Protein quality control (QC) within the endoplasmic reticulum and the related unfolded protein response (UPR) pathway of signal transduction are major regulators of the secretory pathway, which is involved in virtually any aspect of development and reproduction. The study of plant-specific processes such as pathogen response, seed development and the synthesis of seed storage proteins and of particular toxins is providing novel insights, with potential implications for the general recognition events and mechanisms of action of QC and UPR.
Thousands of proteins, collectively termed secretory proteins, start their life in the endoplasmic reticulum (ER), to be eventually secreted or distributed into the membranes or lumen of the different compartments of the endomembrane system. The ER is a nursery where newly synthesized secretory polypeptides fold and oligomeric proteins assemble before being sorted to their final destination. Most ER-resident proteins are folding and assembly helpers that also monitor the success of these events, keeping the newly synthesized proteins in the ER until the correct three-dimensional conformations have been reached and sorting for disposal those that fail to fold or assemble properly (1). These events, termed quality control (QC), are functionally linked to another pathway, termed the unfolded protein response (UPR), which regulates gene transcription and protein translation to adapt the secretory pathway to ER loading and to the overall efficiency of the folding process within this compartment (2). Most strikingly, an increase of misfolded protein in the ER is sensed as an ‘ER stress’ and induces increased synthesis of ER helpers.
Plant biology and plant-specific processes are adding to our understanding of QC and the UPR (see 3–5 for earlier reviews). We will discuss in this study some issues that have been the objects of recent stimulating developments.
What is a stressed ER? Perhaps, the first description at the molecular level regarded mammalian cells infected with a tumor-inducing virus. The resulting proliferation rapidly depleted the medium of glucose, which in turn inhibited N-glycosylation of proteins and induced the accumulation of two proteins much later identified as the ER-resident chaperones of the Hsp70 and Hsp90 family, BiP and endoplasmin, respectively (6). The subsequent discovery that BiP binds structurally defective proteins much more extensively than their normal counterparts opened the way to the now widely accepted conclusion that the synthesis of structurally defective polypeptides is sensed as an ER stress signal. The ER can therefore be considered stressed when it accumulates defective proteins that have difficulty in folding (2). The subsequent increases in the synthesis of molecular chaperones and protein folding catalysts could simplistically be viewed as the molecular landmarks for ER stress. However, other deviations from ER homeostasis can promote similar responses. For example, an increase occurs during (normal) developmental programs that involve enhanced synthesis of secretory proteins such as the development of storage tissues during seed maturation in plants and the differentiation of B lymphocytes into plasma cells in mammals (7–9). During these programs, the ER and other compartments of the endomembrane system increase in dimension. The obvious explanation that ER dilation is triggered by enhanced loading of secretory protein that would require more ‘room’ has been challenged by the observation that the induction of molecular chaperones and protein folding catalysts actually precedes that of seed storage proteins (7), secretory hydrolases produced during plant pathogen responses (8) and immunoglobulins in plasma cells (9) as though the ER somehow senses that it is going to be stressed (Figure 1). This ‘anticipation’ has not yet been fully characterized but may reflect an involvement of ER signaling in integrated networks of adaptive responses. Such an idea is supported by the observation that the genes for many plant ER molecular chaperones have a common upstream regulatory element activated by non-expressor of PR1 (NPR1), a master regulatory protein for systemic acquired resistance (10).
Signaling ER Stress
Perception of ER stress leads to induction of a conserved, cytoprotective UPR that is mediated solely through the transmembrane kinase, inositol-requiring enzyme-1 (IRE1), in yeast but in animals is part of a tripartite signaling response. The ER luminal domains of IRE1 and the other sensors found in metazoans, activating transcription factor 6 (ATF6) and protein kinase-like endoplasmic reticulum kinase (PERK), interact with BiP. When unfolded proteins begin to accumulate within the ER, the sensors release the chaperone, form homodimers (and possibly oligomers) and initiate phosphorylation signaling cascades (2). As a result, PERK activity attenuates general mRNA translation, while all three sensors enhance the transcription of UPR target genes, including those of ER folding helpers and proteins of the machinery that helps with the disposal of misfolded proteins. Upon prolonged ER stress, PERK and ATF6 eventually activate apoptosis.
The luminal domain of plant IRE1 can functionally replace the corresponding domain of yeast Ire1 (11), but its direct downstream targets are not yet known. Two other putative sensors of ER stress in plants are implicated by similarity to animal sensors and transcriptional activation activities, but our understanding of processing pathways and activities in vivo is still incomplete. One of the putative sensors from Arabidopsis thaliana (At), AtbZIP60, encodes an ER-stress-induced leucine zipper (bZIP) transcription factor that is proposed to be a plant counterpart of mammalian ATF6 (12). ATF6, like IRE1, has a COOH-terminal sensor domain in the ER lumen. In response to ER stress, ATF6 traffics to the Golgi complex where it is proteolytically cleaved to release an NH2-terminal domain capable of moving to the nucleus and functioning as a transcription factor (2). In Arabidopsis, AtbZIP60 is normally membrane anchored; when expressed without a membrane anchor, it can move into the nucleus and activate promoters of genes induced as part of the UPR (12). In contrast to ATF6, however, AtbZIP60 lacks domains in the ER luminal portion that have been implicated in BiP interactions and proteolytic cleavage for ATF6 (12). These differences would seem to point to a role in ER stress signal transduction that differs at least mechanistically from that of ATF6. A second membrane-associated transcription factor, AtbZIP28, is also induced by the N-glycosylation inhibitor tunicamycin and by the reducing agent DTT, which inhibits the formation of disulfide bonds (13). Unlike AtbZIP60, AtbZIP28 has a predicted site-1 protease cleavage site and is processed by Arabidopsis S1 protease. When expressed without its membrane anchor, AtbZIP28 induced expression of RNAs corresponding to several molecular chaperones, making it a second likely candidate for activating the UPR (13). The differences in structural domains between AtbZIP28 and AtbZIP60 open the possibility that they have different sensing and/or signal transduction properties. Whether these will also reflect differences in targets remains to be determined.
In addition to primary signal transduction of an ER stress, plants also appear to have signaling pathways that lead to UPR-mediated apoptosis. Arabidopsis has a single heterotrimeric G protein that is an early mediator of oxidative stress and hormone-mediated stress responses. A null mutant for the βG subunit, but not a similar mutant of the αG subunit, is more resistant to tunicamycin and has an alleviated UPR response, with attenuated induction of BiP and apoptosis (14). This G subunit could also be a component of the plant ‘anticipated’ UPR response. Other links between ER stress and apoptosis are suggested by microarray data, response to tunicamycin treatment and synergistic responses to ER and osmotic stresses, but the pathways of signal transduction have yet to be resolved (5,15).
It should also be noted that not all ER folding helpers are equally induced by a given stress. BiP, calreticulin and protein disulfide isomerase transcripts are differently affected by various stress stimuli in tobacco (16). The synthesis of endoplasmin is induced by tunicamycin in tobacco and Arabidopsis seedlings but, unlike that of BiP, not in tobacco protoplasts (17). The lack of fully co-ordinated regulation points to the existence of yet unknown regulatory mechanisms specific for the individual proteins besides the general master switches. Perhaps, such diversity in responses is not surprising given the differences defined to date in ER stress elements in promoters of UPR-responsive plant genes (5). The differences could in part be related to the role of the specific helpers that like BiP can be very promiscuous or like calreticulin and calnexin be specific for glycoproteins or like endoplasmin be limited to a restricted number of secretory proteins (16–18). However, there may be other explanations because even among the three BiP genes of Arabidopsis, there is marked variability in response to tunicamycin (19).
Adaptive Responses to ER Stress
ER stress can be rapidly induced by pharmacological agents that perturb protein folding or alter posttranslational modifications. As described above, unusually severe or prolonged ER stress has been linked to UPR-mediated apoptosis, so drug-induced ER stress is typically assayed soon after treatment (5). In contrast, when ER stress is induced by accumulation of a mutant protein, long-term responses can be measured. A model system of long-term adaptation to ER stress is found in the maize endosperm mutants, floury-2 (fl2), Mucronate (Mc) and Defective endosperm B30 (De-*B30), which have primary defects in genes encoding zeins, the major storage proteins of the seed (20–22). These mutants exhibit an endosperm-specific UPR for several weeks during seed maturation.
To identify genes involved in the transcriptional response to ER stress, several groups analyzed changes in gene expression with RNA profiling techniques. To date, data are available from Arabidopsis seedlings treated with the proline analog, azetidine-2-carboxylic acid (AZC), DTT and tunicamycin; soybean leaves treated with AZC and tunicamycin and the fl2 and Mc maize mutants (23–26). While the drugs can serve as acute inducers of ER stress, the maize mutants offer a complementary chronic induction. As a result, comparison across different experiments provides a means for determining common induction patterns. In addition, analysis of endosperm gene expression in normal maize, in mutants that do not show the UPR, and in mutants that do, might allow discrimination between cellular accommodation of secretory protein flux and ER stress leading to the UPR.
Comparison across the drug studies showed a common increase in expression of genes associated with protein folding, glycosylation, translation/translocation and protein degradation (5,24–26). When expression data from the fl2 and Mc maize mutants are considered, protein folding and degradation categories emerge (23). These data likely represent a minimal overlap as there were fewer gene targets represented on the soybean and maize platforms compared with targets on the Arabidopsis platforms. Within the maize data sets, comparison of gene expression in mutant and normal endosperm did not produce any patterns that would resolve the issue of whether a high secretory flux in seeds mimics a low level UPR. Perhaps, a clearer picture of the plant response to ER stress will emerge when expression can be analyzed across whole genome arrays or more sensitive open-ended platforms, but for now, available data suggest some conservation across kingdoms in the types of genes whose expression is altered during ER stress.
Disposal of Defective Secretory Proteins: the ERAD Pathway
The disposal of defective secretory proteins by ER QC involves recognition, removal from the normal trafficking pathway and their targeting for degradation. Defective secretory proteins are often found in association with BiP for much longer times after synthesis compared with their wild-type counterparts, indicating an inability to mask BiP binding sites (reviewed in 3 for plants). Among the shortest protein domains known to promote BiP interaction in vivo are the helical domains of phaseolin, which are also involved in the trimerization of this vacuolar storage protein (27): trimer formation thus masks these binding sites and allows phaseolin anterograde trafficking from ER, whereas phaseolin mutants unable to trimerize have prolonged interactions with BiP before degradation by QC. Interactions with BiP and other folding factors of the ER most probably inhibit the trafficking of defective proteins along the secretory pathway by mechanisms involving either direct retention or retrieval from the Golgi complex (Figure 2). The two mechanisms appear to be substrate specific, dispelling the earlier notion that defective membrane proteins are likely to be retained, while soluble ones are retrieved (28,29). Interestingly, a recent analysis of glycan processing indicates that, when made in plant cells as an orphan polypeptide in the absence of its B chain, castor bean ricin toxin A chain (RTA) cycles through the Golgi complex before its degradation, whereas the almost identical Ricinus communis agglutinin A chain (RCA A) does not (30).
After prolonged interactions with the folding machinery, the best described eventual fate of defective secretory proteins is their dislocation from the ER to the cytosol, followed by ubiquitination and final degradation by the proteasome. The process is called ER-associated degradation (ERAD) (1) (Figure 2). Three ERAD substrates have been identified in plant cells to date: the above-mentioned RTA and RCA A and point-mutated forms of barley mildew resistance o protein (30–32). The A and B chains of ricin are synthesized as a single precursor polypeptide, proricin, which then matures in seed storage vacuoles. The related RCA undergoes similar processing events. If the ricin or RCA precursor is engineered by deleting the sequence encoding the B chain, the orphan A chain is recognized as a defective protein, dislocated into the cytosol, deglycosylated and in most part degraded by the proteasome (30,31). RTA is a peculiar ERAD substrate, a feature related to the mechanism used by this potent toxin to reach its targets. When ricin comes in contact with mammalian cells, it undergoes endocytosis and retrograde transport using the normal membrane recycling mechanisms of the endomembrane system until it reaches the ER. In this compartment, the reduced RTA chain appears to be recognized as an ERAD substrate for its retrotranslocation, with a small fraction subsequently uncoupling from this pathway to refold in the cytosol into an active enzyme that very efficiently inactivates ribosomes (reviewed in 33). The whole process is quite inefficient, but the toxin is extremely potent, and a few active molecules in the cytosol are sufficient for a cytotoxic effect. Indeed, even in plant cells that contain relatively recalcitrant ribosomes, the toxin is able to inhibit protein synthesis (33). Therefore, RTA dislocation from the ER to the cytosol is at least in part uncoupled from its degradation. Although such an escape from ERAD is unusual, it is not unique [(34) and references therein].
The cells that naturally produce the ricin toxin protect themselves from its action by the concerted action of several mechanisms. First, proricin is inactive: processing in storage vacuoles removes an internal propeptide producing active RTA and the lectin B chain of ricin toxin (RTB). The latter remains associated to RTA through a disulfide bond and favors uptake of the toxin by target cells. Second, an N-terminal propeptide that also negatively affects toxicity is also removed in storage vacuoles (35). Third, RTA retains two Lys residues, which are substrates for polyubiquitination: mutagenesis of these residues enhances toxicity in plant protoplasts (36). Together with the N-terminal propeptide, the potential for ubiquitination would guarantee rapid clearance of any nascent RTA arising from mistargeting, premature termination of preproricin mRNA translation or improper proteolytic events in castor bean seeds that would lead to intoxication (36).
Information is emerging on ERAD machinery in plants. Maize homologues (ZmDerlin1-1 and ZmDerlin1-2) of a yeast ERAD gene Der1 suppressed the mutant phenotype of a yeast strain lacking Der1 and Ire1 (37). Efficient degradation of plant ERAD substrates requires the activity of the cytosolic AAA-adenosine triphosphatase (ATPase) CDC48/p97 (30,32). It should be noted that CDC48/p97 is also involved in the disposal of a mutated RTA that is completely devoid of Lys residues and thus of canonical polyubiquitination sites, challenging the view that the ATPase acts only on ubiquitinated substrates (30). Interestingly, mutagenesis of the same residues in RTA that increase toxicity in plant protoplasts does not enhance ricin toxicity in mammalian cells (36). Putative homologues encoding other ERAD-associated proteins have been identified through sequence similarity analyses, but these do not always contain conserved residues predicted to be important for function (5). Thus, it appears that there are unknown differences in the ERAD process in plants and animals.
Not all defective secretory proteins are degraded by ERAD. Alternative pathways involve degradation by vacuoles (or animal lysosomes). The sorting events are not clear, and at least two pathways are thought to exist. Detailed analysis using yeast mutants indicates that in some cases, the defective proteins undergo vesicular traffic through the Golgi complex and endosome/prevacuolar compartment and that the Vps10 receptor that mediates the normal sorting of most vacuolar proteases is involved [(38) and references therein]. In other cases, QC vacuolar delivery is mediated by autophagy, possibly through selective recognition and sequestration into autophagosomes of ER fragments that contain the misfolded proteins (reviewed in 39). An individual defective protein is not necessarily disposed of by a single pathway: for example, the defective Z variant of α1-antitripsin exists in a soluble form that follows the Golgi-mediated, Vps10-dependent degradation route and an aggregated form that is eliminated by autophagy (38). Therefore, the degree of misfolding or solubility can dictate the degradation pathway.
It has been suggested that the Golgi-mediated QC pathway could involve escape from ER retention/recycling because of the lack of BiP binding sites or other features, followed by recognition by a not yet clearly defined post-ER QC (40). An alternative model has been proposed based on the trafficking properties of BiP in plant cells. Evidence for the delivery of BiP to vacuoles was obtained by taking advantage of the effects of the drug wortmannin (18). Wortmannin inhibits receptor-mediated traffic to vacuoles by blocking the recycling of the plant vacuolar sorting receptor BP80 from the prevacuolar compartment to the Golgi complex, where vacuolar proteins are sorted from proteins destined for secretion. Secretion of BiP from tobacco protoplasts is practically undetectable in normal conditions, an expected feature for an ER resident, but it is greatly stimulated by treatment with wortmannin. This secretion is dependent on coat protein II (COPII) coat formation, consistent with an involvement of the Golgi complex. When BiP is overproduced under a constitutive promoter, it can be detected in vacuoles by immunoelectron microscopy (18). Therefore, a proportion of BiP is delivered to vacuoles through the route and mechanism used by vacuolar resident proteins and is probably degraded there. While this may be the normal route of BiP turnover, the authors also observed that a number of unidentified proteins are found in ATP-sensitive association with BiP in the incubation medium of protoplasts treated with wortmannin and tunicamycin. This led to the suggestion that vacuolar delivery is part of the QC function of BiP. According to this model (18): (i) misfolded proteins that are not available, or poorly available, for retrotranslocation to the cytosol would undergo multiple cycles of ER–Golgi traffic in association with BiP; (ii) the chaperone would have, in addition to the HDEL signal, a vacuolar sorting signal that can be recognized by the BP80 receptor and (iii) at each visit to the Golgi stacks, the BiP–ligand complex would be recognized either by the HDEL receptor and recycled to the ER or by BP80 and disposed in the vacuole (Figure 2). If correct, the model would give a simple explanation for the recognition events that lead to Golgi-mediated sorting of misfolded proteins.
It should be noticed that BiP and other ER folding helpers are also delivered to the plant vacuole and degraded therein, in relevant amounts, when cultures of plant cells reach stationary phase (41). Analysis of the glycans of protein disulfide isomerase delivered to the vacuole indicated that the Golgi complex is not involved in this process (41). Thus, the pathway appears to be distinct from the above-described vacuolar sorting of plant BiP and could be because of autophagy. It is not yet known whether this plant route is selective and could thus represent a form of QC or instead constitutes a normal ER turnover mechanism accelerated in certain conditions.
QC and Insoluble Polymers in the ER
The storage proteins of the prolamin class (such as the zeins discussed earlier) are typical of the cereal endosperm of important crops such as maize and wheat. Prolamins do not have the classical HDEL-like ER localization signals but accumulate in the ER as large insoluble polymers in ER-derived protein bodies (PBs). Within a given plant species, different prolamins exist and co-operate in the formation of heteropolymers. However, some prolamins polymerize and lead to the formation of PB also when expressed individually in vegetative tissues of transgenic plants (42,43), and even individual domains can promote PB formation in chimeric fusions (44,45). Studies of BiP interactions revealed that the chaperone is found in ATP-sensitive association with natural prolamins or chimeric proteins that form PB after they have polymerized and have become insoluble (44,46) (Figure 2). This association is unlike that of the soluble oligomeric storage proteins destined to storage vacuoles, which are released from BiP upon oligomerization. Prolamins seem therefore to take advantage of the chaperone system for their accumulation, but they avoid the degradation process often encountered by defective proteins, which also interact for a prolonged time with chaperones [discussed in Vitale and Ceriotti (4)]. This has some implications for our view of the recognition mechanisms of QC, as we discuss below.
One of the key questions that has not yet been answered is ‘which proteins are targets of BiP during PB formation?’ Although some prolamins are sufficient to nucleate PB formation (e.g. zeins in the γ- and β class), others (e.g. 22 kDa and 19 kDa α-zeins) accumulate only in the presence of the nucleation zeins, and these form an ordered arrangement within the PB (47,48). A second question impacting on QC is whether BiP is important for converting cisternal ER to spherical PB. At present, the answer is not known. In the floury-1 (fl1) mutant (which lacks an ER and PB membrane protein), 22 kDa α-zeins are dispersed within the PB, whereas in wild-type PB, they form a discrete ring between an outer ring of γ-zeins and the PB core (49). Surprisingly, this perturbation neither induces the UPR nor alters size or spherical morphology of PB. In contrast, fl2 and De*-B30 retain 22 kDa α-zeins at the ER/PB membrane where they are anchored by uncleaved signal sequences (20,21). These mutations cause a dramatic increase in the accumulation of molecular chaperones in PB and association of BiP with the defective prolamins. They also negatively affect storage protein accumulation and result in highly lobed PBs that appear to lack the capacity to separate into individual vesicles (20,21,37,50). In these mutants, BiP forms a stable ATP-dependent association with the defective zeins. This spare UPR inducibility indicates that, in spite of one or more PB proteins having affinity for BiP, normal prolamins do not cause severe ER stress.
In endosperm tissue where many different prolamins are synthesized, the heterotypic interactions may markedly affect the destiny of defective prolamins (20–22,50). UPR induction, though, clearly does not lead to the rapid degradation of the fl2 or De*-B30 mutant zeins, which accumulate to high levels. However, it may promote degradation of the 16 kDa γ-zein, which is dramatically decreased in the Mc mutant (22). The mutation consists of a frameshift that alters a large part of the C-terminal domain. In γ-zeins, this domain is very similar to the vacuolar storage proteins of the 2S albumin class, with conserved intrachain disulfide bonds. The frameshift in the Mc mutant most probably leads to severe misfolding of this domain, as also indicated by the altered solubility properties. Possibly, the membrane-anchored fl2 or De*-B30 mutant zeins accumulate to higher levels than the mutant Mc zein because of their tight membrane association and/or still extensive association with the PB. The abundance and insolubility of zeins complicate analysis in maize. However, the results that we will discuss below, obtained by expressing individual prolamins or chimeric proteins, circumvent this problem and help in defining the minimal requirements for polymerization and PB formation and their relationships with QC.
Interchain disulfide bonds are formed by many prolamins and are responsible for their polymerization: Cys residues present in certain domains were found to be the determinants for PB formation and ER retention in studies on chimeric proteins (45,51). The 2S albumin-like domain of γ-zeins is preceded by domains that are rich in Pro residues and also contain additional Cys residues. The Pro-rich domain of 27 kDa γ-zein induces PB formation when fused to phaseolin in the chimeric protein zeolin (44). When disulfide bond formation is inhibited by treatments with reducing agent in vivo or the Cys codons are mutated to Ser codons, zeolin remains soluble, traffics along the secretory pathway and is in relevant amount secreted (51). Consistently, a 27 kDa γ-zein deletion mutant, composed solely of the 2S albumin-like domain that has only intrachain disulfide bonds (52), is also secreted (42). These observations indicate that, per se, the absence of disulfide-bond-mediated polymerization does not make γ-zein domains substrates for ER retention and QC degradation. However, the cytosolic HIV protein Nef or a structurally defective mutated form of phaseolin, both degraded by ER QC in plant cells, are not rescued from their destiny when fused to the same zein fragment used to construct zeolin (53). Nef has three Cys residues that could perturb the formation of correct interchain zein bonds, but the defective phaseolin cannot directly interfere because it is devoid of Cys residues. On the whole, these results suggest a model in which QC is functionally placed before and independently of the sorting between polymerization/insolubilization on one side and traffic along the secretory pathway on the other (Figure 2).
On the basis of the results discussed in this review, Figure 2 illustrates a working model for QC-based decision making within the ER. The model would imply that QC is able to sort out very early after synthesis proteins recognized as permanently defective and that PB are dynamic structures. Whether or not loosely bound defective polypeptides undergo association/dissociation cycles from PB before QC degradation remains an open question.
The ER is an incredibly versatile organelle that must control protein secretory processes for normal cell function and in response to perturbation of ER homeostasis. Our understanding of the cellular and molecular mechanisms by which the plant ER senses and responds to stress is rapidly increasing as we develop the tools and systems needed for analysis. Differences have been found in signal transduction molecules for ER stress among plants, animals and yeast, yet the basic features of the response are conserved. Likewise, recognition of ERAD substrates is not completely conserved across kingdoms. These findings raise intriguing possibilities for discovering the functional basis of the evolutionary differences. As we gain a better understanding of the QC machinery in the plant ER, we will not only learn how the ER orchestrates the secretory, protein folding, protein accumulation and degradation pathways but also have the basis for approaching questions of comparative biology. Separation of the QC process into its component parts has been greatly facilitated by use of reporter constructs and systems that are easily assayed and manipulated. Availability of maize mutants induced for ER stress has allowed analysis of native protein interactions and localization within PB. In the past two decades, we have moved from thinking of ER stress as initiating a linear signaling pathway involving association of proteins with molecular chaperones to being part of a dynamic network involving multiple destinations. We have begun to identify proteins with important functions within the major branches of the ER stress response. Now, we are faced with the challenge of defining the mechanisms that act within the ER QC to direct a given protein to a particular fate. The next era of secretory pathway QC promises to be an even more enlightening time for plant biology as we apply what we have learned to predictive modeling and rational design.
We thank Lynne M. Roberts and Lorenzo Frigerio for giving us access to work in press and Aldo Ceritti for the stimulating discussions. R. S. B. acknowledges support from Department of Energy grant number DE-FG02-00ER150065. A. V. acknowledges support from the European Union Integrated Project ‘Pharma-Planta’ (LSHBCT-2003-503565) and Research Training Networks Contract HPRNCT-2002-00262 (BioInteractions) and the Italian MIUR-FIRB Project (RBNE01TYZF).