Rapid Turnover of Unspliced Xbp-1 as a Factor That Modulates the Unfolded Protein Response*

The mammalian and yeast unfolded protein responses (UPR) share the characteristic of rapid elimination of unspliced Xbp-1 (Xbp-1u) and unspliced Hac1p, respectively. These polypeptides derive from mRNAs, whose splicing is induced upon onset of the UPR, so as to allow synthesis of transcription factors essential for execution of the UPR itself. Whereas in yeast translation of unspliced Hac1p is blocked, mammalian Xbp-1u is synthesized constitutively and eliminated by rapid proteasomal degradation. Here we show that the rate of Xbp-1u degradation approaches its rate of synthesis. The C terminus of XBP-1u ensures its trafficking to the cytoplasm, and is sufficient to impose rapid degradation. Degradation of XBP-1u involves both ubiquitin-dependent and ubiquitin-independent mechanisms, which might explain its unusually rapid turnover. Xbp-1-/- mouse embryonic fibroblasts reconstituted with mutants of XBP-1u that show improved stability differentially activate UPR target genes. Unexpectedly, we found that one of the mutants activates transcription of both Xbp-1-specific and non-Xbp-1-dependent UPR targets in response to tunicamycin treatment, even more potently than does wild type Xbp-1. We suggest that the degradation of Xbp-1u is required to prevent uncontrolled activation of the UPR while allowing short dwell times for initiation of this response.

The mammalian and yeast unfolded protein responses (UPR) share the characteristic of rapid elimination of unspliced Xbp-1 (Xbp-1u) and unspliced Hac1p, respectively. These polypeptides derive from mRNAs, whose splicing is induced upon onset of the UPR, so as to allow synthesis of transcription factors essential for execution of the UPR itself. Whereas in yeast translation of unspliced Hac1p is blocked, mammalian Xbp-1u is synthesized constitutively and eliminated by rapid proteasomal degradation. Here we show that the rate of Xbp-1u degradation approaches its rate of synthesis. The C terminus of XBP-1u ensures its trafficking to the cytoplasm, and is sufficient to impose rapid degradation. Degradation of XBP-1u involves both ubiquitin-dependent and ubiquitinindependent mechanisms, which might explain its unusually rapid turnover. Xbp-1 ؊/؊ mouse embryonic fibroblasts reconstituted with mutants of XBP-1u that show improved stability differentially activate UPR target genes. Unexpectedly, we found that one of the mutants activates transcription of both Xbp-1-specific and non-Xbp-1-dependent UPR targets in response to tunicamycin treatment, even more potently than does wild type Xbp-1. We suggest that the degradation of Xbp-1u is required to prevent uncontrolled activation of the UPR while allowing short dwell times for initiation of this response.
Protein folding in the endoplasmic reticulum (ER) 4 is carried out under the constant scrutiny of the ER quality control machinery (1). The overall capacity of the ER to fold newly synthesized proteins must match the load of client proteins that emerge into the ER. When this amount exceeds the folding capacity of the ER, a signaling pathway emanates from the ER that controls gene transcription, as well as protein translation. This ER to nucleus signaling cascade is referred to as the unfolded protein response (UPR). The overall goal of the UPR is to enhance the clearance of misfolded proteins from the ER, and consequently the UPR alleviates ER stress (2).
In yeast, Ire1p is the only known transducer of the UPR. In response to ER stress conditions, Ire1p dimerizes and undergoes autophosphorylation. This event induces a conformational change that activates a nuclease domain located in its cytosolic tail (3,4). By means of this nuclease activity, activated Ire1p splices the mRNA of Hac1, which in its spliced form encodes Hac1p, a potent transcription factor that induces transcription of many genes that encode ER chaperones, proteins that participate in ER to Golgi trafficking and components of the ER degradation machinery (2).
The mammalian UPR is minimally composed of three transducers: Perk, Atf6, and Ire1 (5). Ire1 is highly conserved from yeast to mammals, but the homolog of Hac1 eluded scientists for many years. The mammalian counterpart of Hac1 was identified as Xbp-1. Xbp-1, a member of the CREB/ATF family of transcription factors, does not share any significant sequence homology with Hac1. It is composed of a basic leucine zipper-containing DNA binding domain located at the N terminus. The C terminus of Xbp-1 operates as a transcription activation domain. In contrast to yeast, in which Ire1p removes a relatively large intron from Hac1 mRNA, only 26 bases are excised by splicing the mRNA of Xbp-1. This splicing induces a frameshift in the coding sequence, replacing the 105-amino acid C terminus of the unspliced Xbp-1 protein with a 226-amino acid domain (6,7). The spliced C terminus potently activates transcription of downstream target genes. Similar to yeast, the specific target genes of Xbp-1 encode proteins that enhance the folding capacity of the ER and participate in the clearance of misfolded proteins from the ER (8).
Despite high levels of the unspliced Hac1 mRNA, no unspliced Hac1 protein is detectable, caused by selective inhibition of its synthesis. The mechanism of the arrest in translation of Hac1 mRNA involves base pairing interactions between the intron and the 5Ј untranslated region (9). In contrast to Hac1, unspliced Xbp-1 (Xbp-1u) is continuously synthesized, but it is unstable and quickly degraded (10). Once Xbp-1 mRNA is spliced, the encoded protein gains stability and allows activation of transcription.
Lee et al. (10) previously demonstrated that an N-terminal segment of Xbp-1, the domain shared between the unspliced and the spliced forms, is relatively stable and exerts dominant-negative activity for expression of genes activated by spliced Xbp-1 (Xbp-1s) targets. This is probably because of direct competition between the N-terminal segment of Xbp-1 and Xbp-1s for the available DNA binding sites. Furthermore, overexpression of a mutant of Xbp-1u, which cannot be spliced because of point mutations in the intron (referred to here as "unspliceable"), also displays dominant-negative characteristics, when assayed under ER stress conditions that robustly generate Xbp-1s. Finally, replacement of lysine residues with arginines at the C terminus of Xbp-1u increases its stability and potentiates its dominant negative activity. These observations implicate the ubiquitin-proteasome system as responsible for Xbp-1u degradation (10). Overall, these data suggest that Xbp-1u is a potential inhibitor of the Xbp-1 pathway of the UPR, and its removal by proteolysis is a prerequisite for proper activation of the UPR.
Here, we examined the localization and stability of Xbp-1u and its C terminus. We show that Xbp-1u is an intrinsically unstable protein, which is degraded in living cells at a rate comparable with its rate of synthesis. The C terminus of Xbp-1u is sufficient to mediate this exceedingly rapid degradation, which occurs predominantly in the cytoplasm involving ubiquitin-dependent and -independent pathways. Against expectation, improvement of the stability of Xbp-1u augmented the transcription of Xbp-1-specific and -nonspecific target genes under conditions that robustly induce the UPR. We conclude that the rapid degradation of Xbp-1u is required to prevent uncontrolled activation of the UPR.

EXPERIMENTAL PROCEDURES
In Vitro Transcription and Translation-pcDNA3.1-encoded Xbp-1 N terminus , Xbp-1u, and Xbp-1s were translated using the TNT coupled reticulocyte lysate system (Promega) according to manufacturer's instructions, in the presence of [ 35 S]methionine (PerkinElmer Life Sciences). To initiate the chase period we added RNase (1 mg/ml final concentration, Roche) together with excess unlabeled methionine. Equal aliquots were taken at the indicated time points, diluted in reducing sample buffer, boiled, and analyzed by SDS-PAGE (12%) followed by fluorography.
Epifluorescence Imaging-HeLa cells were seeded on glass coverslips 18 h before transfection. Vectors encoding GFP fused at its C terminus to the specified constructs were transfected using the calcium phosphate precipitation method (CalPhos, BD Bioscences). 24 h after transfection, nuclei were labeled with Hoechst 33342 (blue fluorescence, Invitrogen). Images were obtained with a Spot RT digital camera mounted on a TE300 Nikon microscope at ϫ40 magnification.
Retrovirus Production-Wild type Xbp-1 and Xbp-1 mutants were cloned into the pMiG MSCV vector harboring an internal ribosomal entry site-GFP element to allow sorting of infected cells. Viral particles were made in 293T cells by triple transfection of the retroviral vector (2 g), pMD-gag-pol (2 g), and pVSV-G (2 g) using Effectene (Qiagen). Cells were infected as previously described (12).

Xbp-1u Is an Intrinsically Unstable
Protein-Previous studies demonstrated rapid turnover of Xbp-1u in different cell types (7,10). To test whether Xbp-1u is an intrinsically unstable protein, we examined its stability in a cell-free assay. Using the T7 TNT in vitro translation system, we generated Xbp-1u, Xbp-1s, and the region of Xbp-1 shared between Xbp-1u and Xbp-1s, amino acids 1-161 (referred to as XBP-1 N terminus ). This approach yields radiochemically pure product under native conditions. The chase period was initiated by inclusion of RNase and excess unlabeled methionine into the reaction mixture. As seen in intact cells, the N terminus of Xbp-1 was stable throughout the chase period (Fig. 1A, upper panel). In contrast, Xbp-1u decayed rapidly. Over 80% of the initial Xbp-1u was degraded within the first hour of chase, the remainder was stable till the end of the experiment (Fig. 1A, middle panel). The percentage of stable material varied from experiment to experiment, but never exceeded 20% of the initial amount of Xbp-1u. We speculate that this material represents Xbp-1u that, by means unknown, is sequestered from the degradation machinery. Xbp-1s decayed more gradually than Xbp-1u ( Fig. 1A, lower panel). Inclusion of proteasome inhibitors, such as ZL 3 VS or MG132 had little if any effect on the degradation observed in vitro (data not shown).
In addition, we looked for the appearance of polypeptides that might indicate their modification by ubiquitin. Longer exposures of the autoradiogram showed the presence of additional polypeptides for Xbp-1u, arranged in a ladder reminiscent of modification by ubiquitin. Note that the reticulocyte lysate used for the generation of in vitro translation products is competent for ubiquitination reactions. These additional polypeptides decayed over time (Fig. 1B, middle panel, labeled by asterisk). Of note, they were observed in the absence of N-ethylmaleimide or proteasome inhibitors, commonly added to enrich for ubiquitinated proteins (15). We neither observed this pattern for the N terminus of Xbp-1 nor for Xbp-1s (Fig. 1B, left and right panels).
To confirm that the multiple bands, which appear at higher molecular weight than Xbp-1u, indeed correspond to products modified by ubiquitin, we translated Xbp-1u in the presence of recombinant HAtagged ubiquitin. The HA tag increases the molecular mass of Ub by 3 kDa, and thus alters the mobility of the ubiquitinated polypeptides. We indeed observed a shift in the mobility of the high molecular weight polypeptides, indicating modification of Xbp-1u by ubiquitin (Fig. 1C). The instability of Xbp-1u in the cell-free extract and its modification by ubiquitin indicate that Xbp-1u degradation is a robust process, and may well be constitutive.
The N and C Terminus Domains of Xbp-1 Determine Its Overall Intracellular Localization and Its Turnover-Xbp-1u can be divided into two parts ( Fig. 2A). Amino acids 1-161, which are shared between XBP-1u and Xbp-1s, were defined as the Xbp-1 N terminus . The Xbp-1 C terminus(u) starts at the end of the intron (amino acid 168) and extends to the stop codon of Xbp-1u (position 267). Similarly, the segment referred to as Xbp-1 C terminus(s) is the domain unique to Xbp-1s, created by removal of the short intron. To examine the contribution of the different domains to intracellular localization and the turnover of Xbp-1, we generated GFP C-terminal fusions for each. For these experiments, Xbp-1u was used in a form that prevents its splicing (referred to as unspliceable), as described (12). This manipulation was necessary to prevent the splicing of GFP-Xbp-1u mRNA by endogenous Ire1. We also generated an unspliceable mutant in which the two lysines, located at positions 235 and 252 of the Xbp-1u, were replaced by arginines. This mutation, referred to as K2R reportedly enhances the stability of Xbp-1u, probably by inhibiting its ubiquitination (10).
HeLa cells were transfected with the individual GFP constructs and imaged by fluorescence microscopy. The nucleus was visualized by staining with Hoechst 33342 (blue fluorescence). In parallel, turnover of the GFP fusion proteins was measured by pulse-chase analysis. Xbp-1u showed no preferential localization to either the nucleus or the cytoplasm and was evenly distributed in the cell (Fig. 2B). The Xbp-1 N terminus was localized exclusively to the nucleus, most likely because of the basic leucine zipper domain and the three bipartite nuclear targeting sequences (residues 53-69, 68 -84, and 69 -85) (Fig. 2B). Conversely, Xbp-1 C terminus(u) was partially excluded from the nucleus (Fig. 2B, third panel). Our analysis for nuclear export sequences did not reveal any consensus nuclear export signals located in the Xbp-1 C terminus(u) .
However, it should be borne in mind that nuclear export signals are less conserved, and for their function usually require post-translational modifications, such as phosphorylation, which are difficult to predict. We conclude that in the context of the full-length molecule, the Xbp-1 C terminus(u) neutralizes the nuclear localization signals in Xbp-1 N terminus to ensure dynamic trafficking of Xbp-1u in and out of the nucleus.
Examination of the turnover of the GFP fusion proteins by pulsechase analysis showed rapid turnover for GFP-Xbp-1u and GFP-C-Xbp-1 C terminus(u) . Cells were pulse-labeled with [ 35 S]methionine for 30 min. Within 30 min of chase, both GFP-Xbp-1u and GFP-C-Xbp-1 C terminus(u) were degraded completely (Fig. 2B). GFP-Xbp-1 N terminus showed much better stability and was detectable at the 60-min chase point.
In contrast to what was observed for the cell-free products, in live cells GFP-Xbp-1 N terminus was not completely stable and did decay  Xbp-1 C terminus(u) imposes cytoplasmic localization and rapid turnover. A, schematic depiction of the constructs fused to the N terminus of GFP. Light blue, Xbp-1 N terminus ; red, splicing region; white, Xbp-1 C terminus(u) ; gray, Xbp-1 C terminus(s) ; black, unspliceable splicing region. B and C, HeLa cells were transfected with the indicated construct. 24 h after transfection nuclei were stained with Hoechst and fluorescent images were taken using a ϫ40 objective. In parallel, cells were pulse-labeled with [ 35 S]methionine for 30 min. Cells were lysed in 1% SDS, and lysate was then diluted to 0.07% SDS with Nonidet P-40 lysis mixture followed by immunoprecipitation with anti-GFP antibodies and analyzed by SDS-PAGE (12%). D, HeLa cells stably expressing the indicated constructs were pulse-labeled as before in the presence or absence of the proteasome inhibitor ZL 3 VS. Cell lysates and immunoprecipitation were carried out as in B. over time. Because Xbp-1 C terminus(u) is sufficient to direct rapid degradation of an otherwise stable protein like GFP, and because it confers localization to the cytoplasm, we conclude that the degradation of Xbp-1u occurs mostly, but not exclusively in the cytoplasm.
To further investigate the intracellular location of Xbp-1u degradation, we fused the K2R mutant of unspliceable Xbp-1u to GFP. HeLa cells were transfected with GFP-Xbp-1u or its K2R mutant. Under conditions of comparable transfection efficiency, the K2R mutant in its GFP fused form was more stable than the wild type allele, as assessed by pulse-chase analysis (the gel at Fig. 2C represents loading of equal amounts of incorporated [ 35 S]methionine). Imaging of the GFP-Xbp-1-K2R mutant revealed accumulation of GFP signal in the cytoplasm. Therefore, inhibition of Xbp-1u degradation results in accumulation in the cytoplasm, and further shows that this is the major site of degradation.
We applied a similar analysis to Xbp-1s. Xbp-1s was localized predominantly to the nucleus (Fig. 2B). Analysis of Xbp-1 C terminus(s) did not show the preferential location to the cytoplasm, as was seen for Xbp-1 C terminus(u) . Instead, Xbp-1 C terminus(s) was evenly distributed between the nucleus and the cytoplasm. Thus, the intracellular localization of full-length Xbp-1s is governed by its N-terminal segment.
Pulse-chase analyses for GFP-Xbp-1s and GFP-Xbp-1 C terminus(s) indicated that they also decay quite rapidly, although at a slower rate than their unspliced counterparts. We also noticed that Xbp-1s appeared more diffuse at the later chase time points (Fig. 2B, fourth  panel), suggesting post-translational modification(s) in live cells, but not in vitro (compare with Fig. 1A). Because Xbp-1s is localized predominantly to the nucleus, its modification(s) and degradation probably occur there.
To assess the rate at which Xbp-1u is degraded when expressed at more physiological levels, we established stable HeLa cell lines derived from single cell clones expressing either GFP-Xbp-1u in its unspliceable form, or GFP-Xbp-1 C terminus(u) . The expression of these constructs was driven by a cytomegalovirus promoter. Remarkably, both fusion proteins were barely detectable by pulse-chase analysis, unless a proteasome inhibitor was included during the pulse and the chase. Regardless of the pulse labeling time used, GFP-Xbp-1u or GFP-Xbp-1 C terminus(u) disappeared at a chase time equal to the duration of the pulse. We also noticed that inclusion of the proteasome inhibitor did not completely spare these proteins from degradation, suggesting that residual activity of the proteasome, not blocked by ZL 3 VS, is sufficient to bring about degradation of Xbp-1u (Fig. 2D). Alternatively, nonproteasomal proteolysis may contribute as well. These results demonstrate that the rate of Xbp-1u degradation, mediated by its C terminus, approaches its rate of synthesis.
Xbp-1 C terminus(u) Does Not Require Ubiquitination for Its Degradation-To investigate the mechanism by which GFP-Xbp-1 C terminus(u) is destroyed, we generated an HA-tagged version of Xbp-1 C terminus(u) , a protein that only contains the two native lysine residues at the C terminus(u). We also generated the corresponding HA-Xbp-1 C terminus(u) K2R mutant, in which these lysines were replaced by arginines. These constructs were transfected into 293T cells. Pulse-chase analysis was then performed in the presence and absence of the proteasome inhibitor ZL 3 VS, followed by anti-HA immunoprecipitation. Both polypeptides were barely detected by pulse-chase analysis in the absence of proteasome inhibitor, again indicating a degradation rate close to that of their rate of synthesis (Fig. 3A). For both constructs the inclusion of ZL 3 VS markedly improved stability of the protein. Therefore, the presence of primary amines in the sequence of of the C terminus(u) is not required for its proteasome-dependent degradation.
Because the HA epitope tag starts with alanine, it is often acetylated when positioned at the N terminus (16). Therefore, the degradation of HA-tagged Xbp-1 C terminus(u) K2R might represent a proteasome-dependent but ubiquitin-independent degradation pathway. To further address this possibility, we constructed a fusion protein between ubiquitin and the HA-tagged Xbp-1 C terminus(u) separated by a proline residue (Ub-P-Xbp-1 C terminus(u) ). Because ubiquitin C-hydrolases quickly cleave the peptide bond between the ubiquitin Gly 76 residue and a C-terminal extension, this method generates polypeptides with the predetermined N-terminal residue of choice (17). The Ub-P-Xbp-1 C terminus(u) K2R construct is designed to yield a C terminus(u) fragment that carries no lysines and a proline residue at its N terminus, and therefore cannot be ubiquitinated.
The Ub-P-Xbp-1 C terminus(u) and its K2R mutant were transfected into 293T cells. Pulse-chase analysis was conducted 24 h after transfection in the presence or absence of the proteasome inhibitor MG132. We used MG132 instead of ZL 3 VS, because MG132 is a reversible inhibitor that can be removed by washing. As expected, in the absence of proteasome inhibitor very little of the polypeptide was recovered by anti-HA immunoprecipitation (Fig. 3B). This degradation is probably mediated by Xbp-1 C terminus(u) . When chased in the presence of MG132, Ub-P-Xbp-1 C terminus(u) was stabilized and gradually converted into P-Xbp-1 C terminus(u) (Fig. 3B), showing that this construct is processed as expected in living cells. We repeated this experiment followed by a chase period of 90 min in the presence of MG132 to allow maximal conversion into P-Xbp-1 C terminus(u) . Cells were then washed to remove MG132 (Fig. 3C, diagram). We followed the fate of P-Xbp-1 C terminus(u) or its K2R mutant generated from the ubiquitin fusion protein precursor in the presence of MG132, and saw similar degradation profiles for both the P-Xbp-1 C terminus(u) and its K2R mutant. Removal of MG132 did not affect the rate of degradation either. Likewise, we did not observe accelerated degradation of the remaining intact Ub-P-Xbp-1 C terminus(u) (Fig. 3C). Thus, we conclude that the removal of MG132 did not restore the proteasomal function. Nonetheless, the fact that the P-Xbp-1 C terminus(u) K2R mutant, although not containing any acceptor sites for ubiquitination, was unstable and decayed at comparable rates to P-Xbp-1 C terminus(u) supports the notion that ubiquitination in this degradation reaction is dispensable (Fig. 3D).
The ts20 cells express a temperature-sensitive E1 activity. Within 2 h at the restrictive temperature (42°C), de novo ubiquitination is blocked and the cells undergo a G 2 cell cycle arrest (11). We used these cells to explore a requirement for ubiquitination in the degradation of Xbp-1u. Unspliceable Xbp-1u was expressed by transfection and the cells were incubated for 12 h at 42°C or kept at 32°C, the permissive temperature. The 12-h exposure to 42°C eliminates completely the high M r ubiquitinated species observable on SDS-PAGE. As an internal control we followed the degradation of Ib, which is degraded by the proteasome in a strictly ubiquitin-dependent manner (18).
As expected, Xbp-1u was rapidly degraded at 32°C. At 42°C we observed a reduction in the rate of degradation, but degradation still continued. In contrast, Ib degradation was completely blocked at the restrictive temperature. Note the different time scale used in the analysis of Xbp-1u and Ib turnover (Fig. 3E). We conclude that an ubiquitinindependent degradation mechanism significantly contributes to the overall elimination of Xbp-1u.
Modulation of Xbp-1u Stability Affects the UPR-To examine whether the unusually rapid degradation of Xbp-1u affects the mammalian UPR, we designed two Xbp-1 mutants that selectively affect Xbp-1u, but not Xbp-1s. The first mutant was engineered to contain a stop codon immediately after the intron. This mutant, designated as "stop," expresses the stable N-terminal fragment (amino acids 1-175) instead of the full-length Xbp-1u. Upon splicing, the stop mutant should yield the exact open reading frame of Xbp-1s (Fig. 4A).
Xbp-1 C terminus(u) contains two tandem proline (PP) elements. We reasoned that the PP elements might impose conformational constraints on the secondary structure of the polypeptide, and might therefore be required for as yet to be determined protein-protein interactions. These interactions might play a role in degradation. We therefore replaced the two PP sequences with leucine-histidine (LH), and refer to this Xbp-1 mutant as PP2LH. Again, this mutation was introduced in a manner that does not alter the open reading frame of Xbp-1s (Fig. 4A).
Wild type Xbp-1 and the two mutants were cloned into the pMig retroviral vector, harboring an internal ribosomal entry site-GFP element. Xbp-1 Ϫ/Ϫ MEFs were transduced with these vectors. An empty pMig retrovirus was used as a negative control. After transduction, GFP positive cells were sorted and propagated in culture. Levels of GFP were comparable for the pMig control and the different Xbp-1 constructs (not shown). We then assayed the expression and stabilities of Xbp-1 by pulse-chase analysis. Wild type Xbp-1u was not detectable unless proteasome inhibitor was included (Fig. 4B). These results again demonstrate that Xbp-1u is degraded at a rate comparable with its rate of synthesis, preventing it from reaching detectable levels. As expected, the stop mutant was readily detected and showed improved stability (Fig.  4B). The PP2LH mutant was detected at time 0, but was rapidly degraded. Incorporation of ZL 3 VS delayed its degradation, but did not fully block it (data not shown). We conclude that whereas the mutation PP2LH indeed improves the stability of Xbp-1u, it does not rescue it from rapid degradation.
To verify that the different Xbp-1 constructs retain the ability of being spliced, we treated the cells for 8 h with tunicamycin. Under these ER stress conditions, Xbp-1s was generated (Fig. 4C). Following 30 min of chase, Xbp-1s appeared as a disperse band at lower intensity, consistent with post-translational modifications and subsequent degradation. Under these conditions, the synthesis of the stop mutant was still detected, suggesting that the synthesis of WT Xbp-1u and PP2LH also continues, but their expression does not reach readily detectable levels because of rapid degradation.
Finally we assayed whether expression of Xbp-1u analogs with enhanced stability would affect the induction of the UPR. To this end, we measured by real-time PCR the level of mRNA of three target genes of the UPR in response to tunicamycin treatment. We measured the mRNA levels of Erdj4, a specific target of the Xbp-1 pathway (8); Bip, a target shared by all UPR pathways (19,20); and Chop, a target downstream to the Perk/Atf4 pathway (19). As expected, Erdj4 transcription was hardly induced in the Xbp-1 Ϫ/Ϫ GFP cells. Reconstitution with the WT Xbp-1 gene fully restored induction. The reconstitution of the Xbp-1 Ϫ/Ϫ cells with the stop mutant attenuated the induction of Erdj4 in comparison to reconstitution with WT Xbp-1. This is probably because of the dominant negative activity of truncated Xbp-1. We consistently observed that cells expressing the PP2LH mutant were superior to WT Xbp-1 expressing control cells with respect to induction of the Xbp-1 specific target gene, Erdj4 (Fig. 5, left panel). Unexpectedly, similar results were obtained when we examined the induction of Bip and Chop (Fig. 5, middle and right panels). We conclude that Xbp-1u, if stabilized, enhances the overall induction of the UPR and may modulate expression not only of Xbp-1-specific but also of other genes not usually considered Xbp-1 targets.

DISCUSSION
Many proteins are short-lived, a property exploited to obtain careful control over protein levels for regulatory purposes. For instance, Ib when phosphorylated by the IKK signalosome, is degraded with a halflife of 5-15 min (21). Class I major histocompatibility complex products are diverted for degradation by the human cytomegalovirus-encoded proteins US2 and US11, also with a half-life of minutes (22,23). Both of these degradation reactions are readily demonstrable by pulse-chase analysis. Here we characterized a much faster degradation reaction. Even when expressed in stable fashion under the control of strong promoters, such as the cytomegalovirus or long terminal repeat, Xbp-1u is not detectable by pulse labeling or immunoblotting. We could also recapitulate the robust degradation of Xbp-1u in a cell-free environment, and detect ubiquitinated intermediates in reticulocyte lysate. We further demonstrate that this unprecedented rate of destruction is not achieved by coincident. Rather, the C terminus of the unspliced product utilizes nuclear export and may override the need for ubiquitination. The net result is that no physiologically relevant levels of Xbp-1u are produced.
Other examples of ubiquitin-independent, 20 S proteasome-dependent degradation are on record. NAD(P)H quinone oxidoreductase 1 was recently demonstrated to bind and regulate the degradation of p53 and ornithine decarboxylase (24,25). Addition of dicoumarol inhibited NAD(P)H quinone oxidoreductase 1 activity and promoted the degradation of both substrates by a ubiquitin-independent mechanism. At leastforp53,thismechanismoccursindependentlyofitsmdm2ubiquitindependent pathway. We think that similarly to p53, mechanisms involving both ubiquitin-dependent and -independent pathways synergize to promote the degradation of Xbp-1u.
Yeast cells have evolved a different mechanism to avoid expression of unspliced Hac1p, the counterpart of Xbp-1. This is caused by a block in translation to prevent the expression of unspliced Hac1p altogether. Why should the expression of the unspliced Xbp-1 and Hac1 proteins be avoided? We speculated that Xbp-1u, unless rapidly removed, may perturb UPR signaling. To address this hypothesis we generated two Xbp-1 mutants, in which only the unspliced open reading frame was affected in its expression. One mutant, termed stop, eliminates completely the C terminus(u) domain. This mutant generates a stable Xbp-1 fragment that localizes to the nucleus and most likely competes with Xbp-1s, and so acts in a dominant negative fashion. The second mutant, termed PP2LH, was engineered to perturb the secondary structure of Xbp-1u, consequently alter its interactions with the degradation machinery, and thus impede its degradation. This approach was only partially successful but did yield a slightly more stable Xbp-1u.
We analyzed the expression of these constructs in Xbp-1 Ϫ/Ϫ MEFs to assess their contribution to the UPR. We observed the synthesis of the Xbp-1 stop mutant even under extreme conditions of ER stress, induced by tunicamycin treatment. This indicates that under physiological conditions Xbp-1u is synthesized constitutively. Indeed, under physiological induction of Xbp-1 splicing, as in the case of B cell differentiation into plasma cells, or in the course of viral infection, the vast majority of the mRNA of Xbp-1, nonetheless, remains in its unspliced form. It suggests that if Xbp-1s and Xbp-1u would have had comparable stabilities, then at any given time Xbp-1u expression would dominate, which might have dire consequences for a stressed cell to respond by induction of the UPR.
The Xbp-1 mutants allowed us to address this possibility. When we analyzed the induction of UPR target genes in MEFs reconstituted with FIGURE 5. PP2LH mutant activates UPR target genes better than WT. Total poly(A) ϩ RNA was isolated from MEF cells either treated or not treated with tunicamycin (1 g/ml, 8 h). The levels of mRNA of Erdj4, Bip, and Chop were quantified by real-time PCR and expressed relative to the levels of ␤-actin mRNA.