A New Branch of Endoplasmic Reticulum Stress Signaling and the Osmotic Signal Converge on Plant-specific Asparagine-rich Proteins to Promote Cell Death*

NRPs (N-rich proteins) were identified as targets of a novel adaptive pathway that integrates endoplasmic reticulum (ER) and osmotic stress signals based on coordinate regulation and synergistic up-regulation by tunicamycin and polyethylene glycol treatments. This integrated pathway diverges from the molecular chaperone-inducing branch of the unfolded protein response (UPR) in several ways. While UPR-specific targets were inversely regulated by ER and osmotic stresses, NRPs required both signals for full activation. Furthermore, BiP (binding protein) overexpression in soybean prevented activation of the UPR by ER stress inducers, but did not affect activation of NRPs. We also found that this integrated pathway transduces a PCD signal generated by ER and osmotic stresses that result in the appearance of markers associated with leaf senescence. Overexpression of NRPs in soybean protoplasts induced caspase-3-like activity and promoted extensive DNA fragmentation. Furthermore, transient expression of NRPs in planta caused leaf yellowing, chlorophyll loss, malondialdehyde production, ethylene evolution, and induction of the senescence marker gene CP1. This phenotype was alleviated by the cytokinin zeatin, a potent senescence inhibitor. Collectively, these results indicate that ER stress induces leaf senescence through activation of plant-specific NRPs via a novel branch of the ER stress response.

NRPs (N-rich proteins) were identified as targets of a novel adaptive pathway that integrates endoplasmic reticulum (ER) and osmotic stress signals based on coordinate regulation and synergistic up-regulation by tunicamycin and polyethylene glycol treatments. This integrated pathway diverges from the molecular chaperone-inducing branch of the unfolded protein response (UPR) in several ways. While UPR-specific targets were inversely regulated by ER and osmotic stresses, NRPs required both signals for full activation. Furthermore, BiP (binding protein) overexpression in soybean prevented activation of the UPR by ER stress inducers, but did not affect activation of NRPs. We also found that this integrated pathway transduces a PCD signal generated by ER and osmotic stresses that result in the appearance of markers associated with leaf senescence. Overexpression of NRPs in soybean protoplasts induced caspase-3-like activity and promoted extensive DNA fragmentation. Furthermore, transient expression of NRPs in planta caused leaf yellowing, chlorophyll loss, malondialdehyde production, ethylene evolution, and induction of the senescence marker gene CP1. This phenotype was alleviated by the cytokinin zeatin, a potent senescence inhibitor. Collectively, these results indicate that ER stress induces leaf senescence through activation of plant-specific NRPs via a novel branch of the ER stress response.
The unfolded protein response (UPR) 5 is activated by stress conditions that disrupt the endoplasmic reticulum (ER) home-ostasis and the proper folding and maturation of secretory proteins. This complex signaling cascade is conserved in eukaryotes, although the mechanism of signal transduction differs across species. The central transducer of the UPR is IRE1, a transmembrane kinase that appears to be the only means of signal transduction in yeast, but is part of a tripartite signaling response in animals, which also includes the ER-transmembrane transducers PERK and ATF6 (1). In plants, there is compelling evidence that the UPR operates through IRE1 (2-7), but characterization of this cytoprotective signaling pathway is still incipient. The only known components of plant UPR are the putative proximal sensors that include two Arabidopsis Ire1p homologues and two ATF6-related proteins, designated AtbZIP28 and AtbZIP60 (8 -10). Although the N-terminal domain of plant IRE1 homologues can replace functionally the yeast IRE1, downstream components of the IRE1 signaling are yet to be identified (8). AtbZIP60 and AtbZIP28 have been described as ER stress-induced leucine zipper (bZIP) transcription factor genes that are anchored to the ER membrane under normal conditions and may serve as ER stress sensors and transducers (9,10). Upon sensing the ER stress, AtbZIP28 is proteolytically released from the membrane and translocated to the nucleus by a mechanism that is predicted to be similar to that acting on mammalian ATF6 transducer (10). Expression of a truncated form of either AtbZIP28 or AtbZIP60, harboring the bZIP domain, up-regulates UPR target genes under normal conditions (9,10).
The UPR up-regulated chaperone BiP (binding protein) is central to this cytoprotective response as the indirect sensor of alterations in the ER environment. In addition to this role as molecular chaperone, in mammalian cells, BiP has been shown to regulate the UPR by controlling the activation status of the three transducers, IRE1, PERK, and ATF6 (11). Under normal conditions, the luminal domains of these sensors are occupied with BiP, which is recruited by unfolded proteins upon ER stress. The dissociation of BiP promotes oligomerization and phosphorylation of IRE1 and PERK, as well as relocalization of ATF6 to the Golgi where its transcriptional domain is proteolytically released from the membrane into the nucleus. Although the UPR has not been extensively characterized in plants, conserved aspects between mammalian and plant UPR include its negative regulation by BiP, possibly through interaction with the putative proximal sensors, and the BiP cytoprotective role (12). In fact, we have previously demonstrated that BiP confers tolerance to water deficit in transgenic plants (13).
Constant challenges by adverse conditions and environmental stresses pose major constraints for development, growth, and productivity of plants. Our knowledge about stress-specific adaptive responses and cross-talk between signaling cascades has advanced considerably with genome-wide analyses and expression-profiling studies in different plant species under different stress conditions (14 -16). A common theme that has emerged is that plants transduce environmental signals through integrated networks between different stress-induced adaptive responses. Despite the potential of ER stress response to accommodate adaptive pathways, its integration with environmental-induced responses is poorly understood in plants.
Recently, we performed global expression profiling on soybean leaves exposed to polyethylene glycol (PEG) treatment or to UPR inducers to identify integrated networks between osmotic and ER stress-induced adaptive responses (4). In addition to uncovering specific responses to ER stress and osmotically regulated changes, a small proportion (5.5%) of the up-regulated genes represented a shared response that seemed to integrate the two signaling pathways. These co-regulated genes had similar induction kinetics and a synergistic response to the combination of osmotic-and ER stress-inducing treatments. Thus, they were considered to be downstream targets of the integrating pathway. Two ESTs (N-richI and N-richII), which showed the strongest synergistic induction, were homologous to genes encoding asparagine-rich proteins that have a plant-specific development and cell death (DCD) domain. Here we investigated the possibility that the integrated pathway might transduce a programmed cell death (PCD) signal generated by ER and osmotic stress. We demonstrated that induction of target genes, such as N-richI and N-richII ESTs, by ER stress-inducing agents occurs via a pathway distinct from the UPR. We also found that the N-rich proteins are critical mediators of osmotic-and ER stress-induced cell death in plants.

Plant Growth, Soybean Suspension Cells, and Stress
Treatments-Soybean (Glycine max) seeds (cultivar Conquista) were germinated in soil and grown under greenhouse conditions (avg: 21°C, max: 31°C, min: 15°C) under natural conditions of light, 70% relative humidity, and approximately equal day and night length. Two weeks after germination, the seedlings were transferred to a 2-ml 10 g/ml tunicamycin (Sigma) solution (DMSO, as control). After 12 and 24 h of treatment, the leaves were harvested, immediately frozen in liquid N 2 , and stored at Ϫ80°C until use.
Alternatively, the aerial portions of 3-week-old plants were excised below the cotyledons and directly placed into 15 ml of 10% (w/v) polyethylene glycol (PEG; MW 8000, Sigma), 10 g/ml tunicamycin (Sigma), or 50 mM L-azetidine-2-carboxylic acid (AZC, Sigma) solutions. The first trifoliate leaves were harvested after 4, 10, or 16 h of PEG and tunicamycin treatments, then immediately frozen in liquid N 2 , and stored at Ϫ80°C until use. Each stress treatment and RNA extraction were replicated in three independent experiments.
Cotyledons cells from the soybean variety Conquista were cultured as described previously (17). Cells were subcultured every 10 days by diluting the culture 1:10 in fresh MS medium. All treatments were performed on 5-day-old subcultures. Tunicamycin was added to cultures by dilution of a 5 mg/ml stock in DMSO into normal growth medium to 10 g/ml and incubated for the intervals indicated in the figure legends. For PEG-induced dehydration, the cells were washed and then cultured with normal growth medium containing 10% (w/v) PEG-8000 for the indicated intervals. Suspension cells were directly treated with 0.5 g/ml cycloheximide, 10 M BAP (6-benzylaminopurine), and 10 M zeatin for the intervals indicated in the figure legends. After treatments, the cells were filtrated under vacuum, washed with 0.25 M NaCl, and immediately frozen in liquid N 2 .
Rapid Amplification of 3Ј-cDNA Ends (3Ј-RACE) and NRP-B cDNA Cloning-3Ј-RACE was performed using the GeneRacer kit (Invitrogen). Total RNA isolated from PEG-treated soybean leaves was used for reverse transcription, and an N-richI ESTspecific primer (FwNRich1, TACAGGCATCCAATTTGGC-GAACC) and oligo dT primer from the kit were used in the polymerase chain reaction. The amplified fragment was cloned in pCR4 to generate pCR4-NRPB, which harbors the full-length NRP-B cDNA.
Plasmid Construction-For transient expression in protoplasts, NRP-B and NRP-A (gi:57898928) cDNAs were amplified and inserted into the BglII and EcoRI sites of pMON921 (18). The resulting clones pUFV967 and pUFV849 contain NRP-B or NRP-A cDNAs, respectively, under the control of the cauliflower mosaic virus 35S promoter and the 3Ј-end of the pea E9 rbcS gene. For agroinfiltration in tobacco leaves, NRP-A and NRP-B cDNAs were amplified by PCR and introduced by recombination into the entry vector pCR8/GW/TOPO (Invitrogen) to generate pUFV908 and pUFV874, respectively. NRP-A and NRP-B cDNAs were then transferred by recombination to the plant transformation binary vector 35S-YFP-cas-setteA-Nos-pCAMBIA1300 yielding pUFV937 (35S:NRP-B) and pUFV938 (35S:NRP-A), which contain NRP-B or NRP-A cDNA, respectively, under the control of the 35S promoter and 3Ј-ends of nos. For subcellular localization, NRP-A and NRP-B cDNAs were amplified by PCR with appropriate extensions and introduced by recombination into the entry vector pDONR201 (Invitrogen) and then transferred to the binary vector pK7FWG2 to generate pK7F-NRP-B and pK7F-NRP-A, which contain the GFP gene fused in-frame after the last codon of the respective cDNAs. The U16322 plasmid harboring a full-length NIG (NSP-interacting GTPase, Ref. 19). cDNA was obtained from the Arabidopsis Biological Resource Center (ABRC) and used as a control. For this purpose, the full-length NIG cDNA was amplified by PCR from U16322 cDNA, cloned into pDONR201 and then transferred to the binary vector pK7WG2 to obtain pK7-NIG, which contains NIG cDNA under the control of the CaMV 35S promoter.
Real-time RT-PCR Analysis-For quantitative RT-PCR, total RNA was extracted from frozen leaves or cells with TRIzol (Invitrogen) according to the instructions from the manufacturer. The RNA was treated with 2 units of RNase-free DNase (Promega) and further purified through RNeasy Mini kit (Qiagen) columns. First-strand cDNA was synthesized from 4 g of total RNA using oligo-dT (18) and Transcriptase Reversa M-MLV (Invitrogen), according to the manufacturer's instructions.
Real-time RT-PCR reactions were performed on an ABI7500 instrument (Applied Biosystems), using SYBR Green PCR Master Mix (Applied Biosystems). The amplification reactions were performed as follows: 2 min at 50°C, 10 min at 95°C, and 40 cycles of 94°C for 15 s and 60°C for 1 min. To confirm quality and primer specificity, we verified the size of amplification products after electrophoresis through a 1.5% agarose gel, and analyzed the T m (melting temperature) of amplification products in a dissociation curve, performed by the ABI7500 instrument. The primers used are listed in supplemental Table  S1. For quantitation of gene expression in soybean protoplasts and seedlings, we used RNA helicase (4) as the endogenous control gene for data normalization in real-time RT-PCR analysis. For quantitation of gene expression in tobacco leaves, we used actin as a control gene.
Fold variation, which is based on the comparison of the target gene expression (normalized to the endogenous control) between experimental and control samples, was quantified using the comparative Ct method: 2Ϫ (⌬Ct Treatment Ϫ ⌬Ct Control ). The absolute gene expression was quantified using the 2-⌬CT method, and values were normalized to the endogenous control.
Soybean Transformation-A plant expression cassette containing the BiPD gene was constructed by insertion of the 2.0-kb XbaI cDNA insert of pUFV24 (17) into the pBS35SdAMVNOS2 vector. The resulting clone pBS35SdAMVNOS2-BiP contains the BiP cDNA under control of a duplicated cauliflower mosaic virus 35S promoter with an enhancer region from the alfalfa mosaic virus and the polyadenylation signal of the nos gene. The Arabidopsis thaliana ahas gene (that confers tolerance to the herbicide imazapyr) was removed from the vector pAC321 (20) with XbaI and inserted into the vector pFACM1 to generate pFACMahas. The BiPD expression cassette was released with SalI and NotI from pBS35SdAMVNOS2-BiP and cloned into the vector pFAC-Mahasm to yield pahasBip. The vector pahasBip was used to transform soybean (cv. Conquista) as previously described (20). Primary transformants were selected by PCR using the primers bipsoy235 (5Ј-GAGAGACTAATTGGAGAGG-CTG-3Ј) and bipsoy645c (5Ј-ATAGGCAATGGCAGCAG-CAGTG-3Ј), which amplify a 410-bp sequence from the BiPD gene coding region and cover an intron region from the genomic BiPD sequence. Segregation analyses of independently transformed soybean lines were performed by PCR, and accumulation of BiP was monitored in each subsequent generation by immunoblotting.
RNA Gel Blotting-Total RNA was extracted from frozen leaves of control and tunicamycin-treated wild type and soybean transgenic seedlings, which were treated with tunicamycin or control DMSO for 24 h, with TRIzol (Invitrogen) according to the instructions from the manufacturer. Equal amounts of total RNA were denatured by formamide/formaldehyde and resolved on 1% agarose gels containing formaldehyde. The RNA was transferred to a nylon membrane by capillary transfer and immobilized by UV cross-linking (Stratalinker, Stratagene). The membrane was hybridized at high stringency conditions using the soyBiPD cDNA (17) as probe. The hybridization probe was radiolabeled with [␣-32 P]dCTP by random primed labeling (Amersham Biosciences). Autoradiography was performed at Ϫ80°C using a Lightning-Plus intensifying screen. The amount of RNA and the integrity of ribosomal RNA were monitored by rehybridizing the membranes with an 18 S rDNA probe.
Immunoblot Analysis-Total protein was extracted from control and tunicamycin-treated leaves of wild-type and soybean transgenic seedlings as previously described (17). SDS-PAGE was carried out, and the proteins were transferred from 10% SDS-polyacrylamide gels to nitrocellulose membranes by electroblotting. Immunoblot analyses were performed using polyclonal anti-BiP-carboxyl antibodies (2), an anti-calreticulin serum (21) or an anti-GmSBP2 serum (22) at a 1:1000 dilution and a goat anti-rabbit IgG alkaline phosphatase conjugate (Sigma) at a 1:5000 dilution. Alkaline phosphatase activity was assayed using 5-bromo-4-chloro-3-indolyl phosphate (Sigma) and p-nitro blue tetrazolium (Sigma).
Transient Expression in Soybean Protoplasts-Protoplasts were prepared from soybean suspension cells as essentially described by Fontes et al. (18). The protoplasts were isolated 5 days after subculture by digestion for 3 h, under agitation at 40 rpm, with 0.5% cellulase, 0,5% macerozyme R-10, 0.1% pectolyase Y23, 0.6 M mannitol, 20 mM MES, pH 5.5. The extent of digestion was monitored by examining the cells microscopically at each 30-min interval. After filtration through nylon mesh of 65 m, protoplasts were recovered by centrifugation, resuspended in 2 ml of 0.6 M mannitol, 20 mM MES, pH 5.5, separated by centrifugation in a sucrose gradient (20% (w/v) sucrose, 0.6 M mannitol, 20 mM MES, pH 5.5), and diluted into 2 ml of electroporation buffer (25 mM Hepes-KOH, pH 7.2, 10 mM KCl, 15 mM MgCl 2 , 0.6 M mannitol). Transient expression assays were performed by electroporation (250 V, 250 F) of 10 g of expression cassette DNA, and 30 g of sheared salmon sperm DNA into 2 ϫ 10 5 Ϫ 5 ϫ 10 6 protoplasts in a final volume of 0.8 ml. Protoplasts were diluted into 8 ml of MS medium supplemented with 0.2 mg/ml 2,4-dichlorophenoxyacetic acid and 0.6 M mannitol, pH 5.5. After 36 h of incubation in the dark, the protoplasts were washed with 0.6 M mannitol, 20 mM MES, pH 5.5, and frozen in liquid N 2 until use.
Caspase-3-like Activity and in Situ Labeling of DNA Fragmentation (TUNEL)-Total protein was extracted from soybean cells 36-h post-electroporation. Caspase-3-like activity was determined using the ApoAlert Caspase-3 Colorimetric Assay kit (Clontech) according to the manufacturer's instructions. The substrate was DEVD-pNA, and the inhibitor of caspase-3 activity was the synthetic tetrapeptide DEVD-fmk supplied by the kit. Free 3Ј-OH in the DNA was labeled by the terminal deoxynucleotidyl transferase-mediated dUTP nickend labeling (TUNEL) assay using the ApoAlert DNA Fragmentation Assay kit (Clontech) as instructed by the manufacturer. Samples were observed with a Zeiss LSM 410 inverted confocal laser scanning microscope fitted with the configuration: excitation at 488 nm and emission at 515 nm. As a positive control, samples were treated with DNase1.
Transient Overexpression in Nicotiana tabacum by Agrobacterium Infiltration-N. tabacum plants were grown in a greenhouse with natural day length illumination. Three weeks after germination, the plants were transferred to a growth chamber at 21°C with a 16-hour light and 8-hour dark cycle. Agrobacterium infiltration was performed in the leaves of N. tabacum with Agrobacterium strain GV3101 carrying pUFV937 (35S: NRP-B) or pUFV938 (35S:NRP-A), NIG (35S:NIG), rpl10 (35S: rpL10) as previously described (23).
Subcellular Localization of Proteins-For subcellular localization of proteins, N. tabacun leaves were agroinoculated with pK7F-NRP-B and pK7F-NRP-A. About 72-h post-agroinfiltration, 1-cm 2 leaf explants were excised incubated with 0.6 M mannitol for 10 min, and GFP fluorescence patterns were examined in epidermal cells with ϫ40 oil immersion objective and a Zeiss LSM510 META inverted laser scanning microscope equipped with an argon laser as excitation source. For imaging GFP, the 455-nm excitation line and the 500 -530-nm band pass filter were used. The pinhole was usually set to give a 1-1.5-m optical slice. Post-acquisition image processing was done using the LSM 5 Browser software (Carl-Zeiss) and Adobe Photoshop (Adobe Systems). Microsomal fractions were prepared from agroinoculated leaves, as described (24), and immunoblotted with an anti-GFP serum.
Determination of Chlorophyll Content and Lipid Peroxidation-Total chlorophyll content was determined spectrophotometrically at 663 and 646 nm after quantitative extraction from individual leaves with 80% (v/v) acetone in the presence of ϳ1 mg of Na 2 CO 3 (25). The extent of lipid peroxidation in leaves was estimated by measuring the amount of MDA (malondialdehyde, a decomposition product of the oxidation of polyunsaturated fatty acids). MDA content was determined by the reaction of thiobarbituric acid (TBA) as described by Cakmak and Horst (26).
Ethylene Determination-For ethylene measurements as a function of NRP-A and NRP-B expression, whole leaves of 3-week-old tobacco plants were agroinoculated with pK7-NRP-A, pK7-NRP-B, or control binary vector. Agroinfiltrated tobacco plants were placed in 25-ml flasks containing a Petri dish with 3 ml of 1 M HgClO 4 ⅐4H 2 O and sealed with rubber serum caps. After 5 days, the solution was collected into a sealed assay tube, and ethylene was released with the injection of 1 ml of 1 M KCl. 1 ml of gas was collected with a syringe, and ethylene was measured on a model 5840A gas chromatography system (Hewlett-Packard Canada Ltda) equipped with a flame ionization detector. A Porapak N (80 -100 mesh, 100 cm ϫ 6 mm) column was used at 60°C. The injection port and detector temperatures were 100 and 150°C, respectively. The flow of nitrogen, air, and hydrogen was at 30, 130, and 20 ml min Ϫ1 , respectively.

N-richI and N-richII ESTs Are Both Encoded by the NRP-B
Gene-The N-richI and N-richII ESTs correspond to different regions of a contiguous soybean genomic sequence (scaf-fold_78:2951370.2949024). To determine if both ESTs corresponded to the same mRNAs, we obtained the full sequences of N-richI and N-richII cDNAs through RACE. A comparison of the cDNA sequences revealed that these ESTs represented the same gene which we designated NRP-B (N-rich protein B; supplemental Fig. S1A). The deduced NRP-B protein has an estimated M r of 37092 and pI 7.05 and is most closely related to the previously identified NRP from soybean (27), here designated NRP-A. The two NRPs share a highly conserved C-terminal DCD domain in addition to a high content of asparagine residues at their more divergent N termini. This structural organization places NRP-A and NRP-B in the subgroup I of plantspecific DCD-containing proteins (28). Analysis of several subgroup I DCD domain-containing proteins by ClustalW showed that NRP-A and NRP-B form a subgroup with the Arabidopsis protein of unknown function encoded by At5g42050 (gi 231 18422167, supplemental Fig. S1B). An examination of available microarray data indicated that the At5g42050 locus is induced by cycloheximide, osmotic stress, salt stress, and during senescence (GeneInvestigator).

Integration of ER Stress and Osmotic Signals Leads to Maximal Expression of NRP-A-
The high degree of sequence similarity between NRP-B and NRP-A prompted us to ask whether NRP-A was also a target of the ER and osmotic stress integrating pathway. Expression analysis was carried out by quantitative RT-PCR (Fig. 1). Like NRP-B, NRP-A is up-regulated by the ER stress inducers AZC, dithiothreitol, or tunicamycin that cause protein misfolding in the ER by different mechanisms (supplemental Fig. S2). The L-proline analogue AZC is incorporated into proteins competitively with L-proline, tunicamycin inhibits Asn-linked glycosylation, and dithiothreitol prevents disulfide bond formation. Very likely, induction of NRP-A and NRP-B by these chemicals is a result of accumulation of misfolded proteins in the ER rather than a result of a specific protein dysfunction unrelated to ER stress.
Osmotic and ER stress each promoted rapid induction of NRP-A expression that showed a kinetic pattern similar to that observed for NRP-B induction (Fig. 1A). When combined, AZC and PEG treatments promoted a synergistic effect on accumulation of both NRP-A and NRP-B transcripts (Fig. 1B). Based on these criteria, we concluded that NRP-A, like NRP-B, was a target of the previously described ER and osmotic stress integrating pathway, which we will refer to here as the integrated pathway (4).

Up-regulation of NRP-A and NRP-B by UPR Inducers Occurs through a Novel Signaling Branch of the ER Stress Response-
UPR target genes that encode components of the ER protein processing machinery have been shown to exhibit a differential and antagonistic regulation by ER stress inducers and PEGinduced dehydration (4). The UPR genes, such as BiPD, BiPC, CNX (calnexin), and PDI (protein-disulfide isomerase), are not only down-regulated by osmotic stress but generally show a progressively enhanced effect with an increase in the duration of the PEG treatment (supplemental Fig. S3). In contrast, upregulation of the integrated pathway, represented by the target NRP-A and NRP-B genes, showed a persistent pattern throughout prolonged PEG and tunicamycin treatments. These results suggest that the integrated pathway genes and the UPR-specific genes may be regulated by osmotic stress via independent pathways. In addition, they raise the question as to whether the integrated pathway shares components with the UPR pathway or is a separate, parallel pathway. To address this issue we examined NRP-A and NRP-B induction under conditions that prevent activation of UPR. Because plant UPR has not been fully characterized and therefore loss-of-function mutants are lacking in the plant kingdom, we proceeded to obtain gain-offunction transgenic lines that had a dampened UPR.
In both mammals and plants, BiP overexpression has been shown to inhibit activation of UPR (11,12). We used a similar strategy and obtained soybean transgenic lines ectopically expressing the soyBiPD cDNA under the control of the 35S promoter. These independently transformed overexpression (OE) lines constitutively accumulated higher levels of BiPD mRNA ( Fig. 2A) and protein (Fig. 2B) than untransformed wildtype controls. To examine the capacity of the OE lines to induce a UPR, we treated seedlings from the highest OE line with tunicamycin and assayed for expression of BiP and also calreticulin (CRT), which we used as a UPR response gene. The 35S-BiP-4 transgenic line expressed BiP mRNA at levels similar to the untransformed wild-type line after tunicamycin treatment and displayed only marginal variations in the level of BiP mRNA after ER stress-inducing treatments (Fig. 2C). The OE line showed a slight increase in the levels of BiP mRNA after 12 h of tunicamycin treatment and had slightly less BiP mRNA by 24 h of treatment. We also analyzed protein levels, and we found that the BiP-overexpressing line continued to have slightly higher levels of BiP than those observed in the wild-type line after UPR induction (Fig. 2D). As expected, however, induction of calreticulin was seen only in the WT plants; the BiP-overexpressing line (35S-BiP) lacked the capacity to induce calreticulin protein in response to tunicamycin treatment.
Because the Northern probe does not discriminate between the BiP transgene and endogenous BiP genes, we also analyzed expression with gene-specific primers to quantify only the endogenous soyBiPD mRNA. In the soyBiPD transgene construct, the 3Ј-untranslated sequences are replaced with nos 3Ј sequences so a primer from the native 3Ј would not recognize the transgene (29). The homozygous soybean 35S-BiP-4 transgenic line was treated with tunicamycin and the induction of UPR was evaluated by quantifying transcript accumulation of typical UPR-specific targets (BiP and CNX) by real-time RT-PCR (Fig. 2E). Our results confirmed that overexpression of soyBiPD in soybean transgenic lines attenuated UPR activation as revealed by decreased induction of the endogenous BiP (BiPD) and CNX genes under conditions that promote ER stress in the wild-type controls (Fig. 2E). In contrast, inhibition of UPR activation by BiP did not alter the induction profile of NRP-A and NRP-B genes by tunicamycin. In addition, overexpression of NRP-A and NRP-B in soybean protoplasts did not up-regulate UPR targets, such as BiP and CNX (Fig. 2F). These results strongly indicate that induction of NRP-A and NRP-B genes by ER stress inducers occurs through a different signaling branch of the ER stress response than the ER chaperone-inducing UPR branch.
This integrated pathway diverged further from characterized ER-specific branches of UPR as downstream targets were inversely regulated by osmotic stress in untransformed soybean plants (supplemental Fig. S3 and Ref. 4). In addition, the integrated pathway genes NRP-A and NRP-B exhibited a strong synergistic response to the combination of osmotic and ER stress-inducing treatments (Fig. 1B). Collectively, our results describe a novel branch of the ER stress response that integrates the osmotic signal to potentiate transcription of shared target genes, such as NRP-A and NRP-B.

NRP-A and NRP-B Induce Apoptotic-like Cell Death in Soybean Suspension Cells-NRP-A and NRP-B encode a DCD
domain that has been suggested to be involved in the hypersensitive response and programmed cell death (25,26). If so, we predicted that they would be involved in transducing a PCD signal through the integrated pathway. To examine whether these genes are involved in cell death, we assayed several different aspects of PCD: acute induction of PCD by cycloheximide, activation of caspase-like activities, DNA fragmentation, inhibition of the developmental PCD pathway associated with senescence, and induction of senescence markers by overexpression of NRP-A and NRP-B. Treatment with cycloheximide, a cell death inducer in soybean cultured cells (30), resulted in significant increases in NRP-A and NRP-B transcript accumulation (Fig. 3A). A time course experiment in soybean cultured cells demonstrated that both genes are rapidly induced by cycloheximide as soon as 30 min after treatment (supplemental Fig.  S4). The induction kinetics of NRP-A and NRP-B by cycloheximide was similar to their rapid tunicamycin and PEG induction profile and differed from the delayed tunicamycin-induction kinetics of the UPR-specific target CNX, which was not strongly induced by cycloheximide. This result suggests that induction of NRP-A and NRP-B by the cell death inducer cycloheximide may be connected to the integrated pathway.
To directly assess the involvement of NRP-A and NRP-B in cell death, we monitored the induction of caspase-3-like activity, as a biochemical marker of cells committed to apoptosis. Previous studies in plants have shown that different types of proteolytic enzymes, including caspase-3-like activity, are linked to developmental and pathogen-and stress-induced PCD (31). In addition, caspase-3 has been characterized as a downstream protease in the effector phase of PCD in mammals (32), and recent evidence has demonstrated its possible involvement in plant PCD as well (33). We used transient expression of both NRP-B and NRP-A in soybean protoplasts under the control of the 35S promoter (Fig. 3B) as our experimental system. Expression is shown relative to a helicase marker to control for any variation in transformation efficiency. Analysis of caspase-3-like activity (Fig. 3C), in protein extracts from NRP-A-and NRP-B-overexpressing cells revealed a 2.6-and 2.3fold increase in caspase-3-like activity, respectively, as compared with control, pMON921-transformed protoplasts. Inclusion of a caspase-3-specific inhibitor (DEVD-fmk) reduced enzyme activity to basal levels.
To confirm that the increase in caspase-3-like activity was associated with PCD in the soybean protoplasts, we employed the terminal TUNEL technique to measure fragmentation of DNA in individual cells 36-h postelectroporation. A distinct feature of PCD is the extensive cleavage of nuclear DNA into oligonu-

NRP-mediated Senescence by Osmotic and ER Stress
cleosome-sized fragments, which can be detected by the incorporation of fluorescein 12-dUTP at free 3Ј-OH ends in DNA from the samples. The results of the TUNEL assay are shown in Fig. 3D. Control cells transformed with the empty vector exhibited mostly TUNEL-negative nuclei (panel F) with only a background level of staining, whereas NRP-A-and NRP-B-expressing protoplasts had TUNEL-positive nuclei that showed the same extent of staining as DNase-treated positive controls (compare in Fig. 3D, panels B-H). These results suggest that NRP-A and NRP-B promote apoptotic-like cell death when transiently expressed in protoplasts.
NRP-A and NRP-B Induce Senescence in Planta-To further investigate the role of NRP-A and NRP-B in cell death, we assayed for hallmarks of senescence after manipulating levels of cytokinins. The cytokinins BAP (6-benzylaminopurine) and zeatin have been shown to be potent inhibitors of senescence (34). In a soybean cell cultures, both BAP and zeatin strongly repressed NRP-A expression and slightly down-regulated NRP-B (Fig. 3A). To examine the role of NRP-A and NRP-B in planta, we performed gain-of-function experiments by transiently expressing the NRP genes in tobacco leaves. The corresponding cDNAs were transferred to binary vectors for Agrobacterium-mediated plant transformation, and the transient expression of the respective genes was induced by agroinoculation into tobacco leaves (Fig. 4). After 1 week, the leaf sectors that were agroinfiltrated with the DNA constructs  Fig. S5B and data not shown) displayed a chlorotic phenotype characteristic of leaf senescence that was associated with decreased chlorophyll content and consequent yellowing of the leaf sector (Fig. 4, A  and B). This yellowing phenotype was attenuated by co-agroinfiltration of NRP-A or NRP-B with the senescence inhibitor, zeatin. The chlorophyll content of NRP-A and NRP-B-agroinoculated leaf sectors was significantly lower than that of the NIG-agroinoculated leaf sector control (Fig. 4B). Although inclusion of zeatin did not appear to revert NRP-A-and NRP-B-induced chlorophyll loss as indicated by Student's t test analysis (p Ͼ 0.05), it kept a similar level of chlorophyll in the leaf sectors agroinoculated with control and test genes.

NRP-mediated Senescence by Osmotic and ER Stress
We further examined the NRP-A-and NRP-B-induced senescence phenotype by measuring the accumulation of malondialdehyde (MDA), a major product of lipid peroxidation that, along with generation of active oxygen species, is associated with chlorophyll loss (35). A significant enhanced accumu-lation of reactive aldehydes derived from lipid peroxidation was observed in either NRP-A-or NRP-B-agroinoculated leaf slices in comparison with control samples and zeatin-pretreated NRP-A-or NRP-B-agroinoculated samples (Fig. 4C). Leaf yellowing and enhanced accumulation of MDA were a direct result of NRP-A or NRP-B expression and not a result of agrobacterium interactions, because expression of an unrelated, control protein (NIG, supplemental Fig. S5) did not cause yellowing of the agroinoculated sectors (Fig. 4A), and MDA synthesis (Fig. 4C) in the NIG-agroinoculated leaf sector was similar to that in leaf sectors where senescence was inhibited by zeatin.
The response to zeatin and loss of chlorophyll along with an increase in lipid peroxidation suggest the possibility that NRP-A and NRP-B trigger a response similar to senescence, a form of programmed cell death (36,37). To further examine this possibility, we measure the evolution of ethylene synthesis, a well-known senescence hormone, and induction of known senescence-associated genes. For ethylene measurement, the entire leaves from young tobacco plants were individually agroinoculated with NRP-A-or NRP-B-derived binary vectors and transferred to a sealed container. One week post-agroinfiltration, the ethylene content was measured by gas chromatography (Fig. 4D). Ethylene production was higher in NRP-A-and NRP-B-agroinoculated plants than in control plants agroinfiltrated with the unrelated gene NIG. Inclusion of the senescence inhibitor zeatin in the agroinoculum prevented NRP-A-or NRP-B-induced ethylene evolution in agroinfiltrated leaves.
In addition to the physiological responses related to senescence, we assayed expression of the senescence-associated gene marker, CP1 (cysteine protease 1), by quantitative RT-PCR (Fig.  4E). NRP-A and NRP-B induced the expression of CP1 when transiently expressed in tobacco leaves. This effect was antagonized by pretreatment of the leaves with the senescence inhibitor zeatin. Collectively, these results substantiate the argument that both NR proteins are functionally involved in programmed cell death processes.

Subcellular Localization of NRP-A and NRP-B Fused to GFP-
The deduced NRP-A and NRP-B proteins do not contain any predicted signal of subcellular localization, although NRP-A was detected in a cell wall fraction of soybean cells undergoing a hypersensitive response (27). We investigated the localization of NRP-A and NRP-B, to see if it might provide additional clues to the function of these proteins in the integrated pathway. We utilized an Agrobacterium-mediated transient expression assay in epidermal cells of tobacco leaves and imaged NRP-A and NRP-B fused to GFP by confocal microscopy (Fig. 5). Both NRP-A and NRP-B fusions were targetted to the cell surface (Fig. 5A). We induced plasmolysis by treatment of the leaf sectors with 0.6 M mannitol to promote detachment of the plasma membrane from the cell wall. Both NRP-A-GFP and NRP-B-GFP remained in the area lining the detached plasma membrane leaving a fluorescent-free cell wall (Fig. 5B). These results rule out a cell wall location for the GFP-fused proteins when transiently expressed under control of the 35S promoter. However, because of the large central vacuole of leaf epidermal cells, the cytoplasm is pressed against the plasma membrane and hence appears in confocal slices as a narrow band. Therefore, it is not clear whether the observed fluorescence is derived from the plasma membrane, tonoplast, cytosol, or other organelles. Because the fused proteins were detected in microsomal fractions prepared from tobacco leaves transiently expressing NRP-A-GFP or NRP-B-GFP (data not shown), NRPs may be associated with membranes.
We also monitored the functional performance of GFP-fused proteins in planta as means to certify that fusion of GFP to the NRP-A and NRP-B C termini would not have caused mislocalization of the protein.  S6) and confirmed that the GFP-fused proteins were functional. These results suggest that the localization of the fused proteins may reflect that of the intact proteins as well.

DISCUSSION
Previously we performed a global expression profiling on soybean leaves exposed to PEG treatment (osmotic stress) or to ER stress inducers and identified integrated networks between osmotic and ER stress-induced adaptive responses (4). A combination of the ER stress-and osmotic stress-induced treatments promoted a synergistic effect on the induction level of common up-regulated genes, indicating that the ER stress response integrates the osmotic signal to potentiate transcription of shared target genes. Among them, N-richI and N-richII ESTs showed the strongest synergistic induction by ER and osmotic stress. Upon further investigation, we found that the N-richI and N-richII ESTs represent the same gene, named NRP-B, which shares a high degree of identity with a previously identified N-rich protein (NRP) gene that we designate NRP-A. Like NRP-B, NRP-A may be considered to be a target of the integrated pathway, as it is synergistically and rapidly up-regulated by simultaneous induction of ER and osmotic stress. Three lines of evidence indicated that NRP-A and NRP-B mediate an apoptosis-like response in plants presumably through their DCD domains. Firstly, both genes were induced by cycloheximide, an apoptosis inducer, but repressed by the senescence inhibitors BAP and zeatin. Secondly, expression of NRP-A and NRP-B in soybean protoplasts induced caspase-3like activity and DNA fragmentation. Finally, transient expression of NRP-A and NRP-B in planta accelerated senescence, as determined by the rapid development of leaf yellowing, that was accompanied by decreased chlorophyll content, increased MDA production, increased ethylene evolution, and induction of the senescence-associated gene marker CP1. The senescence inhibitor zeatin markedly prevented the development of NRP-A-and NRP-B-induced symptoms. Supplementation of the NRP-A or NRP-B agroinoculum with zeatin reduced leaf yellowing, chlorophyll loss, and MDA synthesis as well as prevented ethylene evolution and induction of senescence-associated gene markers. These results further substantiated the argument that NRP-A and NRP-B promote programmed cell death and leaf senescence.
Our results clearly indicated that ER and osmotic stress converge upon NRP-A and NRP-B to potentiate a senescence-like response. The induction of leaf senescence by drought stress, which shares signaling components with osmotic stress, is a well-documented phenomenon in plants (38). However, a direct link between prolonged ER stress and leaf senescence has not been previously demonstrated, although recent studies have shown that, like in mammalian cells, ER stress induces cell death and apoptotic-like responses in plants as well (33, 39 -42). The recent work by Watanabe and Lam (42) further demonstrated that inactivation of the Arabidopsis BAX (BCL2associated X protein) inhibitor 1 (AtBI1) accelerated the ER stress-induced PCD progression in roots. They proposed that AtBI1 modulates a PCD execution program upon ER stress signaling. Although the mechanisms by which ER stress triggers specific programs of cell death in plants remains unknown, indirect evidence indicates that ER stress-induced PCD may occur through the molecular chaperone-inducing branch of UPR as well as by a distinct signal transduction cascade (43). This hypothesis is based on studies with ER stress inducers that identified AtBI1 and the PCD-related Hsr203J, as UPR genes (5,44). The presence of an ERSE-like cis-acting regulatory element in the AtBI1 promoter suggested that up-regulation of the BI gene by ER stress may occur through the molecular chaperoneinducing branch of plant UPR (5). This hypothesis was strengthened by the recent demonstration that accumulation of unfolded protein in the ER controls the AtBI1 promoter activity (42). In contrast, the delayed kinetics of Hsr203J induc-tion in comparison with that of the UPR-specific target BiP and PDI genes indicated that ER stress may induce the PCD-related Hsr203J gene through a pathway independent from UPR (44). Here we showed that ER stress induces leaf senescence by activating plant-specific NRPs through a pathway distinct from the ER stress-specific branches of UPR. This interpretation is supported by several lines of evidence. First, activation of the NRPmediated senescence-like response is not specific to ER stress but rather is a shared branch of general environmental adaptive pathways. In fact, NRPs are also induced by other abiotic and biotic signals, such as osmotic stress, drought, and pathogen incompatible interactions (27). Secondly, the molecular chaperone-inducing branch of UPR is inversely regulated by osmotic stress, whereas full induction of NRPs is dependent on both ER and osmotic stress signals (Fig. 1B). Finally, while induction of UPR-specific target genes, such as BiP and CNX, is prevented in soybean transgenic lines defective for UPR activation, NRP-A and NRP-B expression is up-regulated to a similar extent as in wild-type plants.
Recently, the subunit G␤ of the heterotrimeric G protein has been demonstrated to associate partially with ER membranes and to be involved in the signaling events that trigger UPRassociated cell death in plants (41). Inactivation of the G␤ gene in Arabidopsis impairs tunicamycin-induced cell death and attenuates the ER stress-induced transcriptional response of UPR-specific target genes, such as BiP and PDI, in G␤-null alleles. Because G␤ signaling seems to be coupled (at least partially) to the molecular chaperone-inducing branch of UPR, it is unlikely that induction of NRPs by ER stress may be connected to the cell death-inducing G␤ signaling. More likely, ER stressinduced leaf senescence through activation of NRP expression describes a novel branch of the ER stress response in plants that shares components with the osmotic stress signaling.
In mammalian cells, ER stress has been shown to trigger both ER stress-specific apoptotic pathways and shared PCD signaling pathways elicited by other death stimuli (11). The NRPmediated PCD signaling fits the concept of a shared pathway induced by ER stress and other stimuli, which is consistent with the localization of NRPs either as cytosolic or membrane-associated proteins. The lack of canonical transmembrane domains in the NRP-deduced primary structure suggests that any possible interaction of NRPs with plasma membrane would be either mediated by post-translational modification or by association with transmembrane proteins. Several cell surface death receptors have been identified in mammalian cells as the mediators of the cell death-inducing extrinsic pathway (45). Upon activation by ligand binding, the cytoplasmic domain of these death receptors recruits adaptor proteins to activate caspase cascades that result in morphological changes associated with apoptosis (32). The cytosolic or membrane-associated localization of NRPs, their rapid response to programmed cell death inducers along with their capacity to induce caspase-3-like activities in soybean protoplasts are consistent with a model in which NRPs would act as adaptor proteins that activate upstream initiator caspases. The identification of downstream targets and induction regulators of NRPs is crucial to elucidate the underlying mechanism of this ER-and osmotic stress-induced cell death signaling that transduces a PCD signal into multiple hallmarks of leaf senescence.