Cytoplasmic IRE1α-mediated XBP1 mRNA Splicing in the Absence of Nuclear Processing and Endoplasmic Reticulum Stress*

Accumulation of unfolded proteins in the endoplasmic reticulum (ER) activates an intracellular signal transduction program termed the unfolded protein response (UPR). In mammalian cells, the UPR is signaled in part through dimerization of ER membrane-localized IRE1α to activate its protein kinase and endoribonuclease activities. Activated IRE1α cleaves XBP1 mRNA at two sites to initiate an unconventional splicing reaction. The 5′ and 3′ fragments are subsequently joined by an RNA ligase activity, thereby removing a 26-base intron. This splicing reaction creates a translational frameshift to produce a functional XBP1 transcription factor. However, the cellular location and physiological processes required for splicing of XBP1 mRNA are not well characterized. To study these processes, XBP1 mRNAs were engineered in which translation of enhanced green fluorescence protein or luciferase required splicing of the XBP1 intron. Using cell lines that continuously or transiently express these reporter constructs, we show that cytoplasmic unspliced XBP1 mRNA is efficiently spliced by activated IRE1α and requires ongoing cellular transcription but not active translation. The XBP1 intron was effectively removed from RNA substrates transcribed from T7 RNA polymerase or delivered directly to the cytoplasm by RNA transfection, thus indicating that the splicing reaction does not require nuclear processing of the RNA substrate. Analysis of nuclear and cytoplasmic RNA fractions demonstrated that XBP1 mRNA splicing occurs in the cytoplasm. Moreover, an artificial Fv-IRE1αΔN was engineered that was able to splice XBP1 mRNA upon chemical-induced dimerization. These findings demonstrate that IRE1α dimerization is sufficient to activate XBP1 mRNA splicing in the absence of the UPR. We propose that XBP1 mRNA cytoplasmic splicing provides a novel mechanism to rapidly induce translation of a transcription factor in response to a specific stimulus.

Accumulation of unfolded proteins in the endoplasmic reticulum (ER) activates an intracellular signal transduction program termed the unfolded protein response (UPR). In mammalian cells, the UPR is signaled in part through dimerization of ER membrane-localized IRE1␣ to activate its protein kinase and endoribonuclease activities. Activated IRE1␣ cleaves XBP1 mRNA at two sites to initiate an unconventional splicing reaction. The 5 and 3 fragments are subsequently joined by an RNA ligase activity, thereby removing a 26-base intron. This splicing reaction creates a translational frameshift to produce a functional XBP1 transcription factor. However, the cellular location and physiological processes required for splicing of XBP1 mRNA are not well characterized. To study these processes, XBP1 mRNAs were engineered in which translation of enhanced green fluorescence protein or luciferase required splicing of the XBP1 intron. Using cell lines that continuously or transiently express these reporter constructs, we show that cytoplasmic unspliced XBP1 mRNA is efficiently spliced by activated IRE1␣ and requires ongoing cellular transcription but not active translation. The XBP1 intron was effectively removed from RNA substrates transcribed from T7 RNA polymerase or delivered directly to the cytoplasm by RNA transfection, thus indicating that the splicing reaction does not require nuclear processing of the RNA substrate. Analysis of nuclear and cytoplasmic RNA fractions demonstrated that XBP1 mRNA splicing occurs in the cytoplasm. Moreover, an artificial F v -IRE1␣⌬N was engineered that was able to splice XBP1 mRNA upon chemical-induced dimerization. These findings demonstrate that IRE1␣ dimerization is sufficient to activate XBP1 mRNA splicing in the absence of the UPR. We propose that XBP1 mRNA cytoplasmic splicing provides a novel mechanism to rapidly induce translation of a transcription factor in response to a specific stimulus.
The endoplasmic reticulum (ER) 3 is a network of interconnected tubules, vesicles, and sacs that serve many specialized functions in the cell: including calcium storage and gated release, biosynthesis of membrane and secretory proteins, and production of lipids and sterols. Given the importance of ER function for normal cellular function, it is not surprising that the ER affects a diverse number of cellular processes such as gene expression (at the transcriptional and translational levels), cell cycle control, intracellular signaling, and programmed cell death. If ER homeostasis is altered, signaling pathways are activated to elicit an adaptive process called the unfolded protein response (UPR) (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12). The UPR activates the transcription of ER stress-response genes, including BiP (GRP78), GRP94, CHOP-10 (GADD153), XBP1, and EDEM (13)(14)(15)(16). These genes encode functions that assist protein folding and secretion, facilitate degradation of misfolded proteins in the ER lumen, or induce cell-death pathways.
The initial characterization of the UPR was performed in the budding yeast Saccharomyces cerevisiae (17,18). A genetic screen revealed that three genes are required for signal transduction from the ER to activate gene transcription in the nucleus. The first, IRE1, encodes a type 1 transmembrane Ser/Thr protein kinase that also has a site-specific endoribonuclease (RNase) activity. The second, HAC1, encodes a basic leucine zipper transcriptional activator for UPR target genes (19,20). Finally, the product of RLG1 (tRNA ligase) joins the 5Ј and 3Ј ends of HAC1 mRNA that are produced through cleavage by Ire1p (21,22). IRE1, the most proximal component of the UPR pathway, has three functional domains (17,18,22,23). The amino-terminal domain resides in the ER lumen and senses the accumulation of unfolded proteins. Accumulation of misfolded proteins in the ER lumen induces dimerization of the IRE1 luminal domain that activates trans-autophosphorylation via the cytoplasmic kinase domain. The activation of the kinase domain then stimulates the site-specific carboxyl-terminal RNase activity (23,24). In yeast, the only known substrate for the site-specific RNase activity of Ire1p is HAC1 mRNA (19,20,25). HAC1 mRNA cleavage by Ire1p initiates an unconventional splicing reaction that is required for efficient translation and accumulation of Hac1p. Unspliced HAC1 mRNA is constitutively transcribed, but the RNA is not translated due to long-range base pairing interactions between the HAC1 5Ј-untranslated region and the intron (26). Upon removal of the intron, HAC1 mRNA is efficiently translated. The splicing of HAC1 mRNA changes the open reading frame of HAC1 so the carboxyl terminus is replaced by a short transcriptional activation domain (27).
Although the UPR was first described in the budding yeast S. cerevisiae (4), higher eukaryotic cells have also maintained the essential and unique properties of IRE1 signaling. However, higher eukaryotes possess additional sensors and control mechanisms to mediate more diverse responses upon activation of the UPR. Three transmembrane proteins, IRE1, PERK, and ATF6, serve as proximal sensors that regulate components acting to attenuate translation, up-regulate the capacity of the ER to fold newly synthesized proteins, and degrade unfolded or misfolded proteins (2)(3)(4).
There are two homologues of yeast Ire1p, referred to as IRE1␣ and IRE1␤, in both the human and murine genomes. IRE1␣ is expressed in most cells and tissues (28), where IRE1␤ expression is restricted to intestinal epithelial cells (29). Both IRE1 proteins respond to the accumulation of unfolded proteins in the ER, undergo kinase activation and, thereby, trigger their RNase activities. Each IRE1 protein cleaves its own mRNA (30), 28 S rRNA (31), and the mRNA encoding the basic leucine zipper transcription factor X-box binding protein 1 (XBP1) (32)(33)(34). Upon activation of the UPR, unconventional splicing initiated by IRE1␣ removes a 26-nucleotide intron from unspliced mammalian XBP1 mRNA (encoding 267 amino acids), to induce a translational frameshift producing a fusion protein encoded from two evolutionarily conserved open reading frames. The translation product from spliced mammalian XBP1 mRNA is a potent transcriptional activator of 371 amino acids that comprises the original amino-terminal DNA binding domain plus an additional transactivation domain in the carboxyl terminus.
This IRE1-mediated splicing mechanism is conserved through evolution as mammalian IRE1 recognizes and cleaves yeast HAC1 mRNA (28). In yeast, the splicing mechanism used by Ire1p is similar to pre-tRNA splicing (35,36). Activated Ire1p cleaves the 5Ј and 3Ј exon-intron junctions in HAC1 mRNA to generate 5Ј-OH and 2Ј,3Ј-cyclic PO 4 ends. tRNA ligase (RLG1/TRL1) joins both exons (21). The ligase leaves a 2Ј-phosphate derived from the splice junction on the 5Ј-end of the joined junction (35). This 2Ј-phosphate is removed by phosphatase Tpt1p (37) in a reaction that requires NAD ϩ as a phosphate acceptor to generate nicotinamide and ADP-ribose 1Ј,2Ј-cyclic phosphate (38). The presence of RNA ligases with similar activities has been demonstrated in wheat germ, Chlamydomonas, and mammalian cells (39). Mammals also have an additional ligase activity that incorporates the junction phosphate into the spliced mRNA (40). Finally, a NAD ϩ -dependent 2Ј-phosphatase is conserved in bacteria, yeast, plants, and mammals (37,41). Therefore, although the identity of the XBP1 mRNA ligase remains unknown, the mechanism of IRE1-mediated RNA splicing is likely conserved in all eukaryotes.
To elucidate the mechanism of UPR signaling, it is essential to identify where and how the IRE1␣-mediated splicing reaction occurs in the cell. In yeast, it is now believed that pre-tRNA splicing occurs not in the nucleus (42,43) but in the cytoplasm because the tRNA splicing RNase that catalyzes intron removal from tRNA precursors is associated with the outer mitochondrial membrane (44). This unexpected cytoplasmic localization of a key RNA splicing enzyme explains several previously puzzling observations, such as why unspliced pre-tRNAs accumulate in nuclei of yeast cells defective in tRNA nuclear export upon mutation of the Ran transport system (e.g. RNA1) or the tRNA export receptor LOS1 (45)(46)(47). Even though it was reported that tRNA ligase Rlg1p is localized to the nuclear pore complex in yeast (48), more recent reports show there is a functional pool of tRNA ligase activity in the cytoplasm (26). In addition, a carboxyl GFP-tagged ligase (NAD ϩ -dependent 2Ј-phosphotransferase, TPT1) that catalyzes the third step of pre-tRNA splicing, removal of the residual 2Ј-phosphate (49), is distributed throughout the cytoplasm (50). Finally, in yeast, HAC1 mRNA splicing can occur on polyribosomes but does not require polysomal association. The cytoplasmic pool of HAC1 mRNA can be spliced independently of transcription (26). However, it is unknown where and how efficient splicing of XBP1 mRNA occurs in UPR-activated mammalian cells. Membrane fractionation experiments show that the majority of endogenous IRE1␣ protein is localized to the inner leaflet of the nuclear envelope (28,51). In this case, the kinase/RNase domain is exposed to the nucleoplasm and could participate in XBP1 mRNA splicing as part of the nuclear export pathway. In this report, we studied the requirements for transcription and translation for XBP1 splicing and the potential for XBP1 splicing to occur in the cytoplasm using versatile reporter systems for XBP1 splicing. Through the use of four different approaches, we show that the cytoplasmic pool of XBP1 mRNA is an efficient substrate for the splicing reaction. Efficient XBP1 splicing requires ongoing transcription; however, once the IRE1␣ RNase is activated the reaction does not require mRNA translation or ER stress.
To generate the F v -IRE1␣ fusion protein, the portion of the mIRE1␣ cDNA encoding both kinase and RNase domain (amino acids 468 -977) was fused to one modified FK506 binding domain (F v 1E) from plasmid pC 4 -F v 1E (ARIAD Pharmaceuticals, Inc.). The resultant vector pF v -IRE⌬N produces a cytosolic F v -IRE⌬N-HA protein. The plasmid pC 4 -F v 1E was treated with SpeI and then ligated to mIRE1␣⌬N cDNA treated with SpeI. The mIRE1␣⌬N cDNA was isolated from pcDNA3.1-mIRE1␣ (54) by PCR with two primers: 5Ј-TTTCCTAC-TAGTCTGAGCGTGCATCAGCAG-3Ј and 5Ј-TTTCCTACTAGT-GAGGGCATATGGAATCACTG-3Ј. To express nuclear-or membrane-localized F v -IRE⌬N proteins, plasmids pNLS-F v and pMyr-F v were generated. To construct plasmid pNLS-F v , F1E of pC 4 EN-F1E (ARIAD Pharmaceuticals, Inc.) was replaced by F v 1E of pC 4 -F v 1E. To generate plasmid pMyr-F v , F v 2E of pC 4 M-F v 2E (ARIAD Pharmaceuticals, Inc.) was replaced by F v 1E of pC 4 -F v 1E. Plasmids pNLS-F v -IRE⌬N and pMyr-F v -IRE⌬N were constructed by inserting mIRE1␣⌬N cDNA fragments treated with SpeI from pF v -IRE⌬N into the SpeI sites of pNLS-F v and pMyr-F v , respectively.
RT-PCR and Analysis-The isolation of nuclear and cytoplasmic RNA is described under " RNA Extraction and Northern Blot Analysis." Total RNA was isolated from Tm (10 g/ml)or thapsigargin (Tg) (500 nM)-treated CHO/mXBP1⌬C(un)-d2EGFP or NIH3T3 cell lines using the TRIzol reagent (Invitrogen) as specified by the manufacturer. Where indicated, CHO/mXBP1⌬C(un)-d2EGFP cells were pretreated with actinomycin D (Act D, 5 g/ml) for 15 min (for Tm) or 1 h (for Tg) prior to treatment with Tm or Tg in the presence of Act D. The cells were also pretreated with ␣-amanitin (54 M, Sigma), 5,6-dichloro-1-␤-D-ribofuranosylbenzimidazole (DRB) (156 M, Sigma), or cycloheximide (10 g/ml, Sigma) for 1 h prior to treatment with Tg (500 nM) in the presence of each drug. The splicing of mXBP1⌬C(un)-EGFP RNA was detected by standard RT-PCR (reverse transcription system (Promega) and the Expand high fidelity PCR system (Roche)) using a random primer and then the specific primers: mXBP1-382 (5Ј-ACACGCTTGGGAATGGACAC-3Ј) and EGFP-870 (5Ј-GGTGGT-CACGAGGGTGGGCCAG-3Ј). PCR products were separated by electrophoresis on 3% agarose gels and visualized by ethidium bromide staining. Densitometry scanning was performed and quantified with NIH Image software. The percent of spliced mXBP1 mRNA(S) in Figs. 2A and 5A was calculated as 100 ϫ (XBP1(S) ϩ 0.5XBP1(H))/(XBP1 (S) ϩ XBP1(H) ϩ XBP1(Un)). RT-PCR with primers FLAG-XBP(ϩ) (5Ј-AAAGACGACGACGACAAGC-3Ј) and Fluc-XBP(Ϫ) (5Ј-CCT-TATGCAGTTGCTCTCC-3Ј) was used to detect spliced mXBP1 transcripts in NIH3T3 cells transfected with pRL-IXFL or pRL-XFL (Fig.  3D). RT-PCR (Fig. 6E) and quantitative real-time RT-PCR ( Fig. 6F) were used to measure endogenous mXBP1 transcripts in NIH3T3 cells transfected with artificial F v -IRE⌬N proteins as previously described (57,58). Spliced mXBP1 mRNA (S) levels were normalized to 18 S rRNA. The relative levels of mXBP1 splicing in NIH3T3 cells expressing the artificial F v -IRE⌬N fusion proteins were normalized to the levels of cells transfected with the F v vectors lacking the IRE⌬N sequence. For NIH3T3 cells treated with/without Tm (10 g/ml, 6 h), the amount of mXBP1 splicing measured in Tm-treated cells was divided by the level measured in cells not treated with Tm.
For immunoblot analyses from NIH3T3 cells expressing artificial F v -IRE⌬N proteins, the cells were plated onto 10-cm dishes the day before transfection. For transfection, a mixture of plasmid DNA (8 g) and 24 l of FuGENE 6 (Roche) was added to NIH3T3 cells for 24 h. Cells were then incubated in culture medium with/without AP20187 (2 nM) for the indicated times. The cells were harvested and stored at Ϫ80°C for Western blot analysis. For immunoblotting, the cells were lysed in 2% SDS lysis buffer (2% SDS, 100 mM Tris-Cl, pH 6.8, 0.5 mM sodium vanadate, 100 mM NaF, 50 mM ␤-glycerolphosphate, 1 mM phenylmethylsulfonyl fluoride, protease inhibitors), sonicated on ice, boiled 15 min, and then clarified by centrifugation at 13,000 ϫ g for 10 min. Cellular proteins (70 g) were resolved on 4 -20 or 6% SDS-polyacrylamide gels and transferred to polyvinylidene difluoride membranes. Immunoblot analyses were performed with a 1/500 dilution of anti-HA monoclonal antibody (Santa Cruz Biotechnology) and a 1/5000 dilution of anti-actin monoclonal antibody (ICN Biomedical).
For analysis of phosphatase-treated proteins, the cells were lysed in 1 ml of RIPA buffer (50 mM Tris-Cl, pH 7.5, 1% sodium deoxycholate, 1% Triton X-100, 0.1% SDS, 150 mM NaCl, protease inhibitors), sonicated on ice, and then clarified by centrifugation at 13,000 ϫ g for 10 min. Soluble proteins were immunoprecipitated for 14 h with 2 g of the anti-HA antibody (Santa Cruz) bound to 20 l of protein G-agarose (Pierce) and washed two times for 5 min with RIPA buffer. The samples were then washed two times for 5 min with washing buffer (50 mM Tris-Cl, pH 7.5, 1% Triton X-100, 100 mM NaCl, 0.1 mM EGTA, 2 mM dithiothreitol, protease inhibitors). The immunoprecipitates were treated with -phosphatase in accordance with the manufacturer's specifications (New England Biolabs) at 30°C for 1 h, and immunoblot analysis was performed with anti-HA monoclonal antibody as described above.
In vitro translation reactions were previously described (60). The samples were analyzed by 12.5% SDS-PAGE followed by autoradiography.
DNA/RNA Transfections and Fluorescence Microscope Analysis-Cells were plated onto 10-cm plates (for NIH3T3 and 3126-2 cells) or 35-mm plates (for DHFR-deficient CHO and CHO/mXBP1⌬C(un)-d2EGFP cells) the day before transfection. For DNA transfection into 3126-2 cells, a mixture of pTM1-mXBP1⌬C(un/s)-EGFP (9 g), pRL-CMV (0.1 g), and 27 l of FuGENE 6 (Roche) was added for 12 h. For DNA transfection into CHO/mXBP1⌬C(un)-d2EGFP cells, each plasmid DNA (2 g) expressing a modified FK506 binding domain (F v 1E) fusion protein was mixed with 6 l of FuGENE 6 (Roche) and applied to the cells for 24 h. For RNA transfection into DHFR-deficient CHO cells, a mixture of capped-mXBP1⌬C(un/s)-EGFP-(A) 30 mRNA (2.5 g), capped-Renilla luciferase-poly(A) n mRNA (0.1 g), and 15 g of DOTAP (Roche) was added to the cells for 4 h. The cells were then further incubated with fresh medium with or without Tm (10 g/ml) for 24 h (for 3126-2 cells) and 7 h (for DHFR-deficient CHO cells) or with AP20187 (2 ng/ml) for the indicated times (for CHO/mXBP1⌬C(un)-d2EGFP cell lines). At 36 (for 3126-2 cells) or 11 h (for DHFR-deficient CHO cells) post-transfection, the culture medium was removed, and the cells were washed three times with PBS. The cells were harvested and stored at Ϫ80°C for both Renilla luciferase assay and Western blot analysis. All transfections were performed at least three times to obtain standard deviations and mean values. Cell pellets were thawed on ice and suspended in Renilla luciferase assay lysis buffer. One-fifth of the total lysate was used for the Renilla luciferase assay and the rest was sonicated. The sonicated cell lysates were centrifuged 10,000 ϫ g for 10 min. Cell proteins (70 g) were electrophoresed on SDS-polyacrylamide gels for immunoblotting of EGFP and actin.
For fluorescence microscopy, cells growing on coverslips were fixed with 3.5% (w/v) paraformaldehyde (Sigma) at room temperature for 15 min. The cells were then washed three times with PBS, permeabilized in 0.1% Triton X-100 at room temperature for 2 min, and washed three times with PBS. For immunostaining of HA epitope-tagged proteins in Fig. 6G, the samples were blocked in PBS containing 1% bovine serum albumin for 30 min at room temperature and then incubated with anti-HA antibody (1/500 dilution) for 1 h at room temperature. After being washed with PBS, the samples were reacted with rhodamine TRITC-conjugated goat anti-mouse IgG (1/500 dilution, Jackson ImmunoResearch Laboratories, Inc.) for 1 h at room temperature. The samples were then treated with 4Ј,6-diamidino-2-phenylindole for 2 min at room temperature. Finally, the coverslips were washed three times with PBS, placed on a glass slide, and sealed with transparent nail polish. The images were captured with a cooled charge-coupled device camera and a Zeiss LSM 510 confocal microscope (Zeiss). The captured images were analyzed to obtain the percentage of GFP-positive cells from HA epitope-positive cells (TRITC-labeled cells). Total counted cell numbers were over 100 HA epitope-positive cells for each analysis.
In Fig. 1B, green fluorescence images from living cells treated for 5 h with Tm (10 g/ml) were acquired by using a Nikon Eclipse TE300 inverted fluorescence microscope with a charge-coupled device camera controlled by SimplePCI software (C-Imaging Inc.). All data were processed with Adobe photoshop software (Adobe, Mountain View, CA).
Renilla and Dual Luciferase Assay-To obtain transfection efficiencies for transient DNA and RNA transfections of Fig. 4, B and C, pRL-CMV DNA (Fig. 4B) or capped-Renilla luciferase-poly(A) n mRNA ( Fig. 4C) were also combined into the transfection mixtures. The saved cell lysates were used for measuring the protein amount by using the DC protein assay kit (Bio-Rad) and analyzed for luciferase activities by using the Renilla luciferase assay system according to the manufacturer's instructions (Promega). Relative transfection efficiencies were obtained by the value of Renilla luciferase activity divided by the protein amount in lysates. All transfections were performed at least three times to obtain S.D. Ϯ mean.
DHFR-deficient CHO cells were plated on 10-cm dishes the day before transfection. Opti-MEM I media (0.75 ml) (Invitrogen) containing 2 g of either pRL-IXFL or pRL-XFL was mixed with 6 l of FuGENE 6 (Roche) and added to the cells for 31 h. The cells were then further incubated with fresh medium with or without Tm (10 g/ml) for 7 h. At 38 h post-transfection, the cells were washed three times with PBS, harvested, and stored at Ϫ80°C for dual luciferase assay, or used directly for total RNA isolation by the TRIzol reagent for RT-PCR analysis. Dual-luciferase assay (Promega) was carried out as described previously (60).
Wild-type and Ire1␣ Ϫ/Ϫ MEFs were nucleofected with Nucleofector TM technology according to the manufacturer's instructions (Amaxa Biosystems). Briefly, the primary MEFs (2 ϫ 10 6 cells) were harvested, washed with phosphate-buffered saline, and then resuspended in the MEF 1 Nucleofector kit solution with the reporter plasmid pRL-IXFL (2 g) and the pED-hIRE1␣-Flag/pED⌬C (6 g) vector. After nucleofection, the cell/DNA mixture was divided and cultured into two 10-cm dishes for 24 h. The cells were then further incubated with fresh medium with or without Tm (10 g/ml) for 8 h. At 32 h post-transfection, the cells were washed once with cold PBS, harvested, and stored at Ϫ80°C for dual luciferase assay (Promega) described above (60).
Cellular Fractionation-Cells were fractionated as described in Topisirovic et al. (61,62). In brief, cells were trypsinized, rinsed twice in ice-cold 1ϫ PBS (pH 7.4), and resuspended with slow pipetting in lysis buffer B (10 mM Tris, pH 8.4, 140 mM NaCl, 1.5 mM MgCl 2 , 0.5% Nonidet P-40, 1 mM dithiothreitol, SUPERsae⅐In TM (20 units/l) (Ambion)). Nuclear suspensions were centrifuged at 1,000 ϫ g for 3 min at 4°C, and the supernatants were saved (cytoplasmic fractions). Nuclear pellets were resuspended in lysis buffer B. One-tenth volume of detergent (3.3% (w/v) sodium deoxycholate, 6.6% (v/v) Tween 40) was added under slow vortexing, and the nuclear suspensions were incubated on ice for 5 min. Nuclei were pelleted by centrifugation at 1,000 ϫ g for 3 min at 4°C, and the supernatant (postnuclear fraction) was saved and added to the cytoplasmic fraction. Together, these are considered the cytoplasmic fraction. The nuclei were rinsed once in lysis buffer B. This protocol yielded intact nuclei as determined by light microscopy, with no significant cytoplasmic contamination as determined from the tRNA Lys content (see Fig. 5C).
RNA Extraction and Northern Blot Analysis-Total RNA and RNA from the cytoplasmic and nuclear fractions were isolated by a TRIzol procedure as specified by the manufacturer (Invitrogen). RNAs were treated with the TURBO DNA-free TM kit (Ambion) to remove contaminating DNA. Purified RNA was quantified by spectrophotometry. A 5-g portion of total and fractionated RNA was loaded onto a 1% formaldehyde/agarose gel and subsequently transferred to a positively charged nylon membrane (Amersham Biosciences). Membranes were prehybridized in ULTRAhyb buffer (Ambion) for 1 h at 45°C and probed with biotinylated mouse tRNA Lys antisense oligonucleotide probe (5Ј-CTCATGCTCTACCGACTGAGCTAGCCGG-GC-3Ј (30 pM)), biotinylated mouse U6 antisense oligonucleotide probe (5Ј-GAATTTGCGTGTCATCCTTGCGCAGGGGCCATGCTAA-3Ј (30 pM)), and mouse glyceraldehyde-3-phosphate dehydrogenase cDNA probe (5 pM) (Ambion) biotinylated using the BrightStar Psoralen-Biotin kit (Ambion) in the same buffer for 16 h at 45°C. Signals were detected using CDP-Star TM chemiluminescence (Ambion) as specified by the manufacturer.

RESULTS
Derivation of a Cell Line That Reports IRE1␣ Activation-We developed a stable CHO cell line that expresses an mRNA encoding the murine XBP1 amino-terminal fragment, including the 26-base intron, fused to the green fluorescence protein reporter d2EGFP. Under non-stressed conditions, the d2EGFP open reading frame within the reporter mRNA is not translated due to a stop codon downstream of the mXBP1⌬C(un) coding region. However, under conditions of ER stress, removal of the 26-nt intron produces a translational frameshift to result in an mXBP1⌬C(s)-d2EGFP fusion protein (Fig. 1A). This construct was transfected into DHFR-deficient CHO cells, and a cell line was then selected as described under "Experimental Procedures." Because the d2EGFP-fused mXBP1⌬C(s) protein is translated in cells that activate the IRE1␣ signaling pathway, UPR activation can be monitored by analysis of d2EGFP fluorescence in living cells (Fig. 1B).
Under non-stress conditions, less than 5% of cells had detectable mXBP1⌬C(s)-d2EGFP fusion protein, analyzed by fluorescence microscopy and fluorescence-activated cell sorting (Fig. 1B, and data not shown). After treatment with an ER stress-inducer (Tm), fluorescence microscopy and fluorescence-activated cell sorting detected a time-dependent increase in greater than 85% of the cells harboring mXBP1⌬C(s)-d2EGFP fusion protein (data not shown). The GFP fluorescence was localized to the nuclei and was variable in intensity (Fig.  1B). mXBP1⌬C(s)-d2EGFP protein is localized to the nucleus because the amino-terminal of mXBP1⌬C(s) contains the DNA-binding domain (63)(64)(65). The fluorescence data ( Fig. 1B and data not shown) correlated with the results of Western blot (Fig. 1C) and RT-PCR analysis ( Fig. 2A). Although there was a low level of mXBP1⌬C(s)-d2EGFP protein expression in the absence of an ER stress-inducing agent, the fluorescence intensity, expression, and splicing of mXBP1⌬C(un)-d2EGFP mRNA all significantly increased after Tm treatment (5 h in  Figs. 1C with 2A). In addition to the RT-PCR products derived from the spliced and unspliced transcripts, there was a slower migrating band (H) that increased upon induction of the UPR. This third slower migrating band was shown to represent a hybrid structure of unspliced and spliced single-stranded DNAs produced during PCR (66). We also confirmed that the slower migrating band (H) represents a hybrid structure. These results show that most of the unspliced reporter mRNAs were spliced and only a small amount of unspliced mRNA remained after 10 h of Tm treatment. Similar results were obtained after treatment with several UPR inducers including Tg and dithiothreitol (Fig. 2, C, lanes 1-4, in D, lanes 1-6, and data not shown). In addition, overexpression of wild-type IRE1␣ increased d2EGFP fluorescence (data not shown) and firefly luciferase activity from the pRL-IXFL reporter construct (Fig. 3E), whereas overexpression of an IRE1␣ RNase mutant (K907A) had no effect (data not shown). We conclude that XBP1 mRNA splicing is not restricted to transcripts derived from the endogenous XBP1 locus and that this reporter cell system is useful to measure UPR activation signaled through the IRE1␣ pathway in intact cells.

IRE1␣-dependent XBP1 mRNA Splicing Requires Cellular Transcription-
Although studies have demonstrated that ER stress induces IRE1mediated splicing of XBP1 mRNA in mammalian cells, the cellular compartments and molecular processes required for efficient splicing are unknown. Therefore, we examined the requirement for mRNA transcription and translation in XBP1 mRNA splicing.
To determine the requirement for transcription in IRE1␣-dependent XBP1 mRNA splicing, we used Act D to block transcription in mammalian cells. Act D, at a high concentration (5 g/ml), intercalates into DNA and inhibits all three classes of RNA polymerase transcription (67,68). RT-PCR analysis of total mRNAs prepared at various times after Act D treatment in the absence or presence of Tm demonstrated that spliced transcripts (S) exist at a reduced abundance (Fig. 2B). The intensity of the RT-PCR product representing spliced mXBP1⌬C(s)-d2EGFP gradually increased to 4 h and then declined. We then determined whether this substantial reduction in spliced XBP1 mRNA correlated with reduced activation of IRE1␣ protein. Because IRE1␣ RNase is activated by dimerization-induced trans-autophosphorylation (30,69,70), we examined IRE1␣ activation by the upward mobility shift of the IRE1␣ protein upon SDS-PAGE. Upon Tm treatment, IRE1␣ autophosphorylation occurred rapidly and was stable during Tm treatment (Fig. 2G,  lanes 1-5). However, cells treated with Act D reproducibly displayed a CHO cells that express the reporter mRNA were treated with Tm (10 g/ml) for the indicated periods of time. The expression of the mXBP1⌬C(s)-d2EGFP protein was directly detected by the green fluorescence in living cells (B) or by Western blot analysis for the mXBP1⌬C(s)-d2EGFP fusion protein using anti-EGFP antibody (C). As a loading control, the same membrane was reprobed with anti-actin antibody. The phase-contrast images were merged with the corresponding green fluorescence images in the lower panels of B. The relative ratios in C represent an average of three independent experiments. slower and reduced autophosphorylation of IRE1␣ (Fig. 2G, compare  lanes 1-4 and 6-9). These results suggest that Act D treatment reduces both IRE1␣ trans-autophosphorylation and RNase activity.
The activation of IRE1␣ kinase/RNase by Tm likely requires the accumulation of unglycosylated glycoproteins in the ER lumen. It is possible that transcriptional inhibition by Act D treatment reduces translation of mRNA at the rough ER. Therefore, we tested the effect of treatment with Tg, a Ca 2ϩ -ATPase inhibitor that induces the UPR by depleting the ER intraluminal calcium pool and does not require ongoing protein synthesis for UPR activation. Act D treatment also reduced the accumulation of spliced XBP1 mRNA in response to Tg (Fig. 2C, lanes 6 -8).
The RT-PCR products representing spliced mXBP1⌬C(s)-d2EGFP appeared at an early time point (1 h) and then subsequently declined (4 h). However, IRE1␣ autophosphorylation from Tg-treated cells was not affected by Act D treatment (Fig. 2H). We then analyzed the phosphorylation status of eIF2␣, the downstream target of the PKR-related ER eIF2␣ kinase (PERK) that is activated in ER-stressed cells (Fig. 2I, lanes  9-16). In the cells treated with Tg alone, eIF2␣ phosphorylation slightly increased (Fig. 2I, lanes 1-4) and then returned to the basal level. This transient phosphorylation is due to transcriptional activation of GADD34, a protein phosphatase 1 subunit that targets eIF2␣ dephosphorylation (15). In addition, the expression of CHOP (GADD 153), the  1-4; D, lanes 1-6; H, lanes 1-5; and I, lanes 1-8), Tg and Act D after a 1-h pretreatment of Act D (C, lanes 5-8; H, lanes 6 -10; and I, lanes 9 -16), and Tg and ␣-amanitin or DRB after a 1-h pretreatment of ␣-amanitin (D, lanes 7-12) or DRB (E ). For F, CHO/mXBP1⌬C(un)-d2EGFP cells were treated without/with Tm or Tg after 7 h pretreatment with the three different RNA polymerase inhibitors. For A-F total RNA samples were prepared and RT-PCR analysis was performed with a primer set flanking the intron in the mXBP1⌬C(un/s)-d2EGFP mRNA. PCR products represent hybrid (H), unspliced (Un), and spliced (S) species. For A, the percent splicing represents an average of three independent experiments and was calculated as described under "Experimental Procedures." Cell lysates were prepared when indicated and immunoprecipitated with rabbit anti-IRE1␣-specific antisera followed by Western blot analysis with mouse anti-IRE1␣ antibody (panel G and H) or directly prepared for Western blot analysis using antibodies against EGFP, phospho-eIF2␣, eIF2␣, and CHOP proteins (panel I ).
downstream target gene product of the PERK pathway, gradually increased (Fig. 2I, lanes 4-8). However, treatment of cells with Tg and Act D gradually and stably increased eIF2␣ phosphorylation, but prevented induction of both CHOP and XBP1⌬C-d2EGFP proteins, as expected from the transcriptional block (Fig. 2I, lanes 9 -16). These findings suggest that in Tg-treated cells, Act D treatment does not affect the kinase/RNase activity of IRE1␣ and/or the kinase activity of PERK, but rather reduces the splicing of XBP1 mRNA and inhibits expression of downstream genes through the transcriptional blockade.
To determine whether the effect of Act D on XBP1 mRNA splicing is mediated via inhibition of RNA polymerase II-dependent transcription, we tested the effect of two specific RNA polymerase II inhibitors, DRB and ␣-amanitin. DRB inhibits CDK7 TFIIH-associated kinase, casein kinase I (CKI), and casein kinase II (CCKII) and prevents phosphorylation of the carboxyl-terminal domain of the large subunit of RNA polymerase II, thereby inhibiting RNA polymerase II transcription in the early stage of elongation (71)(72)(73). ␣-Amanitin binds specifically to the RNA polymerase II large subunit, and at a higher concentration to the RNA polymerase III large subunit, resulting in proteolytic degradation of the polymerases (74 -76). Treatment with either DRB, and to a lesser extent ␣-amanitin, reduced Tg-dependent splicing of XBP1 mRNA (Fig. 2, D and E). Although cells were pretreated with the RNA polymerase inhibitors for 1 h prior to treatment with Tg, it is possible that transcription was FIGURE 3. IRE1␣-dependent splicing of mXBP1⌬C(un)-d2EGFP mRNA does not require translation. A, CHO/mXBP1⌬C(un)-d2EGFP cells were treated with Tg and cycloheximide (CHX). Total RNA samples were analyzed by RT-PCR using a primer set flanking the intron in the mXBP1⌬C(un/s)-d2EGFP mRNA. PCR products represent hybrid (H), unspliced (Un), and spliced (S) species. B, schematic diagrams of the expected dicistronic reporter mRNAs, before and after splicing of the 26-nt intron in the XBP1 mRNA region, produced from transfected pRL-IXFL and pRL-XFL plasmids. From pRL-IXFL transcripts, Renilla luciferase translation is directed by cap-dependent ribosome scanning. The translation of mXBP1⌬N/ ⌬C(un) (before UPR activation) and mXBP1⌬N/⌬C(s)-firefly luciferase (after UPR activation) are directed by IRES-dependent internal initiation. Although Renilla luciferase is continuously translated from pRL-XFL, mXBP1⌬N/⌬C(un) (before UPR activation) and mXBP1⌬N/⌬C(s)-firefly luciferase (after UPR activation) proteins are not produced due to the absence of the IRES. C and D, DHFR-deficient CHO cells were transfected with dicistronic plasmids. At 31 h after transfection, the cells were incubated with fresh medium with or without Tm (10 g/ml) for 7 h and then either total RNA was isolated for RT-PCR analysis, or cell lysates were harvested and stored at Ϫ80°C for dual luciferase assay. D, Renilla luciferase and firefly luciferase activities were measured as described under "Experimental Procedures," and the relative ratio of firefly luciferase to Renilla luciferase activity in each cell lysates was calculated. E, wild-type and Ire1␣/MEFs were nucleofected with the indicated plasmid DNAs (pRL-IXFL reporter plasmid and pED-hIRE1␣-Flag/pED⌬C). At 24 h post-transfection, cells were treated with Tm (10 g/ml) for 8 h and then the cell lysates were prepared for luciferase activity assay. Columns and bars represent the mean Ϯ S.D. of three independent transfection experiments.
not completely blocked at this time. Therefore, the UPR was induced after a 7-h pretreatment with transcription inhibitors (Fig. 2F). These treatments completely blocked XBP1 mRNA splicing compared with treatments of Tm or Tg alone, although unspliced XBP1 mRNA was present. These results suggest that IRE1␣-dependent XBP1 mRNA splicing requires RNA polymerase II-dependent transcription. In addition, in the case of Tm treatment, continued transcription may also be required to maintain the IRE1␣ kinase activity.
IRE1␣-dependent XBP1 mRNA Splicing Can Occur Without Translation-We next examined whether mRNA translation is required for XBP1 mRNA splicing. Splicing of mXBP1⌬C(un)-d2EGFP reporter transcripts was analyzed by RT-PCR of total RNA isolated from cells treated with Tg in the presence or absence of cycloheximide. The intensity of the RT-PCR product representing spliced mXBP1⌬C(s)-d2EGFP (S) continuously increased until 10 h (Fig. 3A) without detectable production of mXBP1⌬C(s)-d2EGFP protein (data not shown). As cycloheximide inhibits the peptidyl transferase reaction of the 60 S ribosomal subunit (77)(78)(79), translation elongation is not required for XBP1 mRNA splicing upon UPR activation.
We then tested whether XBP1 intron removal requires association with translating ribosomes by analysis of two dicistronic reporter vectors. The first vector (pRL-IXFL) encodes Renilla luciferase followed by a multiple cloning site, the IRES from EMCV, and the unspliced murine XBP1 ⌬N/⌬C (437-576 nt with an "AUG" initiation codon) fused to firefly luciferase (Fig. 3B). Under these conditions, Renilla luciferase protein is translated in a cap-dependent manner and spliced mXBP1 ⌬N/⌬C-fused firefly luciferase is expressed by IRES-dependent translation when only the 26-nt intron from mXBP1 ⌬N/⌬C (437-576 nt) is removed. The second vector (pRL-XFL) is identical except it lacks both the EMCV IRES and the AUG initiation codon for the mXBP1 ⌬N/⌬Cfirefly luciferase (Fig. 3B). From pRL-XFL transcripts, Renilla luciferase is continuously translated although ribosomes will not translate the mXBP1 ⌬N/⌬C (437-576 nt) reading frame. After DNA transfection, cap-dependent translation and EMCV IRES-dependent translation were monitored by Renilla luciferase and spliced mXBP1 ⌬N/⌬C-fused firefly luciferase activities in the cell lysates (Fig. 3C). Firefly luciferase activity from the pRL-IXFL vector increased ϳ3-fold upon treatment with Tm. In contrast, pRL-XFL-transfected cells did not show any increase in firefly luciferase activity upon Tm treatment (Fig. 3C). Direct RT-PCR analysis of reporter mRNAs showed that XBP1 mRNA splicing was induced by ER stress irrespective of translation initiation at the murine XBP1 ⌬N/⌬C (437-576 nt) cistron (Fig. 3D, lanes 2 and 4). To verify that the intron from the pRL-IXFL vector was indeed spliced by IRE1␣ RNase, we transfected this vector into Ire1 Ϫ/Ϫ cells. Where the ratio of firefly/Renilla luciferase increased ϳ3-fold in wild-type MEFs, there was little detectable firefly luciferase in Ire1␣ Ϫ/Ϫ MEFs (Fig. 3E). Co-transfection of Ire1␣ Ϫ/Ϫ MEFs with an IRE1␣ expression vector produced high levels of firefly luciferase, irrespective of Tm treatment. This is consistent with constitutive activation of IRE1␣ that occurs upon overexpression (30). These results confirm that firefly luciferase expression from pRL-IXFL depends on IRE1␣. From these observations we conclude that IRE1␣-dependent XBP1 mRNA splicing can occur with- Cytoplasmic Expressed/Delivered mXBP1⌬C(un)-EGFP Transcripts Are Efficiently Spliced Without Nuclear Processing-Although HAC1 splicing can occur in the cytoplasm in yeast (26), there are no studies that localize the splicing of XBP1 mRNA in mammalian cells. As transcription is required for XBP1 mRNA splicing, we asked whether XBP1 mRNA splicing can occur without nuclear processing. Functional mXBP1⌬C(un)-EGFP mRNA splicing templates were directly introduced into the cytoplasm by either cytoplasmic bacteriophage T7 RNA polymerase-dependent transcription or by RNA transfection. In these manners, the RNA substrates should not be exposed to nuclear transcription, modeling, processing, and transport. Before transfecting the cells with DNA or RNA reporter constructs, in vitro transcribed RNA products were analyzed by translation in rabbit reticulocyte lysate. As expected, the capped-or IRES-mXBP1⌬C(un)-EGFP mRNAs produced only an ϳ26-kDa product from unspliced mXBP1⌬C(un) mRNA (Fig. 4A, lanes 1 and 3), consistent with translation termination at a stop codon near the junction between the mXBP1⌬C and EGFP sequences. Translation of the spliced mXBP1⌬C(s)-EGFP mRNA produced a ϳ50-kDa product, the size expected from the spliced mRNA (Fig. 4A, lanes 2  and 4). mXBP1 mRNA splicing was analyzed in vivo using a cytoplasmic expression system based on bacteriophage T7 RNA polymerase and EMCV IRES-dependent translation. T7 RNA polymerase produces stable mRNA transcripts from transfected DNA plasmids that contain the T7 promoter and the EMCV IRES sequence. Previous studies demon-strated that transcripts expressed from cytoplasmic-localized T7 RNA polymerase in this cell line are localized to the cytoplasm (80,81). The EMCV IRES sequence stabilizes the uncapped mRNAs in the cytoplasm and promotes their translation. NIH3T3 cells that express T7 RNA polymerase (3126-2) were transfected with plasmid DNAs. After transfection, the cells were fed with fresh medium containing or lacking Tm and harvested after 24 h for Western blot (Fig. 4B) or fluorescence microscope analysis (data not shown). Cells transfected with a plasmid containing unspliced mXBP1⌬C(un)-EGFP expressed a low level of basal spliced mXBP1⌬C(s)-EGFP protein without induction of ER stress (Fig. 4B, lane 1). However, Tm treatment increased the expression level of spliced mXBP1⌬C(s)-EGFP protein (Fig. 4B, lane 2). In contrast, Tm treatment did not alter expression from the spliced mXBP1⌬C(s)-EGFP mRNA expressed from T7 polymerase (Fig. 4B, lanes 3 and 4). This indicates that 5Ј EMCV IRES-containing unspliced cytoplasmic mXBP1⌬C(un)-EGFP RNA are susceptible to UPR-dependent XBP1 intron removal.
The potential to process the XBP1 intron in the cytoplasm was also evaluated by direct RNA transfection. DHFR-deficient CHO cells were transfected with capped and poly(A)-tailed mXBP1⌬C(un)-EGFP in vitro transcripts and fed fresh medium containing or lacking Tm. After 7 h, the transfected cells were prepared for Western blot or fluorescence microscope analysis. The percentage of fluorescent cells transfected with capped and poly(A)-tailed unspliced mXBP1⌬C(un)-EGFP mRNA was very low in the absence of Tm and significantly increased in the presence of Tm (supplemental Fig. 1). Cells transfected with capped and poly(A)-tailed spliced mXBP1⌬C(s)-EGFP mRNA showed strong fluorescence in the absence or presence of Tm (supplemental Fig. 1). The results correlated well with Western blot analysis using an antibody against the EGFP protein (Fig. 4C). Cells transfected with unspliced mRNA showed a basal level of spliced mXBP1⌬C(s)-EGFP protein (Fig.  4C, lane 1). After Tm treatment, the amount of spliced mXBP1⌬C(s)-EGFP protein increased (Fig. 4C, lane 2). Quantitation demonstrated that ϳ50% of the transfected translatable unspliced mRNA substrates were effectively spliced upon activation of the UPR. The results show that 5Ј capped and 3Ј poly(A)-tailed XBP1 mRNAs directly delivered into the cytoplasm can be efficient substrates for UPR-dependent XBP1 intron removal.
Cytoplasmic mXBP1⌬C(un)-d2EGFP Transcripts Are Efficiently Spliced-To localize the intracellular site of XBP1 mRNA splicing, the CHO/mXBP1⌬C(un)-d2EGFP reporter cells were treated with Tm for 4 h and then nuclear and cytoplasmic fractions were isolated. Cell fractionations were performed by both the protocol of Topisirovic et al. (61) and a commercially available kit (PARIS, Ambion Inc.). However, both methods yielded similar results ( Fig. 5A and data not shown). Similar results were also obtained upon analysis of endogenous XBP1 mRNA splicing in NIH3T3 cells fractionated by a commercially available kit (PARIS, Ambion Inc.) (data not shown). Prior to UPR induction, there was not a significant difference between the RT-PCR products from the nuclear (N) and cytoplasmic (C) fractions where unspliced mXBP1⌬C(un)-d2EGFP transcripts (Un) were equally detected in both fractions (Fig. 5A, lanes 2 and 3). After Tm incubation for 4 h, the RT-PCR product from the cytoplasmic fraction (C) primarily produced the spliced and hybrid species (lane 5). In contrast, the RT-PCR results from the nuclear fraction (N) detected primarily unspliced mRNA (Un) (lane 4). In addition, the amount of spliced mXBP1 in the cytoplasm (C) was 30 times greater than that detected in the nucleus (N) (for calculation, see "Experimental Procedures") and approximately equal to the amount of spliced mRNA (S) in the unfractionated RNA (T) (Fig. 5A,  lanes 1 and 6). After 4 h Tm treatment, most of the previously existing unspliced transcripts (Un) were converted into spliced forms in the cytoplasmic fraction (Fig. 5A, lanes 3 and 5). The nuclear and cytoplasmic fractions were relatively pure, monitored by analysis of tRNA Lys as a cytoplasmic RNA (Fig. 5C), and U6 small nuclear RNA as a nuclear RNA (Fig. 5D), as described previously (61,62). In addition, RT-PCR analysis of total mRNAs from Act D-treated cells indirectly demonstrated that the unspliced XBP1 mRNA is stable in the cytoplasm for up to 10 h (Fig. 2B). Therefore, it is likely that cytoplasmic unspliced transcripts remain in the cytoplasm during the 4-h Tm treatment. As the amount of spliced product in the cytoplasm after UPR induction was approximately equal to the amount of unspliced product in the cytoplasm prior to UPR induction, the spliced products were likely derived from cytoplasmic mRNAs. These results indicate that the unspliced cytoplasmic transcripts are efficiently spliced by ER stress-activated IRE1␣ in the cytoplasm.
XBP1 mRNA Cleavage and Ligation Reactions Do Not Require Activation of the UPR-The experiments described above indicate that XBP1 mRNA cleavage and ligation reactions can occur in the cytoplasm and require ongoing transcription, but not translation. To further test the validity of this hypothesis, we engineered IRE1␣ for localization to the cytoplasm, membrane, or nucleus. In addition, we determined whether ER stress is essential to mediate XBP1 mRNA splicing. IRE1␣ activation is initiated by dimerization-induced trans-autophosphorylation to elicit RNase activity (30,51,69,70). Normally, the stress-sensing ER luminal domain of IRE1␣ mediates dimerization. However, replacing the luminal domain with a basic leucine zipper dimerization domain reconstitutes functional IRE1 (82). Therefore, we fused the kinase and RNase domains of IRE1␣ to a modified FK506-binding domain (F v 1E). The modified FKBP domain (F v 1E) is a high affinity receptor for a cellpermeable synthetic small organic molecule, AP20187. AP20187 is a bivalent ligand that serves as a dimerizer for modified F v 1E domains (83). In addition, we engineered different forms of this chimera to direct cytosol, membrane, or nuclear localization, where F v -IRE⌬N-HA is directed to the cytosol, myristylated Myr-F v -IRE⌬N-HA is directed to membranes, and HA-NLS-F v -IRE⌬N has a nuclear localization signal (Fig. 6, A, B, and G).
Because these three F v -IRE⌬N proteins lack the ER luminal domain, they do not respond to ER stress and should allow for drug-induced activation of IRE1␣ kinase and RNase activities in the absence of ER stress. The three chimeric forms of IRE1␣ were expressed in transiently transfected NIH3T3 cells and analyzed by Western blot analysis. Each protein was expressed efficiently, although expression of Myr-F v -IRE⌬N-HA was reduced severalfold (Fig. 6B). Analysis on a low percentage polyacrylamide gel identified two forms of Myr-F v -IRE⌬N-HA, where the species with reduced mobility was increased in cells treated with AP20187 (Fig. 6C). The slower migrating species disappeared after phosphatase treatment, indicating it is a hyperphosphorylated form of IRE1␣ (Fig. 6C). All three forms of F v -IRE1␣ displayed a phosphorylated and non-phosphorylated species, although the increase in the phosphorylated form upon AP20187 treatment was variable (Fig. 6D). Where the majority of Myr-F v -IRE⌬N-HA became phosphorylated upon AP20187 treatment (Fig. 6D, lane 5), only minor portions of F v -IRE⌬N-HA and HA-NLS-F v -IRE⌬N were phosphorylated in response to AP20187 (Fig.  6D, lane 5). The significant amount of hyperphosphorylated F v -IRE1␣  detected in the absence of AP20187 may be a consequence of overexpression (Fig. 6D, lane 6).
To directly measure RNase activity for the differently localized F v -IRE⌬N proteins, the splicing efficiency of endogenous XBP1 mRNA was evaluated by RT-PCR analysis (Fig. 6E) and quantitative real-time RT-PCR analysis (Fig. 6F) from total RNA isolated from transfected cells. As we expected, Myr-F v -IRE⌬N-HA produced most spliced XBP1 mRNA upon AP20187 treatment (Fig. 6, E and F, lane 12). From quantitative real-time RT-PCR analysis, the RNase activity of Myr-F v -IRE⌬N-HA increased ϳ2-fold upon AP20187 treatment. In contrast, both F v -IRE⌬N-HA and HA-NLS-F v -IRE⌬N proteins displayed a low level of XBP1 mRNA splicing (ϳ2-fold) that did not increase upon AP20187 treatment (Fig. 6, E and F, lanes 7-10). The low level splicing induced by F v -IRE⌬N-HA and HA-NLS-F v -IRE⌬N proteins is likely not mediated through activation of the endogenous IRE1␣ because similar overexpression of a wild-type FKBP domain that does not bind AP20187 fused to IRE1␣⌬N did not increase XBP1 mRNA splicing in the presence or absence of AP20187 (data not shown). These splicing analyses of endogenous mXBP1 transcripts showed that the efficiency of mXBP1 mRNA splicing correlates with the increased amount of hyperphosphorylated IRE1␣ induced by the chemical dimerizer AP20137.
The effect of a chemical dimerizer on RNase activity was also investigated by analyzing the fluorescence from mXBP1⌬C(s)-d2EGFP in the CHO/mXBP1⌬C(un)-d2EGFP reporter cells transfected with the three different F v -IRE⌬N expression plasmids (Fig. 6G). Immunofluorescence analysis using anti-HA antibody demonstrated that each of the different IRE1␣ chimeras was localized to the appropriate compartment (Fig. 6G). Fluorescence microscopy showed that the highest frequency of GFP-positive cells (74%) was observed in cells expressing Myr-F v -IRE⌬N-HA. In contrast, cells transfected with F v -IRE⌬N-HA or HA-NLS-F v -IRE⌬N displayed a low frequency of GFP-positive cells (33%). From these XBP1 mRNA splicing studies upon chemical-induced dimerization of membrane-localized IRE1␣, we suggest that most unspliced XBP1 mRNA transcripts are spliced in the cytoplasm or near the ER membrane. The results also suggest that XBP1 mRNA cleavage and ligation reactions do not require activation of the UPR, and that the UPR is only required to dimerize IRE1␣ to lead to RNase activation that initiates cytoplasmic XBP1 mRNA splicing.

DISCUSSION
To date, there are no reports that describe the cellular environment and molecular processes required for XBP1 mRNA splicing in mammalian cells. Previous studies in S. cerevisiae indicated that splicing of HAC1 mRNA can occur in the cytoplasm and does not require cellular transcription or translation (26). In addition, it was recently reported that in yeast, pre-tRNA intron removal occurs in the cytoplasm (44 -47) and that mammalian neuronal cells have RNA splicing capability in the cytoplasm (84). Our findings indicate that mammalian IRE1␣-dependent splicing of XBP1 mRNA transcripts displays some of the same properties for HAC1 mRNA splicing characterized in S. cerevisiae. In mammals, the XBP1 mRNA splicing reaction can occur in the cytoplasm, does not require active translation, and is very efficient where all previously expressed cytoplasmic XBP1 mRNAs can be spliced. However, we have extended these findings by showing that in contrast to yeast, the splicing reaction in mammalian cells requires ongoing transcription. At present, we do not know why transcription is required, as RNA polymerase inhibitors are known to interfere with a variety of different cellular processes. We have also shown that unspliced XBP1 transcripts directly introduced into the cytoplasm are effectively spliced by IRE1␣ in response to ER stress. Therefore, nuclear transcription and/or processing are not required for XBP1 mRNA splicing. In addition, a membranetargeted form of IRE1␣ that lacks the ER luminal domain can mediate XBP1 mRNA splicing, thus indicating that cytoplasmic XBP1 mRNA splicing can occur in the absence of UPR activation.
Several lines of evidence support our conclusion that XBP1 mRNA splicing occurs in the cytoplasm. First, cell fractionation experiments demonstrated that all previously existing cytoplasmic XBP1 mRNA is spliced upon activation of the UPR (Fig. 5A). Second, IRE1␣ targeted to cellular membranes, but not to the cytosol or nuclear compartments, was able to effectively splice XBP1 mRNA (Fig. 6, D-G). We do not know if the membrane-targeted IRE1␣ was more effective because it was more efficiently autophosphorylated or whether XBP1 substrates are also localized to membranes. In this respect, it is interesting to note that translation of XBP1 mRNA was recently demonstrated to occur at the ER membrane (85). Finally, IRE1␣ was able to splice substrates produced by T7 transcription in the cytoplasm or introduced by RNA transfection. Therefore, IRE1␣ cleavage substrates do not require nuclear transcription and/or processing. However, these experiments cannot rule out the possibility that some of the substrate RNA enters the nucleus where the splicing reaction may occur. Therefore, although the preponderance of our data indicates that cytoplasmic splicing of XBP1 mRNA occurs, we cannot rule out that the reaction also occurs in the nucleus.
In yeast, HAC1 mRNA is associated with polysomes, however, Hac1p is not translated due to long-range base pairing between the HAC1 intron and the 5Ј end of the HAC1 mRNA (26). In contrast, mammalian XBP1 mRNA does not utilize such a translational control mechanism (32). Because the product from unspliced XBP1 mRNA encodes the DNA binding and dimerization domains, continued translation of unspliced XBP1 mRNA would produce a trans-dominant negative form of XBP1 that was shown to act as a repressor of spliced XBP1(s) protein (86). In addition, Yoshida et al. (65) recently showed that unspliced XBP1(un) protein forms a complex with spliced XBP1(s) protein and excludes it from the nucleus. The complex is rapidly degraded by the proteasome due to a degradation motif contained in unspliced XBP1(un). Thus, unspliced XBP1(un) acts as a negative feedback regulator of spliced XBP1(s) and it was suggested that rapid degradation of unspliced XBP1(un) by the proteasome is required for temporal control of the UPR (65,86). In view of these recent findings, we propose that cytoplasmic XBP1 mRNA splicing also provides an additional important mechanism to rapidly change the function/activity of the translated XBP1 polypeptide to activate UPR transcription.
We have demonstrated UPR-induced expression of an EGFP fusion protein that is mediated by IRE1␣-dependent splicing of an XBP1 intron. The UPR-induced rapid expression of EGFP provides a unique mechanism for regulated protein expression that can be engineered into expression vectors. This type of an approach could have a variety of applications. For example, expression of an XBP1-toxin fusion protein could be delivered to induce death in cells that have an activated UPR, such as virally infected cells or transformed cells. However, our studies do show a significant basal level of expression in non-stressed cells. This background basal expression was present in only a subpopulation of the cells. At present we do not know if this background expression reflects a low level of UPR being sensed in some cells or whether it is a normal physiological response, possibly associated with some aspect of the cell cycle. It is also possible that a translational frameshift may occur in some cells to generate EGFP in the absence of XBP1 splicing. It should also be noted that mXBP1⌬C(s)-d2EGFP expressed in our reporter cell line contains the XBP1 DNA binding domain so it may inhibit the function of endogenous XBP1(s). However, efficient XBP1 intron removal from the mXBP1⌬N/⌬C(un)-firefly luciferase mRNA splicing reporter indicates that the presence of the XBP1 DNA binding domain is not essential for splicing. Because the mXBP1⌬C(s)-d2EGFP fusion protein localizes to the nucleus, endogenous XBP1(s) may not be exposed to cytosolic proteasomes as suggested by Yoshida et al. (65). In addition, the PERK-dependent UPR pathway was not affected by expression of mXBP1⌬C(un) or mXBP1⌬C(s)-d2EGFP protein as eIF2␣ phosphorylation and CHOP expression were observed (Fig. 2I). The cell line that specifically reports IRE1␣ activation by EGFP expression could provide a useful reagent for screening for compounds that activate or inhibit the IRE1␣ subpathway of the UPR.