain the homeostasis of the ER by activating the unfolded protein response (UPR) (Ron and Walter, 2007).

UPR signaling is transduced across the ER membrane by a transmembrane protein present in the ER (Mori, 2000). The budding yeast Saccharomyces cerevisiae expresses Ire1p, a transmembrane protein kinase/endoribonuclease in the ER, which upon ER stress initiates unconventional splicing of HAC1 mRNA. This in turn results in production of the UPR-specific transcription factor Hac1p, leading to the transcriptional induction of hundreds of genes encoding proteins working at various stages of secretion, including both ER chaperones and ERAD components. Induced proteins help yeast cells to deal with unfolded proteins accumulated in the ER (see the Discussion for details). Importantly, the number of such UPR transducers has increased with evolution, allowing higher organisms to cope with ER stress in a more sophisticated way (Bernales et al., 2006).

Mammalian ER expresses three transmembrane UPR transducers, which carry characteristic effector domains in their cytoplasmic regions (Schroder and Kaufman, 2005). These are IRE1 (Ire1p homologue), PERK (transmembrane protein kinase) and ATF6 (transmembrane transcription factor). Thus, in contrast to yeast cells, which cope with ER stress only by inducing transcription, mammalian cells are capable of decreasing the burden on the ER by attenuating translation generally via the activation of PERK, which phosphorylates the α subunit of eukaryotic translation initiation factor 2 (eIF2α) (Ron, 2002). In addition, mammalian cells are capable of inducing the transcription of a variety of sets of genes by activating three transcription factors downstream of the three UPR transducers. Activated PERK-mediated translational attenuation paradoxically induces the translation of transcription factor ATF4 (Harding et al., 2000a). Activated IRE1 initiates unconventional splicing of XBP1 mRNA to produce the highly active transcription factor pXBP1(S), a functional homologue of yeast Hac1p (Calfon et al., 2002; Yoshida et al., 2001a). ATF6 is converted to an active transcription factor by ER stress-induced regulated intramembrane proteolysis (Mori, 2003). A full understanding of the molecular mechanisms and biological significance of the mammalian UPR requires that both the differential as well as overlapping roles of these three transcriptional induction pathways be determined.

ATF6, consisting of the closely related ATF6α and ATF6β in mammals, is constitutively synthesized as a type II transmembrane protein in the ER, designated pATF6α/β(P) (Haze et al., 2001; Haze et al., 1999). Upon ER stress, pATF6α/β(P) relocates from the ER to the Golgi apparatus to be cleaved by the sequential action of site-1 and site-2 proteases (Nadanaka et al., 2004; Shen et al., 2002a; Ye et al., 2000). The resulting cytoplasmic fragment liberated from the membrane, designated pATF6α/β(N), enters the nucleus to activate transcription of its target genes (Yoshida et al., 2000; Yoshida et al., 2001b). We have recently generated ATF6α- and ATF6β-knockout mice, which developed normally, and found that their double knockout caused embryonic lethality (Yamamoto et al., 2007). Analysis of mouse embryonic fibroblasts (MEFs) deficient in either ATF6α or ATF6β showed that ATF6α but not ATF6β is required for transcriptional induction of not only ER chaperones but also ERAD components, and that ATF6α–/– MEFs are sensitive to ER stress. Wu et al. independently generated and characterized ATF6α-knockout mice as well as MEFs deficient in ATF6α, and reached similar but not identical conclusions (Wu et al., 2007) (see the Discussion for details).

Our previous analysis focused on selected canonical target genes of the UPR. Here, to unambiguously clarify the role of ATF6α, we performed a genome-wide search for ATF6α-target genes. Based on the results, we propose that ATF6 is a transcription factor which is specialized in the regulation of ER quality control proteins.

Experimental Procedures

Preparation, culture and transfection of ATF6α+/+ and ATF6α–/– MEFs

Male heterozygotes of ATF6α (ATF6α+/–) (Yamamoto et al., 2007) were backcrossed to female wild-type mice (C57BL/6J) eight times to obtain ATF6α N8-heterozygotes. Crosses between male and female ATF6α N8-heterozygotes were dissected on embryonic day 13.5 and MEFs were isolated by trypsinization of embryos. Primary N8-MEFs were cultured in Dulbecco’s modified Eagle’s medium (glucose at 4.5 g/liter) supplemented with 10% fetal bovine serum, 2 mM glutamine, and antibiotics (100 U/ml penicillin and 100 μg/ml streptomycin) at 37°C in a humidified 5% CO₂/95% air atmosphere. Transfection was performed using FuGENE6 (Roche) according to the manufacturer’s instructions. pECFP-N1-tsVSVG and pECFP-N1-A1AT to express tsVSVG-CFP and A1AT-CFP fusion proteins, respectively, were as described previously (Nadanaka et al., 2004). PERK+/+ and PERK–/– MEFs (Harding et al., 2000b) were the generous gift of Dr. David Ron (New York University). XBP1+/+ and XBP1–/– MEFs (Lee et al., 2003) were the generous gift of Dr. Laurie Glimcher (Harvard Medical School).

Microarray analysis

Total RNA extracted from ATF6α+/+ and ATF6α–/– N8-MEFs by the acid guanidinium/phenol/chloroform method using ISOGEN (Nippon Gene) was further purified using RNeasy Mini (Qiagen), and checked for quality with an RNA 6000 Nano Assay using an Agilent 2100 Bioanalyser (Agilent Technologies). Five hundred nanogram aliquots of total RNA prepared from N8-MEFs untreated or treated with 2 μg/ml tunicamycin for 8 h were converted to cDNA by reverse transcription. cDNA obtained from untreated and tunicamycin-treated N8-MEFs was then labeled by transcription with cyanine 3-CTP and cyanine 5-CTP, respectively, using an Agilent Low Input Linear Amplification kit. After purification through RNeasy Mini, 825 ng each of the labeled cRNA probes was mixed and hybridized with a 4×44K Agilent oligo microarray (Whole Mouse Genome), on which 44,000 mouse genes were spotted, using an Agilent Gene Expression Hybridization Kit. Cyanine 3 and cyanine 5 fluorescence intensities of a spot were obtained after subtraction of respective background intensity of the spot using a GenePix 4000B (Axon). Fold induction caused by tunicamycin treatment was defined as the ratio of cyanine 5 intensity to cyanine 3 intensity. Because our analysis was based on fold induction values, genes showing extremely low fluorescence intensity were eliminated to ensure accuracy. To this end, background intensity obtained for 44,000 spots was summed and the average background intensity was determined for both cyanine 5 and cyanine 3: if cyanine 5 or cyanine 3 intensity for a spot was less than half the respective average background intensity, the fold induction value was not determined. We carried out four independent experiments and obtained fold induction values for 14,729 of 44,000 genes at least three and mostly four times.

Northern blot hybridization

Total RNA was extracted from cultured N8-MEFs using ISOGEN. Northern blot hybridization was performed according to standard procedures (Sambrook et al., 1989). Digoxigenin (DIG)-labeled cDNA probes were prepared using PCR according to the manufacturer’s instructions (Roche) and hybridized with RNA electrophoresed and blotted on a membrane. Subsequent reaction with anti-digoxigenin antibody (Roche) and treatment with the chemiluminescent detection reagent CDP-star (GE Healthcare Biosciences) were performed according to the manufacturer’s specifications. Chemiluminescence was visualized using an LAS-3000mini LuminoImage analyzer (Fuji Film).

Immunological techniques

Immunoblotting analysis was carried out according to the standard procedure (Sambrook et al., 1989) as described previously (Okada et al., 2002) using Western Blotting Luminol Reagent (Santa Cruz Biotechnology). Chemiluminescence was detected using an LAS-3000mini LuminoImage analyzer (Fuji Film). ATF6α was detected with rabbit anti-ATF6α polyclonal antibody (Haze et al., 1999). Anti-GFP monoclonal antibody (mixture of clone 7.1 and 13.1) was purchased from Roche. Mouse anti-KDEL antibody was purchased from Stressgen. Immunoprecipitation was carried out essentially as described previously (Nadanaka et al., 2004).

Results

Identification of ATF6α-target genes

Using total RNA isolated from ATF6α+/+ and ATF6α–/– MEFs which had been untreated or treated for 8 h with tunicamycin, an inhibitor of protein N-glycosylation known to evoke ER stress (Kaufman, 1999), we conducted four independent microarray analyses, and obtained fold-induction values for 14,729 of 44,000 spotted mouse genes at least three and mostly four times (see Experimental Procedures). The number of genes induced more than 2-fold by tunicamycin treatment was 680 and 556 in ATF6α+/+ and ATF6α–/– MEFs, respectively. ATF6α targets were defined as genes whose fold induction value in ATF6α+/+ MEFs was more than 2 and in ATF6α–/– MEFs was less than half that in ATF6α+/+ MEFs. This process produced a total of 30 genes as ATF6α-target genes (Table I), which were categorized into several functional groups as detailed below.

Requirement of ATF6α for transcriptional induction of most ER chaperones

Microarray analysis showed that mRNA encoding the major ER chaperone BiP (Sitia and Braakman, 2003) was induced 11-fold in ATF6α+/+ MEFs in response to tunicamycin treatment (Fig. 1A). This induction was mitigated to less than 5-fold in ATF6α–/– MEFs. Similarly, the transcriptional induction of ER-localized molecular chaperones ORP150/GRP170, GRP94 and calreticulin (Sitia and Braakman, 2003) observed in ATF6α+/+ MEFs was greatly mitigated in ATF6α–/– MEFs (Fig. 1A). These results are consistent with our previous northern blot hybridization analysis of BiP as well as immunoblotting analysis of BiP and GRP94 (Yamamoto et al., 2007), demonstrating that ATF6α is generally required for transcriptional induction of ER-localized molecular chaperones, with the possible exception of calnexin (Sitia and Braakman, 2003) and STCH (Otterson et al., 1994). Transcriptional induction of GRP75, a mitochondrial molecular chaperone (Szabadkai et al., 2006), did not depend on ATF6α, as expected.

View Details

Fig. 1. Effect of the absence of ATF6α on transcriptional induction of molecular chaperones. (A) ATF6α+/+ (WT, open bars) and ATF6α–/– (KO, closed bars) MEFs were untreated or treated with 2 μg/ml tunicamycin for 8 h. Fold induction of various molecular chaperones indicated on the abscissa was determined by microarray analysis and is shown as the means ± S.D. of four independent experiments. P<0.01 for the difference in induction of BiP, ORP150/GRP170, GRP94 and calreticulin (CRT) between ATF6α+/+ and ATF6α–/– MEFs. There was no statistically significant difference in the induction of calnexin (CNX), STCH or GRP75. (B) Fold induction of various folding enzymes indicated on the abscissa was determined and is shown as in (A). P<0.01 for the difference in induction of ERp72, P5, GRP58 and ERO1β. There was no statistically significant difference in the induction of ERO1α. (C) ATF6α+/+ and ATF6α–/– MEFs were treated with 2 μg/ml tunicamycin (Tm, left panel) or 300 nM thapsigargin (Tg, right panel) for the indicated periods. Total RNA was isolated and analyzed by northern blot hybridization using a DIG-labeled cDNA probe specific to mouse BiP, ERO1α, ERO1β or GAPDH.

ERp72, P5 and GRP58 are protein disulfide isomerase-like folding enzymes containing thioredoxin motifs (Sitia and Braakman, 2003). As shown in Fig. 1B, transcriptional induction of these three enzymes observed in ATF6α+/+ MEFs was lost or greatly mitigated in ATF6α–/– MEFs. After oxidization of substrates, protein disulfide isomerase is reoxidized by the action of Ero1, which consists of closely related ERO1α and ERO1β in mammals (Sitia and Braakman, 2003). Microarray analysis showed that weak transcriptional induction of ERO1α occurred similarly in ATF6α+/+ and ATF6α–/– MEFs, whereas the marked transcriptional induction of ERO1β observed in ATF6α+/+ MEFs was lost in ATF6α–/– MEFs (Fig. 1B). These observations were well confirmed by northern blot hybridization analysis of MEFs treated with tunicamycin or thapsigargin, an inhibitor of ER-Ca²⁺-ATPase known to evoke ER stress (Kaufman, 1999), as shown in Fig. 1C.

p58IPK (Rutkowski et al., 2007), ERdj3 (Shen and Hendershot, 2005) and ERdj4 (Shen et al., 2002b) are DnaJ domain-containing cochaperones inducible by ER stress. Microarray analysis showed that the induction of p58IPK and ERdj3 mRNA observed in ATF6α+/+ MEFs was lost or greatly mitigated in ATF6α–/– MEFs (Fig. 2A), consistent with our previous northern blot hybridization analysis (Yamamoto et al., 2007). In contrast, microarray analysis showed that induction of ERdj4 mRNA in ATF6α–/– MEFs was nearly half that in ATF6α+/+ MEFs (Fig. 2A), though our previous northern blot hybridization analysis showed that induction of ERdj4 mRNA occurred similarly in ATF6α+/+ and ATF6α–/– MEFs (Yamamoto et al., 2007). The only difference here is that our previous northern blot hybridization analysis was conducted with MEFs isolated from N1 mice, whereas our microarray analysis was conducted with MEFs isolated from N8 mice. We therefore performed northern blot hybridization analysis of N8-MEFs and found that induction patterns of ERdj3 and ERdj4 were identical to those previously reported in N1-MEFs, as shown in Fig. 2B, indicating that the inconsistent results of microarray analysis of ERdj4 were within the acceptable range of experimental error.

View Details

Fig. 2. Effect of the absence of ATF6α on transcriptional induction of ER cochaperones. (A) Fold induction of three ER cochaperones was determined and is shown as in Fig. 1A. P<0.01 for the difference in induction of p58IPK, ERdj3 and ERdj4. (B) ATF6α+/+ and ATF6α–/– MEFs were treated with 2 μg/ml tunicamycin (Tm). Total RNA was isolated and analyzed by northern blot hybridization using a DIG-labeled cDNA probe specific to mouse BiP, ERdj3, ERdj4 or GAPDH.

Proteins analyzed in Fig. 1 and Fig. 2 represent all ER-localized molecular chaperones, cochaperones and folding enzymes induced by tunicamycin treatment in our microarray analysis. From these, we thus concluded that ATF6α is required for transcriptional induction of most ER chaperones.

Requirement of ATF6α for transcriptional induction of some ERAD components

As shown in Fig. 3A, transcriptional induction of Herp, SEL1L, HRD1, EDEM1, known components of ERAD (Kawaguchi and Ng, 2007), largely depended on ATF6α, consistent with our previous northern blot hybridization analysis (Yamamoto et al., 2007). p97/VCP, a cytosolic AAA-ATPase which plays a key role in extracting ERAD substrates (Ye et al., 2001), was not inducible during ER stress. In contrast, transcriptional induction of RAMP4, an ER membrane protein of unknown function (Hori et al., 2006), occurred similarly in ATF6α+/+ and ATF6α–/– MEFs (Fig. 3A), consistent with our previous northern blot hybridization analysis (Yamamoto et al., 2007). Similarly, transcriptional induction of Derlin-1 and Derlin-2, proteins that span the ER membrane four times and are required for ERAD (Lilley and Ploegh, 2004; Oda et al., 2006; Ye et al., 2004), was not significantly affected by the absence of ATF6α (Fig. 3A). This observation was confirmed by northern blot hybridization as shown in Fig. 3B. Further, VIMP and EDEM3, known components of ERAD (Kawaguchi and Ng, 2007), also did not depend on ATF6α for induction (Fig. 3A). Interestingly, marked induction of Derlin-3, a protein more closely related to Derlin-2 than to Derlin-1 (Oda et al., 2006), absolutely required ATF6α (Fig. 3A), and this observation was firmly confirmed by northern blot hybridization analysis (Fig. 3B). As the proteins shown in Fig. 3A are all known ERAD components which we found inducible during tunicamycin treatment in our microarray analysis, we concluded that ATF6α is required for transcriptional induction of some but not all ERAD components.

View Details

Fig. 3. Effect of the absence of ATF6α on transcriptional induction of ERAD components. (A) Fold induction of various ERAD components indicated on the abscissa was determined and is shown as in Fig. 1A. P<0.01 for the difference in induction of Herp, SEL1L, HRD1, EDEM1, and Derlin-3. P<0.05 for the difference in induction of Derlin-2 and Derlin-1. There was no statistically significant difference in the induction of p97/VCP, RAMP4, VIMP or EDEM3. (B) ATF6α+/+ and ATF6α–/– MEFs were treated and analyzed as in Fig. 1C using a DIG-labeled cDNA probe specific to mouse BiP, Derlin-1, Derlin-2, Derlin-3 or GAPDH.

Cross talk among UPR mediators

As shown in Fig. 4A, the absence of ATF6α had almost no effect on expression levels of two other UPR transducers, IRE1α and PERK. Transcriptional induction of XBP1 and ATF4, downstream transcription factors of the IRE1 and PERK pathways, respectively, occurred similarly in ATF6α+/+ and ATF6α–/– MEFs, consistent with our previous northern blot hybridization and immunoblotting analyses (Yamamoto et al., 2007). The transcription factor CHOP is known to be regulated by both ATF6 and PERK pathways (Ma et al., 2002; Okada et al., 2002). Accordingly, the transcriptional induction of CHOP observed in ATF6α+/+ MEFs was mitigated considerably in ATF6α–/– MEFs (Fig. 4A). Transcriptional induction of GADD34 (Novoa et al., 2001) and asparagine synthetase (Barbosa-Tessmann et al., 2000), known targets of the PERK pathway, was not affected by the absence of ATF6α, as expected.

View Details

Fig. 4. Effect of the absence of ATF6α on the levels of two other UPR transducers and transcriptional induction of their downstream proteins, as well as effect of the absence of PERK on activation of ATF6α. (A) Fold induction of UPR transducers and their downstream proteins indicated on the abscissa was determined and is shown as in Fig. 1A. P<0.05 for the difference in induction of CHOP. There was no statistically significant difference in the induction of XBP1, ATF4, PERK, IRE1α, GADD34 or asparagine synthetase (ASN-S). (B) PERK+/+ and PERK–/– MEFs were treated with 10 μg/ml tunicamycin (Tm) for the indicated periods. Cell lysates were prepared and analyzed by immunoblotting using anti-ATF6α or anti-KDEL antibody. The migration positions of pATF6α(P), pATF6α(P*), pATF6α(N) and BiP are marked. pATF6α(P*) denotes the nonglycosylated form of pATF6α(P). The positions of full-range rainbow molecular weight markers (GE Healthcare Biosciences) are indicated on the left.

It was previously shown that transcriptional induction of ER chaperones was significantly mitigated in MEFs deficient in PERK (Harding et al., 2000a; Wu et al., 2007) or in MEFs in which the phosphorylation site of eIF2α had been replaced by alanine (Scheuner et al., 2001), but that the downstream transcription factor ATF4 does not bind to the cis-acting ER stress-response element responsible for induction of ER chaperones (Ma et al., 2002). We therefore examined the effects of the absence of PERK on activation of ATF6α. As shown in Fig. 4B, pATF6α(P) was cleaved to produce pATF6α(N) 2 h after treatment with tunicamycin in both PERK+/+ (lane 2) and PERK–/– (lane 8) MEFs, but this activation was not sustained in PERK–/– MEFs (lanes 9–12) in contrast to the case of PERK+/+ MEFs (lanes 3–6). Accordingly, BiP was less induced in PERK–/– MEFs than in PERK+/+ MEFs. These findings indicated that the absence of PERK exerted its inhibitory activity on the induction of ER chaperones via the blocking of ATF6α activation, the mechanism of which is currently under investigation.

Dispensability of ATF6α in transcriptional induction of proteins involved in translocation into and transport from the ER

A number of proteins work together for the translocation of newly synthesized secretory and transmembrane proteins into the ER as well as their transport from the ER to the Golgi apparatus, some of which are known to be upregulated during the UPR. We therefore examined whether these translocation and transport proteins are regulated by ATF6α. As shown in Fig. 5A, microarray analysis revealed that SRP54 (Bernstein et al., 1989) showed the highest inducibility among the various proteins involved in translocation into the ER and that its induction was not affected by the absence of ATF6α. This observation was well confirmed by northern blot hybridization analysis (Fig. 5C): induction of BiP mRNA as a control was greatly mitigated in ATF6α–/– MEFs as compared with ATF6α+/+ MEFs. Further, transcriptional induction of several other proteins involved in translocation was also not dependant on ATF6α, although the extent of their induction was low.

View Details

Fig. 5. Effect of the absence of ATF6α on transcriptional induction of proteins involved in translocation into the ER and transport from the ER. (A) Fold induction of various proteins involved in translocation into the ER indicated on the abscissa was determined and is shown as in Fig. 1A. There was no statistically significant difference in the induction of any of the 9 genes. (B) Fold induction of various proteins involved in transport from the ER to the Golgi apparatus indicated on the abscissa was determined and is shown as in Fig. 1A. P<0.01 and P<0.05 for the difference in induction of VDP and SAR1a, respectively. There was no statistically significant difference in the induction of the other 6 genes. (C) ATF6α+/+ and ATF6α–/– MEFs were treated and analyzed as in Fig. 1C using a DIG-labeled cDNA probe specific to mouse BiP, SRP54, SEC23b, VDP, BET1 or GAPDH.

Similarly, three proteins involved in the ER-Golgi transport, namely SEC23b (Paccaud et al., 1996), VDP (Allan et al., 2000) and BET1 (Zhang et al., 1997), were also induced in ATF6α+/+ and ATF6α–/– MEFs, as determined by microarray analysis (Fig. 5B) and confirmed by northern blot hybridization analysis (Fig. 5C). Importantly, transcriptional induction of SEC23b, VDP and BET1 as well as SRP54 required XBP1 as their mRNA was induced in XBP1+/+ MEFs but not in XBP1–/– MEFs (Fig. 6); BiP mRNA as a control was induced similarly in XBP1+/+ and XBP1–/– MEFs, as reported previously (Lee et al., 2003; Lee et al., 2002). These results indicated that ATF6 is not required for transcriptional induction of either translocation proteins or transport proteins.

View Details

Fig. 6. Effect of the absence of XBP1 on transcriptional induction of proteins involved in translocation into the ER and transport from the ER. XBP1+/+ and XBP1–/– MEFs were treated and analyzed as in Fig. 1C using a DIG-labeled cDNA probe specific to mouse BiP, SRP54, SEC23b, VDP, BET1 or GAPDH.

We finally examined whether the absence of ATF6α affects the transport of cargo proteins from the ER to the Golgi apparatus using a temperature-sensitive mutant of vesicular stomatitis virus G protein (tsVSVG) as a model protein. At the non-permissive temperature of 39.5°C, tsVSVG is misfolded due to a point mutation in its luminal domain and retained in the ER. After shift to the permissive temperature of 32°C, in contrast, it is rapidly folded and then transported to the plasma membrane through the Golgi apparatus (Nehls et al., 2000). As shown in Fig. 7A, after temperature downshift to 32°C, tsVSVG fusing to cyan-emitting green fluorescent protein (tsVSVG-CFP) showed doublet protein bands (upper panel), which were resolved more clearly when samples were treated with endoglycosidase H (lower panel). As the endoglycosidase H-resistant form (upper migrating band) represents the protein modified at the Golgi apparatus, tsVSVG-CFP moved to the Golgi apparatus with similar kinetics in ATF6α+/+ and ATF6α–/– MEFs.

View Details

Fig. 7. Effect of the absence of ATF6α on transport of cargo proteins from the ER. (A) ATF6α+/+ and ATF6α–/– MEFs were transfected with pECFP-N1-tsVSVG to express tsVSVG-CFP fusion protein. Four hours later, transfected MEFs were incubated at the non-permissive temperature of 39.5°C for 18 h and then at the permissive temperature 32°C for the indicated periods in the absence (–) or presence (+) of 1 mM dithiothreitol (DTT). Cell lysates were prepared, treated with (+) or without (–) endoglycosidase H (Endo H), and analyzed by immunoblotting using anti-GFP monoclonal antibody. Migration positions of tsVSVG-CFP, a mixture of Endo H-resistant and -sensitive forms (upper panel) as well as the endo-H resistant form, tsVSVG-CFP (mature), and the deglycosylated form, tsVSVG-CFP (-CHO), (lower panel) are shown. (B) ATF6α+/+ and ATF6α–/– MEFs were transfected with pECFP-N1-A1AT to express A1AT-CFP fusion protein. Seventeen hours later, transfected MEFs were fed with fresh medium and then incubated at 37°C for the indicated periods in the presence of 1 mM dithiothreitol (DTT) or 2 μg/ml tunicamycin (Tm). Cell lysates were prepared and analyzed by immunoblotting using anti-GFP monoclonal antibody, whereas medium was collected and subjected to immunoprecipitation using anti-GFP monoclonal antibody. Migration positions of A1AT-CFP as well as monoclonal antibody (loading control) are shown. A1AT-CFP* denotes the nonglycosylated form of A1AT-CFP.

We also examined whether secretion of α1-antitrypsin (A1AT), a serum glycoprotein, was affected by the absence of ATF6α. We expressed A1AT-CFP fusion protein by transfection and found that A1AT-CFP was secreted similarly in ATF6α+/+ and ATF6α–/– MEFs even in the presence of 1 mM dithiothreitol, a reducing reagent known to cause ER stress (Kaufman, 1999), as shown in Fig. 7B: as A1AT contains only one cysteine residue, dithiothreitol should have no effect on the folding of A1AT (Nadanaka et al., 2004). In ATF6α+/+ or ATF6α–/– MEFs treated with tunicamycin, in contrast, A1AT was not secreted and its deglycosylated and thus misfolded form was accumulated intracellularly (Fig. 7B). These results indicated that cargo proteins are transported from the ER normally in ATF6α–/– MEFs if they are folded correctly. We concluded that transport from the ER to the Golgi apparatus was not affected by the absence of ATF6α.

Discussion

A prototype of the UPR was first discovered in the 1970s, when analysis of virus-transformed mammalian cells identified GRP78 and GRP94 as proteins inducible by glucose starvation (Shiu et al., 1977). Substantial analysis of the UPR began in the late 1980s, when the accumulation of unfolded proteins in the ER was found to trigger the induction of GRP78 (found to be identical to the major ER chaperone BiP) and GRP94 (found to be a molecular chaperone of the Hsp90 family) (Kozutsumi et al., 1988). This finding in turn implied that the UPR is a homeostatic response which maintains the protein folding environment in the ER by suppressing the proteotoxicity of accumulated unfolded proteins. The use of the budding yeast Saccharomyces cerevisiae as a model in the 1990s led to major progress in understanding the molecular mechanisms of the UPR, and to molecular cloning of the UPR transducer Ire1p present in the ER (Cox et al., 1993; Mori et al., 1993), as well as the UPR-specific transcription factor Hac1p (Cox and Walter, 1996; Mori et al., 1996); followed by discovery of Ire1p-mediated splicing of HAC1 mRNA, which connects the event in the ER to that in the nucleus (Cox and Walter, 1996; Kawahara et al., 1997). A number of ER chaperones were identified as targets of the Ire1p-Hac1p pathway in the yeast UPR (Mori et al., 1998). Importantly, microarray analysis, a new technology invented in the late 1990s, increased the number of UPR target genes drastically, to 381 out of a total 6,000 yeast genes, including not only ER chaperones but also numerous proteins working at various stages of secretion, specifically proteins involved in translocation, protein folding, protein degradation, glycosylation in the ER, lipid/inositol metabolism, ER-Golgi transport, Golgi-ER retrieval, glycosylation in the Golgi apparatus, vacuolar targeting, distal secretion and cell wall biogenesis. Based on these findings, it was proposed that activation of the yeast UPR induces specific remodeling of the secretory pathway to minimize the amount, concentration, or both of unfolded proteins in the ER (Travers et al., 2000).

In metazoan cells, the UPR transducer Ire1p is well conserved, whereas the target transcription factor of Ire1p-mediated unconventional splicing is switched from Hac1p to XBP1 (Calfon et al., 2002; Shen et al., 2001; Yoshida et al., 2001a). The IRE1-XBP1 pathway is responsible for induction of most canonical UPR target genes in C. elegance (Shen et al., 2005) and D. melanogaster (Hollien and Weissman, 2006). Interestingly, however, the IRE1-XBP1 pathway is dispensable to the induction of major ER chaperones such as GRP78/BiP and GRP94 in mammalian cells, but is required for induction of a subset of ER chaperones and most ERAD components (Lee et al., 2003; Lee et al., 2002; Oda et al., 2006; Yoshida et al., 2003), thus revealing diversity in the transcriptional induction system in mammals. Extensive genome-wide searches using microarray or chromatin immunoprecipitation analyses by several laboratories consistently showed that mammalian XBP1 is involved in the expression of not only genes working for protein folding and degradation in the ER but also many genes working for the secretory pathway and others, similarly to the case of S. cerevisiae, indicating that XBP1 is a multifunctional transcription factor (Acosta-Alvear et al., 2007; Lee et al., 2003; Shaffer et al., 2004; Sriburi et al., 2007).

The UPR transducer ATF6 is present in metazoan genomes but not in the yeast genome. Worm and fly genomes contain a single ATF6 gene, whereas mammalian genomes contain two closely related genes, ATF6α and ATF6β. Although worm and fly ATF6 has little role in the UPR (Hollien and Weissman, 2006; Shen et al., 2005), we recently showed that mammalian ATF6α but not ATF6β is required for transcriptional induction of major ER chaperones as well as of ERAD components. Detailed analysis showed that ATF6α is solely responsible for the transcriptional induction of major ER chaperones and that ATF6α heterodimerizes with XBP1 for the induction of major ERAD components (Yamamoto et al., 2007). Thus, ATF6 has gained the ability to induce the canonical UPR target genes in higher eukaryotes.

In the present study, we conducted a genome-wide search for ATF6α-target genes by microarray analysis of MEFs deficient in ATF6α. Among the 14,729 mouse genes we could successfully analyze, 680 and 556 genes were induced more than 2-fold in response to tunicamycin treatment of ATF6α+/+ and ATF6α–/– MEFs, respectively. When the following criteria were applied, however, only 30 genes were identified as ATF6α-target genes: the ATF6α-target gene had to be induced more than 2-fold in ATF6α+/+ MEFs and its fold induction value in ATF6α–/– MEFs had to be less than half that in ATF6α+/+ MEFs. Among these 30 genes, notably, 7 were ER chaperones, 5 were ERAD components and 6 were ER proteins; six ER proteins are SDF2L1 for protein O-glycosylation (Fukuda et al., 2001); CRELD2 with unknown function (Ortiz et al., 2005); ARMET with unknown function (Mizobuchi et al., 2007); PIG-A for glycosylphosphatidylinositol anchor biosynthesis (Watanabe et al., 1996); SERCA2, Ca²⁺-ATPase (Thuerauf et al., 2001); and ORMDL2 for protection from ER stress (Hjelmqvist et al., 2002) (Table I). The remaining 12 targets were miscellaneous and could not be further categorized. ATF6α therefore appears to have limited function as compared with XBP1, and may regulate ER quality control proteins primarily and specifically.

It is particularly noteworthy that proteins involved in translocation into the ER (SRP54, SSR3 and SPCS3) as well as ER-Golgi transport (SEC23b, VDP and BET1) are not ATF6α targets, but rather XBP1-targets (Figs. 5 and 6). We did not observe any defects in the transport of cargo proteins (tsVSVG and A1AT) out of the ER to the plasma membrane or extracellular space via the Golgi apparatus in ATF6α–/– MEFs as compared with ATF6α+/+ MEFs (Fig. 7). Wu et al. independently generated ATF6α-knockout mice and characterized ATF6α-deficient MEFs (Wu et al., 2007). Their conclusions were similar to ours as far as regulation of ER chaperones and ERAD components are concerned. Nonetheless, they found that transport of transferin receptor from the ER to the Golgi apparatus or secretion of alkaline phosphatase into the extracellular space was inefficient in ATF6α–/– MEFs treated with thapsigargin as compared with untreated ATF6α–/– MEFs or ATF6α+/+ MEFs treated with thapsigargin, leading them to propose that ATF6α is required to optimize protein folding, secretion and degradation during ER stress. In their microarray analysis, however, only Rab38 was identified as an ATF6α-target among numerous proteins involved in vesicular transport if our criteria for ATF6α-target genes were applied, and Rab38 appeared to control post-Golgi trafficking (Wasmeier et al., 2006). In our analysis, signals of Rab38 were extremely weak and thus Rab38 was excluded from the 14,729 genes we analyzed. Therefore, the molecular basis of their proposal regarding secretion defects of ATF6α–/– MEFs seems unsupported. Rather, we suspect that the inefficient transport of transferin receptor or alkaline phosphatase in ATF6α–/– MEFs treated with thapsigargin as compared with similarly treated ATF6α+/+ MEFs may be due to folding defects resulting from inefficient induction of ER chaperones. It is markedly difficult to separate transport defects from folding defects experimentally, because only correctly folded cargo proteins are transported. In our experiments with tsVSVG (Fig. 7A), transfected cells were incubated at the non-permissive temperature 39.5°C for 18 h to cause misfolding and retention of tsVSVG in the ER, which should have evoked significant ER stress in both ATF6α+/+ and ATF6α–/– MEFs. Yet, tsVSVG was transported to the Golgi apparatus similarly in ATF6α+/+ and ATF6α–/– MEFs after shift to the permissive temperature 32°C. We therefore concluded that ER-Golgi transport is not regulated by ATF6α and that ATF6α is a transcription factor which is specialized in regulating the expression of ER quality control proteins.

Based on the present and previously published results, we envision the following scenario to explain the evolution of the UPR. First, the IRE1 pathway evolved to counteract the accumulation of unfolded proteins in the ER. Because cells such as S. cerevisiae keep synthesizing proteins even under ER stress conditions at this evolutional stage, activated IRE1 induces transcription of not only ER chaperones and ERAD components but also numerous proteins working at various stages of secretion to minimize the amount, concentration or both of unfolded proteins accumulated in the ER (Travers et al., 2000). Next, the PERK pathway emerged via gene shuffling between IRE1 and GCN2: GCN2 encodes a protein kinase which phosphorylates eIF2α in response to amino acid starvation (Silverman and Williams, 1999). Cell such as C. elegance would now be able to attenuate translation and decrease the burden on the ER, which was perhaps advantageous in handling unfolded proteins in multicellular organisms, as these encounter not only environmental but also physiological ER stress during development or differentiation. The ATF6 pathway then evolved, which is specialized in regulating the transcription of ER quality control proteins. Because of the difference between the IRE1 and ATF6 pathways in their mechanisms of activating downstream transcription factors, cells such as those in mammals are able to perform a time-dependent phase shift to cope with ER stress in accordance with the quality, quantity, or both of unfolded proteins accumulated in the ER (Yoshida et al., 2003). Finally, various tissue-specific UPR transducers, such as IRE1β (Bertolotti et al., 2001) and ATF6-like membrane-bound transcription factors, i.e. OASIS (Kondo et al., 2005), CREBH (Zhang et al., 2006), Luman/LZIP (Raggo et al., 2002) and Tisp40 (Nagamori et al., 2005), were developed to strengthen local quality control capability in the ER. It appears that ER stress is so threatening to the cell that ever more sophisticated tactics were created in accordance with the complexity of biological systems in the cell or organism constructed during evolution.

Acknowledgements

We are grateful to Dr. D. Ron for his provision of PERK+/+ and PERK–/– MEFs and Dr. L. Glimcher for XBP1+/+ and XBP1–/– MEFs. We thank Ms. Kaoru Miyagawa for her technical and secretarial assistance. This work was supported in part by grants from the Ministry of Education, Culture, Sports, Science and Technology of Japan (15GS0310 and 19058009 to K. M.).