BINDING PROTEIN Is a Master Regulator of the Endoplasmic Reticulum Stress Sensor/Transducer bZIP28 in Arabidopsis

BINDING PROTEIN (BiP) is a major chaperone in the endoplasmic reticulum (ER) lumen, and this study shows that BiP binds to the C-terminal tail of the stress sensor/transducer bZIP28, a membrane-associated transcription factor, retaining it in the ER under unstressed conditions. In response to ER stress, BiP dissociates from bZIP28, allowing it to be mobilized from the ER to the Golgi where it is proteolytically processed and released to enter the nucleus. Under unstressed conditions, BiP binds to bZIP28 as it binds to other client proteins, through its substrate binding domain. BiP dissociates from bZIP28 even when bZIP28 ’ s exit from the ER or its release from the Golgi is blocked. Both BiP1 and BiP3 bind bZIP28, and overexpression of either BiP detains bZIP28 in the ER under stress conditions. A C-terminally truncated mutant of bZIP28 eliminating most of the lumenal domain does not bind BiP and is not retained in the ER under unstressed conditions. BiP binding sites in the C-terminal tail of bZIP28 were identi ﬁ ed in a phage display system. BiP was found to bind to intrinsically disordered regions on bZIP28 ’ s lumen-facing tail. Thus, the dissociation of BiP from the C-terminal tail of bZIP28 is a major switch that activates one arm of the unfolded protein response signaling pathway in plants.

INTRODUCTION BINDING IMMUNOGLOBULIN PROTEIN, or simply Binding Protein (BiP), known in animal systems as 78-kD Glc-regulated protein or heat shock 70-kD protein 5, and calnexin/calreticulin are the two major chaperone systems in the endoplasmic reticulum (ER) lumen (Otero et al., 2010). BiP is primarily involved in the maturation and folding of nonglycosylated proteins (Hendershot, 2004). BiPs form complexes with the HSP40-like cochaperones containing J domains (ERdj3) and stromal-derived factor-2 (Jin et al., 2008;Nekrasov et al., 2009;Schott et al., 2010). These BiP complexes maintain nascent proteins in a competent state for subsequent folding and oligomerization (Anelli and Sitia, 2008).
BiP also plays an important role in the unfolded protein response (UPR) by regulating stress transducers, such as ACTIVATING TRANSCRIPTION FACTOR6 (ATF6), Protein kinase RNA-like ER kinase, and Inositol Requiring Enzyme1 (IRE1) in animal cells (Bertolotti et al., 2000;Shen et al., 2002). In response to stress, BiP and ATF6 rapidly dissociate, and ATF6 becomes a cargo in the ER-to-Golgi trafficking system . Shen et al. (2002) observed that BiP binds to three regions in the lumenal domain of ATF6. They reasoned that binding to these sites might mask Golgi localization signals (GLSs) on ATF6. They located the GLSs in two regions of the lumenal domain of ATF6 by showing that when these regions were deleted, transport of ATF6 to the Golgi and its processing by the Golgi-resident proteases were blocked. To demonstrate that BiP masked GLSs under unstressed conditions, they developed constructs in which the BiP binding sites were deleted, but putative GLSs were retained. These constructs were constitutively translocated to the Golgi, supporting the idea that BiP retains ATF6 in the ER by blocking its GLSs .
In plants, BiP is reported to play a role in the defense against various stresses. In particular, the overexpression of BiP in Nicotiana tabacum (tobacco) has been shown to protect against tunicamycin (TM) inhibition of seed germination and against water stress during plant growth (Alvim et al., 2001). When BiP was overexpressed in Glycine max (soybean), leaves of transgenic plants were more resistant to wilt and more tolerant of the loss of water potential than control plants when subjected to drought or osmotic stress (Valente et al., 2009). The greater stress resistance has been attributed to an attenuation of a cell death signal produced by ER and osmotic stress through N-RICH PROTEIN (NRP) and NAC6-mediated pathways (Reis et al., 2011). This implies that BiP is a negative regulator of stress-induced NRPmediated cell death. In plants, BiP has also been shown to be a limiting factor in the folding of certain secreted proteins under ER stress conditions. Leborgne-Castel et al. (1999) reported that the production of a-amylase was reduced in a tobacco transient expression system subjected to ER stress but could be restored by cotransfection with BiP. BiP2 loss-of-function mutants are defective PATHOGENESIS RELATED1 protein secretion in response to salicylic acid elicitation and are more susceptible to Pseudomonas syringae infection (Wang et al., 2005).
BiP plays a dual role in plant UPR. First, the genes encoding BiP are upregulated by the UPR. There are three BiP coding genes in Arabidopsis thaliana, and BiP3, in particular, is most highly upregulated by abiotic stress or by ER stress agents (Koizumi, 1996;Martínez and Chrispeels, 2003;Iwata and Koizumi, 2005;Liu et al., 2007b;Iwata et al., 2008Iwata et al., , 2010Tajima et al., 2008;Liu and Howell, 2010). Since BiP is a chaperone, it is thought to mitigate stress by binding to misfolded proteins in the ER, preventing their aggregation during refolding processes. Second, BiP is thought to regulate the activity of the ER stress sensor/transducers, bZIP17 and bZIP28, plant homologs of mammalian ATF6 and related factors. bZIP17 and bZIP28 like ATF6 are membrane-associated transcription factors activated by various stresses in a process that involves their mobilization from the ER to the Golgi where they are processed and released by site 1 and site 2 proteases (S1P and S2P) (Liu et al., 2007a(Liu et al., , 2007b.
In this study, we examined the role of BiP in retaining bZIP28 in the ER under unstressed conditions and allowing it to mobilize to the nucleus in response to stress to upregulate target genes. BiP binds to the C-terminal, lumen-facing tail of bZIP28, and in response to ER stress, BiP dissociates from bZIP28, releasing it from the ER. The deletion of the bZIP28 C-terminal tail prevents bZIP28's retention in the ER under unstressed conditions and allows it to relocate to the nucleus to constitutively upregulate stress response genes.

BiP Binds to bZIP28 under Unstressed Conditions
BiP binds to bZIP28 in unstressed Arabidopsis seedlings as demonstrated by the coimmunoprecipitation of BiP with myc-bZIP28 in transgenic lines expressing myc-bZIP28 ( Figure 1A). Arabidopsis encodes three BiP isoforms, BiP1 to BiP3, that are predicted to be ER lumenal proteins (see Supplemental Figure 1 online). The anti-BiP antibody used in these studies does not discriminate between the BiPs. Therefore, to determine which isoforms of BiP bind to bZIP28, we FLAG-tagged BiP1 and BiP3 and used the epitope-tagged forms in immunoprecipitation experiments with myc-bZIP28 transiently expressed in Nicotiana benthamiana leaves. BiP2 was not used in these experiments because BiP1 and BiP2 are nearly identical in sequence (see Supplemental Figure 1 online). We observed that both BiP1-flg and BiP3-flg coimmunoprecipitate with myc-bZIP28 ( Figure 1B). We were also able to pull down myc-bZIP28 using BiP antibodies from unstressed Arabidopsis plants (see Supplemental Figure 2 online). These results confirmed that bZIP28 does indeed interact with BiP.

Characteristics of BiP Binding to bZIP28
We performed further experiments to determine the characteristics of BiP binding to bZIP28. BiP appears to bind myc-bZIP28 as it does to other client proteins. BiP coimmunoprecipitated (A) myc-bZIP28 was detected in extracts and immunoprecipitates (iP) from of roots of unstressed 7-d-old transgenic (T) and nontransgenic control (NT) Arabidopsis seedlings. Immunoblots were probed with anti-BiP and anti-myc antibodies. The anti-BiP antibody that was used for immunoprecipitations did not bind to agarose beads alone. (B) Both BiP1-flg and BiP3-flg bind to myc-bZIP28. myc-bZIP28 was immunoprecipitated from extracts of N. benthamiana leaves transiently expressing BiP1-flg and BiP3-flg. Immunoblots were probed with anti-flg and anti-myc antibodies. (C) BiP bound to myc-bZIP28 extracted from myc-bZIP28 expression lines is released by ATP. myc-bZIP28 was immunoprecipitated with anti-myc antibodies and incubated for 30 min with or without 2 mM ATP and 2 mM MgCl 2 . Immunoblot was probed with anti-BiP and anti-myc antibodies. (D) Mutant BiP with a defect in substrate binding does not bind bZIP28. myc-bZIP28 coexpressed with BiP1-flg or BiP1P503L-flg in the transient expression system was immunoprecipitated with anti-myc antibodies. Immunoblot was probed with anti-flg and anti-myc antibodies.
with carboxypeptidase Y star-green fluorescent protein (CPY*-GFP), a known BiP substrate (Izawa et al., 2012), when it was expressed from a transgene in Arabidopsis (see Supplemental Figure 3 online). Client proteins can be released from BiP by ADP/ATP exchange in BiP's N-terminal nucleotide binding domain . BiP was released from myc-bZIP28 upon incubation of the immunoprecipitate with ATP and MgCl 2 and was not released in the absence of ATP ( Figure 1C). In addition, BiP did not bind to myc-bZIP28 when BiP's substrate binding domain was disabled by substituting Pro at position 503 for Leu. In mammalian systems, the equivalent substitution of a Pro for Leu at position 495 (BiPP495L) gives rise to a form that is defective in substrate binding (Shen et al., 2005). When Arabidopsis BiP1P503L-flg was expressed along with myc-bZIP28, -BiP1P503L-flg did not coimmunoprecipitate with myc-bZIP28 ( Figure 1D). Thus, we concluded that under unstressed conditions, BiP binds to bZIP28 similarly to the manner in which BiP binds to other client proteins.

BiP Dissociates from bZIP28 under Stress
When seedlings are treated with ER stress agents, such as TM, bZIP28 is transported from the ER to the nucleus via the Golgi apparatus (Liu et al., 2007b;Srivastava et al., 2012). We were interested in determining whether BiP dissociates from bZIP28 in response to stress and whether the dissociation of BiP is correlated with other events involved in the mobilization of bZIP28. Therefore, we performed coimmunoprecipitations of myc-bZIP28 (A) Seven-day-old Arabidopsis seedlings expressing myc-bZIP28 were treated with 2 mg/mL TM, and proteins were extracted (from whole seedlings) at times indicated following TM treatment. Extracts were immunoprecipitated with anti-myc antibodies, and immunoblot was probed with anti-BiP and anti-myc antibodies. The fast migrating band visualized by the anti-myc antibody is the proteolytically processed form of myc-bZIP28. (B) Confocal images of roots of Arabidopsis seedlings expressing YFP-bZIP28. Seedlings were treated for 1 h with 2 mg/mL TM. Roots were stained with propidium iodide to show cell outlines. Bars = 50µm. (C) BiP dissociates in response to stress from mutant forms of bZIP28 that are prevented from exiting the ER or from being released from Golgi bodies. Seven-day-old Arabidopsis expressing various forms of myc-bZIP28 were treated with TM for 1 h, and proteins were extracted, immunoprecipitated with anti-myc, subjected to immunoblot analysis, and probed with anti-BiP and anti-myc antibodies. bZIP28KK320AA is blocked in exiting the ER, and bZIP28G329A is not proteolytically cleaved by S2P and therefore is not released from Golgi bodies. and BiP during a time course of movement of bZIP28 from the ER to the nucleus as described by Srivastava et al. (2012). BiP progressively dissociated from myc-bZIP28 during an hour after treating seedlings with TM ( Figure 2A; see Supplemental Figure 2 online). The loss of BiP from myc-bZIP28 immunoprecipitates was due to dissociation and not to degradation of BiP because BiP levels did not decline during the time course of these experiments (see Supplemental Figure 4 online). At ;30 min after the start of stress treatment, bZIP28 begins to exit the ER (Srivastava et al., 2012), and by 1 h, myc-bZIP28 was being proteolytically processed ( Figure 2A) and yellow fluorescent protein (YFP)-bZIP28 appeared in the nucleus ( Figure 2B; see Supplemental Figure 5 online). Since BiP dissociation and YFP-bZIP28 movement occurred at about the same time, we asked whether the dissociation of BiP depends on exit of bZIP28 from the ER. To do so, we looked for the dissociation of BiP from myc-bZIP28KK320AA, a mutant that is incapable of exiting the ER. bZIP28KK320AA has an altered pair of Lys residues on the cytoplasmic side of the membrane, which impedes its interaction with COPII vesicle components and its exit from the ER (Srivastava et al., 2012). BiP bound to myc-bZIP28KK320AA under unstressed conditions and dissociated from it following TM treatment ( Figure 2C). Since it takes about an hour of TM treatment for most of BiP to dissociate from wild-type myc-bZIP28, the question can be asked whether the dissociation is only apparent because the myc-tagged N terminus of bZIP28 is cleaved off in the Golgi. To test this, we determined whether BiP dissociated from myc-bZIP28G329A, a form of bZIP28 that has a mutation in its transmembrane domain, making it resistant to proteolytic cleavage by S2P and incapable of being released from the Golgi (Srivastava et al., 2012). BiP dissociated from myc-bZIP28G329A following TM treatment ( Figure 2C). Therefore, BiP dissociates from bZIP28 in response to ER stress whether or not its exit is blocked from either the ER or Golgi.

BiP Binding Retains bZIP28 in the ER
To test whether BiP retains bZIP28 in the ER by binding to bZIP28's C-terminal tail, two major truncation constructs were developed: YFP-bZIP28D591, which eliminated most of the C terminus downstream of the S1P site; and YFP-bZIP28D355, which removed most of bZIP28's C-terminal tail residing in the ER lumen ( Figure 3A). We tested these truncations of bZIP28 for their ability to bind BiP in stable transgenic overexpression lines. BiP bound normally to myc-bZIP28D591 but did not coimmunoprecipitate with myc-bZIP28D355, the construct lacking most of the C-terminal tail ( Figure 3B). We also examined the localization and movement of YFP-tagged versions of the two truncation constructs. YFP-bZIP28D591 was retained in the ER under unstressed conditions and migrated to the nucleus only when seedlings were subjected to TM stress ( Figures 3C and 3D). However, YFP-bZIP28D355 was located in the nucleus even under unstressed conditions ( Figure 4A) and also under TM stressed conditions ( Figure 4B). To demonstrate that YFP-bZIP28D355 is transcriptionally active in unstressed cells, we looked for the expression of a bZIP28 target gene, BiP3 (Liu et al., 2007b). We found that unstressed seedlings expressing the full-length form of YFP-bZIP28 did not upregulate BiP3, but seedlings expressing YFP-bZIP28D355 upregulated Bip3 expression even under unstressed conditions (see Supplemental Figure 6 online). Thus, bZIP28D355 does not bind to BiP and is not retained in the ER, but instead migrates to the nucleus in unstressed seedlings where it upregulates stress response genes. It is possible that the YFP-bZIP28D355 truncation construct fails to be retained in the ER by shortcutting its normal route, which involves translocation through the secretion pathway to the Golgi where it would be released by S1P and S2P proteolysis. To test whether YFP-bZIP28D355 follows the conventional route to the nucleus, we introduced this construct into an s2p mutant background (Che et al., 2010). Under unstressed conditions, YFP-bZIP28D355 was prevented from moving constitutively into the nucleus ( Figures 4C and 4D). In a wild-type background and under unstressed conditions, YFP-bZIP28D355 was observed not only in the nucleus, but also in small punctate structures that colocalize with a Golgi body marker (see Supplemental Figure  7 online). Thus, YFP-bZIP28D355 moved via the Golgi to the nucleus, and this was prevented by blocking proteolysis of the bZIP28 protein in an s2p mutant.

Identifying BiP Binding Regions in the C-Terminal Tail of bZIP28
Based on the results described above in which bZIP28D591 bound BiP and was retained in the ER under unstressed conditions, but bZIP28D355 did not bind BiP and was not retained in the ER, we focused on a region (positions 376 to 555) between these sites to identify the BiP binding and ER retention domains ( Figure 3A). We initially attempted to use a yeast two-hybrid system to study the interaction between BiP and bZIP28. We had to abandon this approach due to high backgrounds in controls with the lumenal domain proteins. Blond-Elguindi et al. (1993) developed a means for identifying binding sites for mammalian BiP based on a phagepanning assay for octapeptides that bind to BiP. From their findings, they generated a scoring algorithm to identify potential BiP binding sites. However, their scoring algorithm did not allow us to unambiguously identify potential BiP binding sites in the lumenal tail of bZIP28. To better identify potential BiP binding sites in bZIP28, we conducted phage panning experiments using purified, immobilized Arabidopsis BiP1-His (see Supplemental Figure  8 online) and a phage library of 12 overlapping peptides (see Supplemental Table 1 online) from the lumenal domain of bZIP28 displayed in a M13 phage display system ( Figure 5A).
We subjected the overlapping peptide phage library to four rounds of panning. Among the 25 phages sequenced from the second round of panning, we recovered eight of the 12 input (A) Overlapping peptides from residues 376 to 555 in the lumenal domain of bZIP28 were displayed in M13 phage. Tendency for intrinsic disorder in the lumenal domain of bZIP28 is plotted against the map of the lumenal domain for bZIP28. The tendency for disorder was determined by IUPred (http:// iupred.enzim.hu/pred.php).

of 14
The Plant Cell sequences, and the phage displaying peptide 376 led all others in frequency of recovery ( Figure 5B). In the third and fourth rounds, we recovered fewer sequences, but the recovered phage were enriched further for peptides 376 and 471. Thus, in the panning assay, peptides 376 and 471 clearly outcompeted the others in binding to immobilized BiP1. The results of the panning were confirmed by scoring the binding of phage separately to immobilized BiP1 in an ELISA assay ( Figure 5C). When the peptides were not challenged competitively, the difference in the binding of phage displaying peptides 376 and 471 compared with phagedisplaying peptides 441 and 501 was more modest. Nonetheless, the binding of these peptides to immobilized BiP1 was comparable to the binding of phage displaying the entire region covered by the peptide library. Also, to demonstrate that the binding of the peptides to immobilized BiP is not a contextual artifact, we performed protein overlay assays. The four peptides, 376, 441, 471, and 501, were fused to a glutathione S-transferase (GST) tag in a pET42a vector and expressed in BL-21 cells. The GST-tagged peptides were pulled down using glutathione beads and loaded on SDS gels. The blots were probed with purified BiP1-His followed by primary and secondary antibodies to detect the binding of BiP1 to the individual peptides. The four GST-tagged peptides were capable of binding to BiP1 ( Figure 5D). The GST control and peptides that failed to bind to immobilized BiP in the panning procedures did not show any binding in the overlay assays (see Supplemental Figure 9 online). In surveying the region covered by the phage panning analysis, it was observed that the peptides that showed the greatest binding to BiP were derived from regions of the bZIP28 lumenal tail that had the highest tendency for disorder ( Figure 5A). In general, the bZIP28 C-terminal tail has three broad regions, labeled R1 to R3, that showed greater tendency for disorder. The tendency for disorder was determined by IUPred (http://iupred. enzim.hu/pred.php), a prediction program for disorder tendency based on the estimation of stabilizing contacts in the pairwise interactions between residues. Regions with fewer inter-residue interactions are predicted to be intrinsically unfolded regions (Dosztányi et al., 2005). The bZIP28 tail peptides that showed the greatest binding affinity for BiP1 were located in R1 and R2. These regions of bZIP28 showed little correlation with hydrophobicity index (see Supplemental Figure 10 online). It has been reported that BiP tends to bind to solvent-accessible hydrophobic patches on proteins (Caramelo et al., 2003). That does not appear to be case in the binding of BiP to the lumenal tail of bZIP28.
To determine if bZIP28 bearing the R1 region alone at its C-terminal tail could bind BiP, we developed another bZIP28 truncated construct, myc-bZIP28D450 ( Figure 3A). This construct, along with BiP1-flg, was coexpressed in a tobacco leaf transient expression assay, and it was found that BiP1-flg coimmunoprecipitated with myc-bZIP28D450 in the leaf extracts ( Figure 5E). We concluded that myc-bZIP28 bearing only the R1 region, which shows tendencies for disorder, can interact with BiP in planta. These results also demonstrated that BiP interacts independently with the different regions of the bZIP28 lumen-facing tail.

Effect of Overexpression of BiP on the Mobilization of bZIP28
We manipulated the levels of BiP expression to explore its impact on bZIP28 mobilization. Arabidopsis has three BiP genes, but since BiP1 and BiP2 are nearly identical, we overexpressed BiP1 and BiP3 with the 35S promoter in lines containing YFP-bZIP28 (see Supplemental Figure 11 online). To demonstrate that the overexpressed forms are appropriately localized at the subcellular level, we showed that BiP1-YFP-HDEL colocalized with an ER marker, sPMcherry (see Supplemental Figure 12 online). In the wild-type lines (not containing the BiP1 or BiP3 transgenes), YFP-bZIP28 was located in the ER in unstressed seedlings ( Figure  6A) and translocated to nuclei after 2 h of TM treatment ( Figure  6B). However, in BiP1 and BiP3 overexpressors, most YFP-bZIP28 was retained in the ER even after 2 h of TM treatment ( Figures 6C  and 6D). Although BiP1 and BiP3 were overexpressed, TM treatment still induced ER stress as evidenced by the splicing of bZIP60 (see Supplemental Figure 13 online and Deng et al., 2011). Nonetheless, BiP1 or BiP3 overexpression interfered with the release of YFP-bZIP28 from the ER. As a control, we showed that overexpression of BiP1P503L-flg, the construct that is defective in substrate binding, did not prevent bZIP28's movement to nuclei following stress treatment ( Figure 6E).
To determine if we could further restrain the stress mobilization of YFP-bZIP28 by preventing its dissociation from BiP, we introduced into Arabidopsis mutated forms of BiP (BiPG235D) that are known in other systems to interfere with ATP binding and the release of client proteins   (B) Recombinant phages were pooled and panned against immobilized BiP1-His in four rounds of panning. At each round, bound phages were released and the inserts encoding the bZIP28 peptides were sequenced. The frequency in recovering phage expressing the various peptides in progressive rounds of screening is shown. Red asterisks indicate the peptides in phage recovered with the highest frequency in the fourth round of panning. (C) Separate recombinant phage lines were incubated with immobilized BiP1-His, and bound phage were quantified in an ELISA assay. Error bars indicate SE. (D) Overlay immunoblot demonstrating that soluble BiP1-His binds to GST-tagged bZIP28 peptides. The four peptides (441, 471, 376, and 501) enriched in panning were tagged with GST, purified by binding to glutathione beads, eluted, subjected to SDS-PAGE, and transferred to a nitrocellulose filter. The filter was incubated with purified BiP1-His, washed, and incubated with a primary anti-BiP antibody and then a secondary antibody was used to detect BiP binding. The GST-tagged peptides pulled down with glutathione beads and stained with Coomassie blue were used as a loading control.  Snowden et al., 2007). In response to TM-induced stress, the expression of BiP1G233D or BiP3G247D transgenes appeared to prevent the dissociation from myc-bZIP28 in pull-down experiments ( Figure 7A). The overexpression of BiP1G233D with reduced capacity for client protein dissociation prevented the processing of myc-bZIP28 in response to DTT treatment ( Figure  7B) and also effectively blocked the nuclear relocation of YFP-bZIP28 in response to TM-induced stress ( Figures 7C and 7D). DTT and not TM was used as a stress agent in this experiment because TM treatment produces additional nonglycosylated forms of myc-bZIP28 that make the gel patterns difficult to interpret. In our hands, DTT has been just as effective as TM in eliciting ER stress. The overexpression BiP3G247D similarly prevented the relocation of YFP-bZIP28 to nuclei in response to TM-induced stress ( Figure 7E). Thus, we conclude that overexpression of BiP, particularly forms of BiP that interfere with the release of client proteins, prevents the normal mobilization of bZIP28 from the ER to the nucleus by binding to bZIP28 and not allowing its release.  Figure 1D) treated with 2 mg/mL TM for 2 h. YFP fluorescence was imaged by confocal microscopy in roots counterstained with propidium iodide. Bars = 50 µm. (A) Immunoprecipitation of myc-bZIP28 in extracts from Arabidopsis lines expressing flag-tagged mutated forms of BiP (BiP1G233D-flg and BipG247D-flg) that are predicted to be defective in binding ATP and releasing client proteins. Seedlings were treated with 2 mg/mL TM for 2 h to determine whether the BiP constructs dissociated from myc-bZIP28. Immunoblot of extracted proteins was probed with anti-flg and anti-myc.
(B) Processing of YFP-bZIP28 in wild-type (wt) and BiP1G233D-flg overexpressing (OE) backgrounds. Seedlings were either untreated or treated with 2 mM DTT for 2 h. Immunoblot of extracted proteins was probed with anti-myc and antitubulin as a loading control.
(C) YFP-bZIP28 localization in wild-type Arabidopsis seedlings lines treated with TM for 2 h. (D) and (E) YFP-bZIP28 localization in lines overexpressing BiP1G233D (D) or BiP3G247D (E), mutants with defects in ATP binding and release of client proteins. Lines were treated with TM for 2 h. YFP fluorescence was imaged by confocal microscopy in roots counterstained with propidium iodide. Bars = 50 µm.

of 14
The Plant Cell

Effect of BiP Knockouts on bZIP28 Mobilization
Similar experiments were conducted in BiP knockout lines containing YFP-bZIP28. Our interest here was whether the knockout of BiP genes would prevent the retention of bZIP28 in the ER under unstressed conditions. Homozygous T-DNA insertion lines are available for all three BiP genes; however, multiple mutants, such as the bip1 bip2 double mutant, are not viable. In an effort to reduce BiP levels as much as possible, we obtained lines homozygous for either bip1 or bip2 but heterozygous for the other locus as described by Maruyama et al. (2010). The BiP3 T-DNA homozygous insertion line (GK-075D06-011890) was obtained from the Nottingham Arabidopsis Stock Centre and found not to produce BiP3 transcripts under stress conditions (see Supplemental Figure  14 online). Compared with the wild type, detectable amounts of YFP-bZIP28 escaped from the ER to nuclei under unstressed conditions in the bip1/bip1 bip2/+ and in the bip1/+ bip2/bip2 and bip3 mutant lines, indicating a reduced ability of the BiP knockout mutants to retain bZIP28 in the ER ( Figures 8A to 8D). We observed processed myc-bZIP28 in the various knockout mutant lines under unstressed conditions, confirming these results and indicating that bZIP28 is constitutively activated in these lines ( Figure 8E).

DISCUSSION
Our results show that BiP binds to bZIP28 under unstressed conditions and appears to bind to it in the same manner as it does to other client proteins, such as CPY*. We have also shown that BiP dissociates from bZIP28 in response to ER stress, and in doing so bZIP28 is activated, allowing it to make its way to the nucleus where it upregulates stress response genes ( Figures 9A and 9B). The release of bZIP28 from BiP corresponds closely with its exit from the ER, but its release is not dependent on the trafficking of bZIP28 from organelle to organelle. In animal systems, it is thought that BiP binding retains ATF6 in the ER under nonstress conditions, putatively by blocking ATF6's GLSs, preventing it from being transported through the secretory pathway . There are different ideas as to how BiP relinquishes its hold on ATF6 under stress conditions (Parmar and Schröder, 2012). One idea, called the dynamic competition model, is that BiP is competed away from ATF6 by misfolded proteins in the ER (Harding et al., 2002;Kaufman et al., 2002;Kimata et al., 2003). In this model, BiP bound to ATF6 is thought to be in equilibrium with free BiP and BiP associated with misfolded proteins, and when unfolded proteins accumulate in the ER as a result of stress, the binding of BiP to ATF6 would be competed away. Our results in Arabidopsis support such a dynamic competition model. We observed that the overexpression of BiP1 or BiP3 delays or blocks the translocation of YFP-bZIP28 to the nucleus. When BiP is overexpressed, bZIP28 appears not to be fully deployed in response to stress but seems to be largely retained in the ER. The overexpression effect is enhanced using mutant versions of BiP, which interfere with the release of substrates, as has also been shown by Snowden et al. (2007). On the other hand, overexpression of Bip1P503L defective in substrate binding does not block the movement of bZIP28. In animal systems, similar results were also observed by Shen et al. (2002), who overexpressed BiP in HeLa cells and found that ATF6 translocated to the Golgi, but processing by S1P and S2P, the Golgi resident proteases, was delayed following stress treatment. Also, overexpression of BiPT37G, an ATPase mutant compromised in the release of bound proteins, prevented the translocation of ATF6 to the Golgi and its processing by the Golgi-resident proteases . The BiP knockout mutants (bip1/bip1 bip2/+, bip1/+ bip2/bip2, and bip3/bip3) allow some migration of bZIP28 to nuclei under unstressed conditions. (A) YFP-bZIP28 in wild-type unstressed Arabidopsis seedlings. (B) to (D) YFP-bZIP28 in the bip1/bip1 bip2/+ (B), bip1/+ bip2/bip2 (C), and bip3/bip3 (D) mutant lines. YFP fluorescence was imaged by confocal microscopy in roots stained with propidium iodide. Arrows point out nuclear localization. Bars = 50 µm. (E) Proteolytic processing of mycbZIP28 in wild-type seedlings in response to 2 mg/mL TM treatment for 2 h. Processing of myc-bZIP28 in untreated BiP knockout lines as indicated.
BiP Regulates bZIP28 in Endoplasmic Reticulum Stress 9 of 14 All three BiP genes cannot be knocked out because even homozygous bip1 bip2 mutants are not viable (Maruyama et al., 2010). We observed that these partial knockdown BiP mutants appear to be defective in retaining bZIP28 in the ER under unstressed conditions. In all, our results involving the overexpression and underexpression of BiP have demonstrated the critical role of BiP in the retention and activation of bZIP28 in Arabidopsis.
Another model for the release of ATF6 from BiP in animal systems does not invoke dynamic competition, instead positing that the association is stable, but can be disrupted by a signal from misfolded proteins. Several arguments favor a stability model (Shen et al., 2005), with one being that association between BiP and ATF6 is stable enough to survive immunoprecipitations without the need for cross-linking. Second, if BiP is continually binding and dissociating from ATF6, then the T37G BiP mutant with a defect in ATPase activity (protein releasing activity) should build up on ATF6 when introduced into monkey kidney fibroblast cells. Apparently, it does not (Shen et al., 2005). Shen et al. (2005) compared the dissociation of BiP from ATF6 and from unassembled Ig heavy chains, which are also retained in the ER by their association with BiP. They found that in response to stress, BiP dissociated from ATF6, but not from unassembled Ig heavy chains. From this they hypothesized that ATF6 contains BiP binding and release elements, the latter referred to as stress-responsive domains.
Our experiments demonstrate that BiP binds to the C-terminal, lumen-facing tail of bZIP28, and when the tail is eliminated as in bZIP28D355, the protein is not retained in the ER and behaves like an activated form of bZIP28. Truncated bZIP28 constitutively moves to the nucleus where it upregulates stress genes, including BiP3. The movement takes place via the Golgi and requires S2P processing. As an aside, it is interesting that YFP-bZIP28D355 appears to require S2P to relocate to the nucleus and is apparently a substrate for S2P. YFP-bZIP28D355 lacks most of its C-terminal tail and its S1P site. Cleavage at the S1P site is usually considered to be a prerequisite for S2P cleavage (Espenshade et al., 1999;Shen and Prywes, 2004). The implication from this is that S1P cleavage is not required for S2P proteolysis as long as the C-terminal tail on bZIP28 has been removed. BiP is a chaperone that binds to nascent and misfolded proteins. So why does BiP bind to the lumen-facing tail of bZIP28? Our findings suggest that BiP binds to intrinsically disordered regions of the bZIP28 tail, regions that may have the properties of unfolded proteins. There are three regions in bZIP28's C-terminal tail that are predicted to be more disordered based on the algorithm at IUPRED (http://iupred.enzim.hu/pred.php), and peptides from two of these regions, R1 and R2, bind to BiP1 more avidly. The amino acid sequence in R1 is not conserved among various plant species; however, the pattern for disorder in that region is conserved among several species. It should be pointed out, nonetheless, that there is not much support in the literature for chaperones to bind to intrinsically disordered proteins; in fact, one meta-analysis of interactomes shows a negative correlation between the interactions between chaperones and intrinsically disordered proteins (Hegyi and Tompa, 2008). Foresti et al. (2003) identified regions of bean (Phaseolus vulgaris) phaseolin that bound to BiP in plant cells. Earlier studies had shown that phaseolin mutants lacking the C-terminal domain failed to assemble into trimers (Pedrazzini et al., 1994(Pedrazzini et al., , 1997Frigerio et al., 2001). Based on the supposition that BiP binding increases the efficiency of the folding and assembly of secretory proteins, they searched for BiP binding sites in the C-terminal domain. To look for BiP binding sites, they appended the C-terminal domain containing three a-helical segments onto GFP and showed that the domain was sufficient to bind BiP. Foresti et al. (2003) attributed the binding to the hydrophobicity of the C-terminal domain. However, C-terminal domain of phaseolin is not more hydrophobic than several other regions of the protein. The C terminus is predicted to be somewhat more disordered than the rest of the protein, but the subterminal region to which BiP also binds is not particularly disordered. It is possible that preferred BiP binding sites might depend on other properties in addition to the tendency for disorder. Those properties need further definition because existing algorithms are not sufficiently predictive to identify BiP binding sites (Blond-Elguindi et al., 1993).
To sum up, our experiments demonstrate that the BiPs in Arabidopsis are master regulators of the membrane-associated transcription factor arm of the UPR signaling pathway in response to stress and function by binding to the intrinsically disordered regions of the C-terminal tail of bZIP28.

Plant Material and Stress Treatments
Arabidopsis thaliana ecotype Columbia-0 was used in this study. Seeds were stratified at 4°C for 3 d prior to germination. Plants were grown under continuous white light at 23 to 25°C in soil or on Linsmaier/Skoog (LS) medium (13 LS salts, 1% Suc, and 0.8% agar). Agrobacterium tumefaciensmediated transformation was performed using the floral dip method (Bechtold et al., 1993). Agrobacterium strain GV3101 was used in all transformation experiments. For stress treatment, 7-d-old seedlings were treated with 2 mg mL 21 TM or 2 mM DTT in LS medium for the times indicated.
T-DNA mutant line with an insertion in the first exon of BiP3 (GK-075D06-011890) was obtained from the Nottingham Arabidopsis Stock Centre (N744092). The T-DNA lines were genotyped by PCR using genespecific primers and a left border T-DNA primer (see Supplemental Table 2 online). The insertion lines were further treated with TM for 30 min to affirm the loss of BiP3 transcripts. The bip1 and bip2 mutant lines were kindly provided by the Shuh-ichi Nishikawa Lab (Nagoya University). The lines were described by Maruyama et al. (2010). Since the double homozygous lines are female sterile, they are maintained as bip1/+ bip2/bip2 or bip1/bip1 bip2/+ strains. The s2p mutant line was described by Che et al., (2010) and was kindly provided by Ping Che (DuPont Pioneer Hi-Bred International).

Plasmid Construction
Plasmid constructs were prepared as summarized in Supplemental Table  3 online using the primers listed in Supplemental Table 2 online.
The open reading frame of bZIP28 (At3g10800) was amplified from 1-week-old seedlings by RT-PCR using primer pair Zip28 (see Supplemental  Table 2 online). The product was cloned into pSKM36 (Liu et al., 2007b) at the AscI and SpeI sites, resulting in pMZIP28, which was used as a template for further constructs. pMZip28 was tagged with monomeric GFP by amplifying it from a GFP vector using GFPAsc primers and inserting it into the AscI site to generate N-terminal monomeric GFP-tagged bZIP28. A 43 epitope myc tag (EQKLISEEDLRN) was amplified from pSKM36 using the primer MycAsc (including the ATG before the myc sequence) and inserted into pMZIP28 at the AscI site to generate N-terminal myc-tagged bZIP28. The construction of bZIP28KK311AA, bZIP28KK320AA, and bZIPG329A mutants has been described previously by Srivastava et al. (2012). The vectors pSKM36 and pSKY are described by Srivastava et al. (2012). The open reading frames of BiP1 and BiP3 were amplified from 10-d-old seedlings (for Bip3; cDNAs were generated from the RNA of seedlings treated for 30 min with TM) using the primers Bip1SKHDELStopPr and Bip3HDELStopPr listed in Supplemental Table 2 online. The templates were called Bip1SKHDELStop and Bip3HDELStop. The Flag and YFP tags with HDEL sequences (as required) were attached to the C terminus of BiP1 and BiP3 using the enzyme SpeI. The BiP1G233D and BiP3G247D mutations were prepared by overlapping PCR using the primers listed and the templates Bip1SKHDELStop and Bip3HDELStop. The signal peptide chitinase mCherry marker was prepared by amplifying the CSP-YFP-HDEL obtained from Edgar Cahoon (University of Nebraska) (Chen et al., 2008) using the primers spMcherry and ligated into the vector pCHF1 with HDEL sequence at its C terminus to retain the protein in the ER.
Cloning of the peptides from the C terminus of bZIP28 was performed in pCANTB SE at the SfiI and NotI sites using pMZIP28 as template and the primers are listed in Supplemental Table 2 online. bZIP28 was cloned in pET28 at the BamHI and NotI sites and expressed in BL-21 cells.

Transient Transformation in Tobacco
Transient expression in tobacco (Nicotiana tabacum) leaf epidermis was performed as described previously (Batoko et al., 2000) using Agrobacterium (OD 600 = 0.05) containing the binary vectors pSKY and pSKM36.

Immunoprecipitations
To prepare plant extracts for immunoprecipitations of myc-tagged proteins, 5 g of seedlings was ground in liquid nitrogen and suspended in 25 mM Tris/HCl, pH 7.2, 150 mm NaCl, 0.1% Nonidet P-40 (Calbiochem), and 10% glycerol. Anti-c-myc agarose conjugate (Sigma-Aldrich) was added to the filtered lysate and incubated for 2 h at 4°C. The mixture was rotated at 4°C for 2 h, and beads were washed four times for 5 min each with the buffer described above. The recovered beads were resuspended and boiled in 23 SDS buffer for 5 min, and the eluted material was subjected to immunoblotting. Plant crude extracts (input material for the immunoprecipitation reactions) were analyzed for the presence of the tagged protein using c-myc antibody (9E10; Santa Cruz Biotechnology) as probe. BiP-bound proteins were immunoprecipitated from 2 g of plant material using a BiP antibody, ADI-SPA-818 D from Enzo Life Sciences. The mixture was rotated at 4°C for 2 h. This was followed by binding to protein agarose 916-157 (Millipore), and bound beads were washed four times for 5 min each with the buffer described above followed by loading on SDS gel. Flag antibodies (F1804; Sigma-Aldrich) and GFP antibodies (11814460001; Roche) were used in immunoprecipitations and in probing immunoblots.

Immunoblot Analysis
Immunoblots were performed as described by Liu et al. (2007aLiu et al. ( , 2007b. To examine the processing of bZIP28, plants were grown vertically on Petri plates containing agar medium. Ten-day-old seedlings were treated with 2 mm DTT in LS medium, and 300 mg of root material was harvested from the treated plants. Roots were homogenized in liquid nitrogen, and 30 mg of protein was loaded per lane on gels. T8203 monoclonal anti-atubulin antibody from Sigma-Aldrich was used to detect tubulin as a loading control.

Confocal Microscopy
Subcellular localization and protein movement experiments for fluorescenttagged proteins were performed using a NikonC1si confocal scanning system attached to a 90i microscope (Nikon Instruments). Roots were used for microscopy from plants pretreated with TM, and untreated plants were used as controls. The roots were observed under 320 and 360 water lenses. Some roots were counterstained with 50 mg mL 21 propidium iodide and Syto Red. The emission signals for YFP, propidium iodide, and SYTO Red 59 were acquired using sequential scanning mode to eliminate crosstalk and emission signal bleed-through. Fluorescence emission was obtained by laser excitation of YFP at 488 nm, and for mCherry, propidium iodide, or Syto Red, excitation was at 591 nm. Emission was in the range 500 to 575 and 590 to 700 nm, respectively. Syto Red 59 (S-11341) was obtained from Molecular Probes.

Gene Expression Analysis
Total RNA was isolated from ground plant tissues using an RNeasy kit, treated with RNase-free DNase I, according to the manufacturer's instructions (Qiagen), and was quantified by 260/280-nm UV light absorption. A 1-mg portion of total RNA was reverse transcribed using the Supertranscript III RT kit (Invitrogen). A 2-mL volume of cDNA was used for RT-PCR. All primers are listed in Supplemental Table 2 online.

Construction of Phage-Displayed Overlapping Peptide Library
Twelve overlapping regions of the lumenal domain of bZIP28 were generated by PCR using bZIP28 cDNA as a template and forward and reverse primers to incorporate SfiI and NotI sites. The PCR fragments were then digested with SfiI and NotI, ligated into a similarly digested pCANTAB 5E phagemid vector, and subsequently transformed into Escherichia coli XL1-Blue cells (Stratagene). Single colonies from each plate were inoculated into 5 mL 2YT/carbenicillin/tetracycline media, grown to OD 600 of ;0.2 to 0.3 at 37°C and then infected with helper phage-VCSM13 (Stratagene). After 1 h, the cell culture was transferred to 25 mL 2YT/Kan media and further incubated overnight at 37°C. After removing cell debris by centrifugation, phage particles were precipitated from the supernatant using 7.5 mL 20% polyethylene glycol solution containing 2.5 M NaCl. The precipitate was resuspended in 1 mL PBS, and the phage concentration was determined by measuring absorbance at 268 l (OD 268 = 1.0 for a solution containing 5 3 10 12 phage per mL). A peptide library was prepared by mixing equal concentrations of individual phage peptide.

Production and Purification of BiP for Phage Panning Experiment
BiP1 was His tagged at its N terminus in the pET28a vector. The construct was introduced into E. coli strain BL21, and cells were induced with 300 mM isopropyl b-D-1 thiogalacopyranodise overnight at 16°C. His-tagged BiP1 was purified using nickel-nitrilotriacetic acid agarose beads (Qiagen). The incubation buffer contained 50 mM sodium phosphate buffer, pH 8.0, 300 mM NaCl, 20 mM imidazole, and 0.05% Tween. The mixture was rotated at 4°C for 2 h and washed four times with incubation buffer. The protein was eluted by increasing the imidazole concentration in the incubation buffer to 250 mM. The beads were resuspended and boiled in 23 SDS buffer for 5 min and subjected to SDS-PAGE to check for purity. The purified BiP-His was used for panning experiments, ELISA, and overlay immunoblotting.

Phage Panning
His-tagged BiP1 protein was immobilized in the wells of a Nunc Maxisorp ELISA plate by aliquoting 100 mL of protein at a concentration of 10 µg/mL in 50 mM NaHCO 3 , pH 9.6, at room temperature with gentle rotation for 2 h. Wells were then blocked with PBS containing 0.2% BSA for 1 h followed by three washings with PBS containing 0.05% Tween 20 (PBST) and then incubated with 100 mL of the phage peptide library from the lumenal domain of bZIP28 for 3 h at room temperature with gentle rotation. After removing unbound phage by washing with PBST five times, bound phage was eluted by incubating with 500 mL of 0.1 M HCl for 5 min at room temperature with shaking. The eluted phage were immediately neutralized by the addition of one-third phage volume of 1 M Tris-HCl buffer, pH 8.0, followed by infection of XL1-Blue cells (grown to <0.6 OD) with the phage. Infected XL1-Blue cells were incubated at 37°C for 20 min followed by addition of the helper phage (VCSM13) and again incubating at 37°C for 30 min. The phage-infected XL1-Blue cells were transferred into a conical flask containing 50 mL of 2YT medium containing 10 µg/mL of tetracycline and 100 µg/mL of ampicillin that was further incubated at 37°C overnight with shaking at 210 rpm. Phage was prepared as described earlier, yielding the first round of enriched phage. The entire process was repeated for four rounds, and after two, three, and four rounds, phageinfected XL1-Blue cells were grown on 2YT ampicillin plates. Plasmids were prepared from 25 randomly selected colonies and their DNA sequenced.

Phage ELISA for Binding Specificity
For phage ELISAs, 100 mL of His-tagged BiP1 protein and BSA (control) (10 µg/mL in 50 mM NaHCO 3 , pH 9.6) were immobilized in the wells of an ELISA plate at room temperature with gentle rotation for 2 h. The plate was then washed two times with PBS followed by blocking with PBST containing 0.2% BSA for 1 h. Subsequently, after three washings with PBST, wells were incubated with the 100 mL of phage displaying lumenal domain peptides diluted in PBST containing 0.2% BSA for 2 h at room temperature with gentle shaking. The plate was again washed three times with PBST followed by incubation with anti-M13 HRP conjugated antibody for 1 h. After washing four more times with PBST, bound phage in each well was detected by incubating for ;10 min with 50 mL of a solution containing 0.01% hydrogen peroxide and 0.8 mg/mL o-phenylenediamine dihydrochoride. Reactions were terminated by the addition of 50 mL 3M HCl, and absorbance of the developed yellow color was measured at 490 nm.

of 14
The Plant Cell

Overlay Immunoblotting
The four peptides (441, 471, 376, and 501) were tagged on their N terminus with a GST tag in pET42a vector. For in vitro GST pull-down assays, GST-441, GST-471, GST-376, GST-501, and GST-H6 (control) were bound to glutathione agarose beads (Sigma-Aldrich). The pull-down mixture contained 25 mM Tris-HCl, pH 7.4, 75 mM NaCl, 0.5 mM EDTA, 0.5 mM DTT, 0.05% Nonidet P-40, and 1 mg/mL BSA (NEB). The mixture was rotated at 4°C for 2 h and washed four times with pull-down buffer. The beads were resuspended and boiled in 23 SDS buffer for 5 min, and the eluted material was subjected to SDS-PAGE. GST-tagged peptides were transferred onto a nitrocellulose membrane after separation on SDS-PAGE, and the membrane was blocked with 1% BSA in TBS buffer. Ten micrograms of BiP protein was allowed to bind to the respective peptides by incubation in T-TBS buffer containing 0.1% BSA. This was followed by incubation of the blot with the BiP antibody, which was detected by a secondary antibody, as described in the immunoblotting procedure.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure 1. Sequence Alignment of the Three BiP Proteins from Arabidopsis.
Supplemental Supplemental Figure 11. Expression Levels of BiP in Untreated BiP-Overepressing Lines.
Supplemental Table 1. List of Peptides Used in the Phage Display Experiments.
Supplemental Table 2. Primers Used in This Study.