ER quality control can lead to retrograde transport from the ER lumen to the cytosol and the nucleoplasm in plants


For correspondence (fax +44 1865 483955; e-mail


Quality control in the secretory pathway is a fundamental step in preventing deleterious effects that may arise by the release of malfolded proteins into the cell or apoplast. Our aims were to visualise and analyse the disposal route followed by aberrant proteins within a plant cell in vivo using fluorescent protein technology. A green fluorescent protein (GFP) fusion was detected in the cytosol and the nucleoplasm in spite of the presence of an N-terminal secretory signal peptide. In contrast to secreted GFP, the fusion protein was retained in the cells where it was degraded slowly, albeit at a rate much higher than that of the endoplasmic reticulum (ER)-retained derivative GFP-HDEL. The fusion protein could not be stabilised by inhibitors of transport or the cytosolic proteasome. However, the protein is a strong lumenal binding protein (BiP) ligand. Complete signal peptide processing even after long-term expression in virus-infected leaves rules out the possibility that the documented accumulation in the cytosol and nucleoplasm is because of the bypassing of the translocation pores. The data are consistent with the hypothesis that the fusion protein is disposed off from the ER via a retrograde translocation back to the cytosol. Moreover, accumulation in the nucleoplasm was shown to be microtubule dependent unlike the well-documented diffusion of cytosolically expressed GFP into the nucleoplasm. The apparent active transport of the GFP fusion into the nucleoplasm may indicate an as yet undiscovered feature of the ER-associated degradation (ERAD) pathway and explain the insensitivity to degradation by proteasome inhibitors.


As nascent polypeptides are translocated into the endoplasmic reticulum (ER) lumen through the Sec61 pore complex (Brodsky, 1998), the signal peptide is cleaved and the proteins are likely to attain the final conformation only after the import is complete (Gething and Sambrook, 1992; Hartl, 1996). Malfolded and incompletely assembled proteins emerge as by-products in the ER. Several errors, such as the incorrect incorporation of amino acids or premature termination of translation, can give rise to mutant or truncated proteins. A variety of quality control mechanisms operate to prevent their release into the cell or the apoplast where they could be detrimental to the physiology of the plant. This process also leads to their disposal so that valuable amino acids can be recycled.

Degradation of newly synthesised malfolded or unassembled proteins in the secretory pathway is an essential function of the quality control system in the ER so that only correctly folded, processed, and completely assembled proteins exit from this compartment for further transport. One strategy for protein quality control in the secretory pathway involves retention in the ER followed by ER-associated degradation (ERAD; Hiller et al., 1996; Ivessa et al., 1999; Ward et al., 1995). Retrograde transport is ATP dependent and requires the luminal ER chaperone BiP (Brodsky et al., 1999; Pempler et al., 1997) and the ER membrane translocation pore protein Sec61p (Brodsky et al., 1999; Huges et al., 1997; Wiertz et al., 1996).

The ERAD pathway in yeast and animal cells has mainly been demonstrated indirectly via inhibition of degradation by the cytosolic proteasome (Amshoff et al., 1999; Hiller et al., 1996; Huges et al., 1997; McCracken et al., 2000; Pempler et al., 1997; Simpson et al., 1999; Wiertz et al., 1996). More importantly, it is still unclear how permanently malfolded proteins are distinguished from folding intermediates that may yet assume the correct conformation.

More recently, it has been demonstrated that the ERAD route may not require full unfolding of the ERAD substrate, as a green fluorescent protein (GFP) could be tagged onto an ERAD substrate and followed from the ER to the cytosol (Fiebiger et al., 2002). In other words, retrograde translocation has different properties compared to the classic forward translocation, as predicted from genetic studies (Zhou and Schekman, 1999).

In plants, little is known about the mechanisms and proteolytic systems that mediate protein quality control in the secretory system. A suggestion of the existence of a protein retrograde transport and degradation pathway in plant cells has arisen from the analysis of the location of the ricin catalytic A subunit in tobacco protoplasts (Frigerio et al., 1998). It has also been demonstrated that the ricin A chain in tobacco protoplasts is subject to de-glycosylation and degradation after ER retrotranslocation (Di Cola et al., 2001). However, there is currently no evidence of a malfolded protein being targeted from the ER to the cytosol for degradation, although it is known that malfolded luminal proteins can be targeted to the vacuole via a Golgi-mediated pathway (Coleman et al., 1996; Pueyo et al., 1995). Some other malfolded proteins are degraded via a brefeldin A and heat shock-independent pathway in a so far unidentified cellular compartment (Pedrazzini et al., 1997).

Here, we show that it is possible to visualise the ERAD route of malfolded proteins, subjected to the quality control processes in vivo, using chimeras of calreticulin subunits and GFP. We demonstrate that the transport of malfolded proteins, after retrograde exit from the ER, is cytoskeleton dependent and results in accumulation within the nucleoplasm. The slow degradation is possibly because of this unusual transport route and leads to relatively easy detection of ERAD activity.


The calreticulin P-domain deviates GFP from the secretory pathway to the cytosol and the nucleoplasm

To generate vital markers suitable for the visualisation of quality control routes in plant cells, we took advantage of previous observations that small deletions or N-terminal fusions to plant calreticulin cause severe malfolding, induction of the ER chaperone BiP and lack of secretion (Crofts and Denecke, unpublished). Therefore, a secreted derivative of GFP was fused to various calreticulin domains, and chimeric proteins were tested for increased ER retention and vacuolar or cytoplasmic location. Figure 1 shows a comparison, in Nicotiana tabacum leaf protoplasts, of the behaviour of a GFP fusion protein containing the P-region of calreticulin (Michalak et al., 1992) fused to the C-terminus of GFP (sGFP-P) and sGFP-HDEL, a well-documented ER marker. As expected, the latter accumulated in the mobile elements of the ER and the nuclear envelope (Figure 1a, arrowheads). In contrast, sGFP-P showed much weaker fluorescence and was only faintly detected in the ER, nuclear envelope, the cytosol and the nucleoplasm (Figure 1b). Visualising GFP in the protoplasts expressing sGFP-P was very difficult because of a very low signal to noise ratio of the GFP in these cells.

Figure 1.

Retention and retrograde transport of GFP fusion proteins.

Tobacco protoplasts expressing sGFP-HDEL and sGFP-P were analysed by confocal microscopy 72 h after DNA transfer.

(a) The fluorescence of sGFP-HDEL accumulates in the ER (empty arrowhead) and in the nuclear envelope (full arrowhead, scale bar = 10 µm).

(b) The fluorescence because of sGFP-P is weak in comparison to sGFP-HDEL. However, it is detectable in the ER, nuclear envelope, cytosol and nucleoplasm (full arrowhead, scale bar = 10 µm). Imaging of sGFP-P required on average five times higher laser intensity with respect to that of sGFP-HDEL for the same pinhole aperture. The empty arrowhead indicates autofluorescent chloroplasts.

(c–i) Imaging differently targeted GFPs in virus-transformed Nicotiana bentamiana leaf epidermal cells.

(c) Ten days post virus infection expression of PVX.sGFP-HDEL is located in the ER (empty arrowhead) and the nuclear envelope (full arrowhead, scale bar = 20 µm).

(d) Expression of PVX.sGFP-P is located in the cortical ER (empty arrowhead) and the cytoplasm (full arrowhead, scale bar = 10 µm).

(e) An optical section through the middle of an epidermal cell shows that sGFP-P also accumulates in the nucleoplasm (full arrowhead) and the cytoplasm. Note that small vacuoles (empty arrowhead) can be seen in negative contrast and the apoplast is delimited by two layers of cytoplasmic fluorescence (star, scale bar = 10 µm).

(f) Double labelling the nuclei with propidium iodide (50 mg l−1, 10 min) after an RNase treatment (50 mg l−1, 5 min) in the same cells as in (e) confirms their identity (full arrowhead) in sGFP-P-expressing cells.

(g,h) Trichomes expressing PVX.sGFP-P and double labelled with FM64 dye (Molecular Probes, California), which stains the plasma membrane (red, arrowhead), confirms the absence of the apoplastic location of sGFP-P (scale bar = 15 µm).

We also expressed sGFP-HDEL and sGFP-P in the epidermal cells of Nicotiana benthamiana leaves using the viral vector pTXS.P3C2 (Boevink et al., 1996), thus permitting expression to be followed for longer periods of time. Figure 1(c) shows that sGFP-HDEL labels the nuclear envelope and the classical cortical ER network. In contrast, sGFP-P labels not only the nuclear envelope and the ER, but also the cytosol and the nucleoplasm (Figure 1d,e,g). This pattern was not altered in cells subjected to cold shock for 30 min or brefeldin A (100 µg ml−1) for 3 h (data not shown). Also in tobacco protoplasts, the amount of cellular sGFP-P could not be increased by treatment with brefeldin A or inhibition of ER export via co-expression with Sec12 (Phillipson et al., 2001), suggesting that the protein was not transported to distal locations of the secretory pathway and then degraded (data not shown). The presence in the cytosol and the nucleoplasm was confirmed by double labelling with propidium iodide (compare Figure 1e,f). Even though the sGFP-P hybrid did not contain a C-terminal HDEL motif, no significant levels of secretion could be detected (Figure 1g,h). Under the same experimental conditions, secreted GFP lacking the HDEL retention signal accumulated in the apoplast (data not shown; also see Boevink et al., 1999), while free GFP (cGFP) resulted in diffuse fluorescence in the cytosol and the nucleoplasm but not in the ER (Figure 8).

Figure 8.

Photobleaching analysis of the fluorescence of cells transformed with PVX.cGFP, PVX.sGFP-P and PVX.sGFP-C.

The fluorescence of the nucleus of an untreated epidermal cell transformed with PVX.cGFP, releasing cytoplasmic GFP (a, arrowhead; scale bar = 10 µm) is not appreciably quenched in 1 min (b), resulting in a little difference in the fluorescence intensity in the recovery period (c). Parallel results obtained with the control in which the nuclei of colchicine-treated cells (d, arrowhead, scale bar = 10 µm) do not show appreciable loss of fluorescence after photobleaching (e) and therefore recovery (f). The fluorescence of a nucleus of an untreated epidermal PVX.sGFP-P-transformed cell (g, arrowhead; scale bar = 10 µm) shows little quenching after 1 min of photobleaching (compare (h) with recovery image (i), taken 30 sec later). A nucleus of a cell transformed with PVX.sGFP-P treated with colchicine (j, arrowhead; scale bar = 10 µm) was photobleached for 1 min (k) and does not show the recovery of fluorescence within the nucleoplasm after 1 min (l) or even 5 min (not shown), but recovery can be seen in the nuclear envelope. (m–o) Epidermal cells transformed with PVX.sGFP-C showed accumulation of GFP in the ER, cytosol and nucleoplasm, similar to those transformed with PVX.sGFP-P. Photobleaching the nuclei of untreated epidermal cells (m, arrowhead, scale bar = 10 µm) did not induce an appreciable loss of fluorescence after 1 min of photobleaching (n), and no recovery was therefore detectable within 5 min (o). (p–r) Photobleaching the nuclei of colchicine-treated cells expressing PVX.sGFP-C (p, arrowhead, scale bar = 10 µm). After photobleaching for 1 min, the nucleus of cells expressing sGFP-C in the presence of colchicine showed loss of fluorescence (q) and no recovery even after 5 min (r). White horizontal line indicates the transect along which pixel intensity was measured and plotted as arbitrary fluorescence units in the overlaid graph.

Thus, we can conclude that the calreticulin P-domain deviates the GFP fusion from the secretory pathway, even though it is fused to the C-terminus of the protein and should be synthesised only after signal peptide-mediated co-translational translocation (Vitale et al., 1993) has been initiated.

Cytosolic and nuclear localisation of sGFP-P is not a result of bypassing translocation into the ER lumen

One possible explanation for the above result could be that high expression levels of sGFP-P lead to significant bypassing of the translocation pores in the rough ER membrane. If so, then this would be accompanied by a lack of signal peptide cleavage. Such an event can easily be detected, as the presence of a signal peptide is known to cause an increase in the molecular weight of the nascent secretory protein produced in vitro in the absence of microsomes (Denecke et al., 1990). We thus generated a similar GFP fusion protein that lacked an N-terminal signal peptide (cytosolic GFP-P or cGFP-P) and compared its molecular weight with that of sGFP-P after production in protoplasts, in virus-infected leaves and in Escherichia coli (Figure 2).

Figure 2.

Cellular location of sGFP-P is not because of the bypassing of translocation into the ER.

Cellular extracts of protoplasts transformed with sGFP-P (lane 2), of leaves virally transformed with sGFP-P (lane 3), cGFP-P (i.e. lacking the signal peptide, lane 4), of Escherichia coli expressing cGFP-P (lane 5) and sGFP-P (lane 6) were analysed by SDS–PAGE and blotting with GFP antiserum. The presence of one band in all the lanes with the exception of the sGFP-P expressed in E. coli indicates cleavage of the sporamin signal peptide and therefore absence of protein mistargeting. Lane 1 = negative control.

Expression of sGFP-P and cGFP-P in E. coli clearly showed an additional higher molecular weight band when the signal peptide was present (Figure 2, compare lanes 5 and 6). Cytosolic GFP-P was produced at much higher levels (data not shown), and the sample was diluted 10-fold compared to the E. coli extract from sGFP-P producers to allow comparison of the molecular weights. The data are consistent with the generally lower capacity of periplasmic protein production in E. coli compared to that of cytosolic production. The additional band is because of incomplete signal peptide processing, hence providing a molecular weight marker to compare it with the proteins expressed in planta.

In contrast, sGFP-P and cGFP-P produced in virus-infected leaves have an identical mobility on SDS–PAGE (Figure 2, compare lanes 3 and 4), which is also similar to that of sGFP-P produced in tobacco leaf protoplasts (lane 2). No higher molecular weight form containing a signal peptide could be detected, indicating faithful translocation into the ER, regardless of the method of gene transfer.

The data confirm that signal peptide processing in planta is generally complete and difficult to bypass (Denecke et al., 1990). This is a strong evidence against the bypassing of translocation, together with the fact that the onset of translocation and probably signal peptide processing is likely to be completed well before the P-domain emerges from the ribosome. Therefore, we propose that sGFP-P first enters the ER lumen at least partially, undergoes efficient signal peptide processing by signal peptidase, and is later re-located to the cytosol and the nucleoplasm.

sGFP-P associates with BiP and is degraded

To obtain evidence for malfolding of the C-terminal portion of the fusion protein, we tested if sGFP-P can interact with BiP in an ATP-dependent fashion. Virus-infected leaf discs were subject to in vivo labelling followed by extraction and immunoprecipitation with anti-BiP serum (Figure 3). Plants expressing sGFP-P exhibited higher synthesis rates for BiP compared to the control (Figure 3, compare lanes 1 and 2), indicating ER stress possibly because of the presence of malfolded protein. Furthermore, an abundant protein of lower molecular weight co-precipitated with BiP. This protein was clearly absent in the control lane in which in vivo labelling of leaf discs from untransformed plants revealed only BiP in the immunoprecipitate.

Figure 3.

Anti-BiP immunoprecipitation of extracts of leaves transformed with PVX.sGFP-P.

Discs of untransformed and PVX.sGFP-P-transformed leaves of Nicotiana benthamiana were labelled for 3 h with 35S-methionine and 35S-cysteine. After immunoprecipitation, proteins were analysed by SDS–PAGE and fluorography. Extracts of labelled leaves, untransformed (lane 1) and transformed with PVX.sGFP-P (lane 2) were immunoprecipitated with anti-BiP (α-BiP). The precipitate was then washed with ATP (+ATP, lane 3 (untransformed) to lane 4 (transformed)), and ATP washes were immunoprecipitated with maize anti-calreticulin (α-Cal, lane 5 (untransformed) to lane 6 (transformed)].

To confirm that the band corresponds to sGFP-P, pellets were washed with ATP (3 mm) in BiP release buffer. Figure 3 shows that the majority of the co-precipitating protein was removed from BiP (lane 4) but it could be re-precipitated from the supernatant with anti-maize calreticulin antibodies (Figure 3, lane 6). Thus, the result confirmed the identity of the co-precipitated protein as the GFP fusion carrying the P region of calreticulin.

To test whether the interaction with BiP was because of GFP itself or the fused P-domain, we compared cytosolic GFP with secreted GFP (sGFP), sGFP-HDEL and sGFP-P (Figure 4). To avoid the variation inherent to virus infections, the efficiency of which is difficult to control or quantify, as well as differences in the physiology of individual leaves which could influence BiP levels, we used transient expression in tobacco leaf protoplasts as a model system. All experimental parameters were thus the same, except for the transfected plasmid.

Figure 4.

Evidence showing that sGFP-P constitutes a typical BiP-ligand with strong ATP-dependent affinity compared to other GFP derivatives.

(panel a) Pulse chase of 70% 35S-methionine and 30% 35S-cysteine-labelled tobacco protoplasts transformed with DNA encoding for cytosolic GFP (cGFP), secreted GFP (sGFP), ER-retained GFP (sGFP-HDEL) and secreted GFP fused to the P region of calreticulin (sGFP-P). sGFP co-precipitates BiP but is degraded during the chase period.

(panel b) Quantification of pulse-chased sGFP-HDEL and sGFP-P immunoprecipitation, confirming the instability of sGFP-P.

(panel c) Line scans of pulse-chase samples from an experiment in which sGFP-P and sGFP-HDEL were co-expressed in the same protoplasts. Note the much faster degradation of sGFP-P compared to that of sGFP-HDEL.

Protoplasts were transfected and metabolically pulse-labelled (P) followed by a 10 h chase period (C), and cell extracts and culture medium were immunoprecipitated with anti-GFP serum. Figure 4 shows that sGFP-P co-precipitates much more BiP than the three controls (panel a). Line scans are shown for sGFP-HDEL and sGFP-P to illustrate that the differences are quantitative (panel b).

Figure 4 also shows that sGFP-P is slowly degraded, but much more rapidly than sGFP-HDEL whose half-life is well beyond the entire chase period. Quantification of the signal from phosphorimaging revealed that after the chase, labelled sGFP-P dropped to 6.7% during the chase period. This corresponded to an estimated half-life of 2.5 h. It can be concluded that sGFP-P is degraded because it does not chase into the medium (data not shown). In contrast, the appearance of sGFP in the medium is easily detected after the chase (not shown).

To confirm these results, we co-expressed sGFP-HDEL and sGFP-P in the same protoplast suspension to rule out variance in the pulse-chase experiments. The line scan reveals the signal for sGFP-HDEL, sGFP-P and BiP, and again illustrates the much shorter half-life of sGFP-P compared to that of sGFP-HDEL (Figure 4c).

The weak association of BiP with sGFP-HDEL may indicate that GFP folds very slowly. The apparent increase in the amount of co-precipitating BiP with sGFP-HDEL after the chase (Figure 4b) indicates that labelled BiP remains stable during the chase period and can be associated with de novo synthesised (cold) sGFP-HDEL. This is not surprising as the total amount of sGFP-HDEL increases significantly during the experimental period. In conclusion, the data clearly suggest that sGFP-P has increased BiP-binding properties and is also degraded more quickly than the GFP controls.

GFP-P does not produce the typical degradation intermediate shown by secreted GFP

The results from Figure 4 indicate the presence of a degradation intermediate of lower molecular weight for sGFP and sGFP-HDEL. This intermediate was identical in molecular weight in both cases, although the precursors differed in the presence of the HDEL motif. We concluded that GFP was most likely proteolytically cleaved near its C-terminus, yielding identical proteins regardless of the presence of the HDEL motif. This band of lower molecular weight could also be seen in Western blots, but was only detected in the cells (Figure 5b) and not in the medium (Figure 5a). It should be noted that the P-domain was also fused to the C-terminus as the HDEL peptide, but interestingly, this C-terminal processing intermediate was not observed for sGFP-P (Figure 5). The absence of the typical intermediate suggested that degradation of this fusion protein occurred in a different manner.

Figure 5.

Despite the absence of a retention/retrieval signal, the sGFP-P fusion is not secreted.

Tobacco protoplasts were electroporated in the absence of foreign DNA or with DNA encoding for cytosolic GFP (cGFP), secreted GFP (sGFP), ER-targeted GFP fused to HDEL (sGFP-HDEL) or to the P region of maize calreticulin (sGFP-P). Proteins were detected by SDS–PAGE (10%) blotting with GFP antiserum.

(a) Immunodetection of proteins in the protoplast medium. The efficiency of secretion of sGFP is high compared to that of the other constructs as detected in blots on 4× concentrated medium. Despite the absence of a retention/retrieval signal, sGFP-P is not secreted (arrowhead).

(b) Immunoblots of cellular extracts show bands with expected gel mobility and partial degradation products of sGFP and sGFP-HDEL, but not of sGFP-P (arrowhead).

The ratio between precursor and degradation intermediate was shifted towards the precursor when the HDEL motif was present. Given the proteolytically benign environment of the ER lumen, we postulated that the degradation product was a result of ER export, and occurred downstream of the ER in the secretory pathway, not in the medium.

To provide further evidence for this, we conducted pulse-chase labelling of sGFP-HDEL-producing cells and separated cells into soluble and microsomal fractions via an established osmotic shock procedure (Denecke et al., 1992). Figure 6 shows that the degradation intermediate was only detected after a prolonged chase (14 h) but exclusively in the soluble fraction, which also contains vacuolar contents. The results are consistent with the hypothesis that a portion of sGFP-HDEL is transported out of the ER and may explain the generally low fluorescence of secreted GFP in the apoplast (Batoko et al., 2000; Boevink et al., 1999). Figure 6 again demonstrates the stability of sGFP-P, which is generally lower than that of sGFP-HDEL.

Figure 6.

Proteasomal inhibition with clasto-lactacystin β-lactone has no detectable effect on sGFP-P.

Protoplasts co-electroporated with sGFP-P and sGFP-HDEL were treated with the proteasome inhibitor clasto-lactacystin β-lactone (100 µm) after 1 h pulse. Samples were then chased for 3 and 14 h. Chased protoplasts were subjected to osmotic shock. sGFP-P and the lower molecular size band present in sGFP-HDEL lane were not stabilised by the inhibitor. – = control; βL = clasto-lactacystin β-lactone; S = soluble fraction; P = pellet.

Degradation of sGFP-P is not because of a clasto-lactacystin β-lactone-sensitive machinery

Recent work has shown that the ricin A chain exploits ER retrotranslocation before being degraded in the protoplast cytosol upon de-glycosylation (Di Cola et al., 2001). The degradation is partially reduced by the proteasome inhibitor, clasto-lactacystin β-lactone. On the basis of these studies, it has been suggested that ricin A may be degraded by the proteasome, even if the protein is not necessarily malfolded.

We pulsed co-electroporated protoplasts with sGFP-P and sGFP-HDEL as control and chased them at 3 and 14 h in the absence or presence of 100 µm clasto-lactacystin β-lactone. sGFP-HDEL appeared in the soluble and the pellet fractions. The presence of this protein in the soluble fraction might be because of the release of the luminal content of the endomembranes during preparation of the samples.

The sGFP-P fusion did not appear to be stabilised by the proteasome inhibitor (Figure 6). An instability of the sGFP-P in the cytosol might indicate the lack of increase in the soluble and pellet ratio over the time course of the experiment.

The slower degradation of sGFP-HDEL and the appearance of the lower molecular weight form was also not influenced by the presence of the proteasome inhibitor.

sGFP-P first accumulates in the ER and then in the nucleoplasm

It is not possible to show that the microsomal sGFP-P chases out into a cytosolic form. Such classical pulse-chase result is only possible if the chase product is at least as stable or more stable than the precursor, but this does not seem to be the case. However, the complete signal peptide processing, as well as the typical ER fluorescence pattern, indicates that sGFP-P enters the ER and folds sufficiently to support fluorescence. To completely rule out that sGFP-P in the ER/nuclear envelope and the cytosol/nucleoplasm derived from two different pools exhibiting different targeting routes, we monitored fluorescence distribution in leaf epidermal cells as a function of time. sGFP-HDEL accumulated in the nuclear envelope and in the ER 3 days after virus infection (Figure 7a). The location of the protein fusion did not change with prolonged virus infection (Figure 7b). Figure 7(c) shows that 3 days after virus infection, sGFP-P was primarily detected in the ER and the nuclear envelope. However, prolonged incubation for 2 weeks resulted in the accumulation of sGFP-P in the nucleoplasm with lower levels detected in the nuclear envelope (Figure 7d). The data suggested that the location of sGFP-P shifted from the ER to the cytosol and the nucleoplasm with time. This result argues against the presence of two alternative pools of the fusion protein and is consistent with our hypothesis of retrograde ER transport of sGFP-P.

Figure 7.

Accumulation of sGFP-P in the cytoplasm and nucleoplasm is time dependent.

With sGFP-HDEL, fluorescence can be seen in both the ER and the cross-sections of the nucleus, clearly in the nuclear envelope 3 days (a) and 3 weeks (b) post-infection. With epidermal cells expressing sGFP-P, 3 days post-infection fluorescence can clearly be seen in the nuclear envelope (c); however, 2 weeks post-infection fluorescence accumulates in the cytoplasm and nucleoplasm masking residual fluorescence in the nuclear envelope (d). Relative fluorescent intensity in the nuclear envelope compared to the nucleoplasm can easily be seen via line scans across the images.

It is worth noting that the protoplasts and the epidermal cells appear to exhibit different levels of sGFP-P fluorescence, with the protein showing increased stability in the leaves. This may reflect differences in the species used for the two expression systems or a different metabolic state between the cells in planta and the protoplasts. The higher detection of sGFP-P in the leaves may therefore be because of either a slower degradation of sGFP-P or a quicker saturation of the degradation machinery compared to the protoplasts.

Nuclear targeting of sGFP-P is active and microtubule dependent

Even though cytoplasmic detection of sGFP-P could be explained by the ERAD pathway, the location in the nucleoplasm was unexpected. This could be because of the fusion protein diffusing through the nuclear pores thus bypassing the proteasome. To test this, we monitored fluorescence recovery of the nucleus after photobleaching (FRAP; Figure 8) and investigated the dynamics of the process by comparing the movement of cytosolic GFP and sGFP-P. Both cytosolic GFP and sGFP-P were difficult to bleach in the nucleoplasm because of the rapid recovery of fluorescence, presumably because of rapid diffusion/transport of the proteins back into the nucleus (Figure 8a–c,g–i).

We also analysed the influence of the microtubule-destabilising drug colchicine (1 mm for 30 min) on the transport of the proteins into the nucleus. Colchicine did not prevent the nuclear localisation of cytoplasmic GFP after photobleaching (Figure 8d–f), indicating that this molecule diffused rapidly into the nucleus in a microtubule-independent fashion. The same results were obtained after treatment with another plant-specific microtubule inhibitor, oryzalin (10 µg ml−1 for 30 min; data not shown). However, in sGFP-P-expressing cells treated with colchicine (Figure 8j), the nucleoplasm was bleached but a significant amount of fluorescence persisted in the nuclear envelope (Figure 8k). Although fluorescence subsequently recovered fully in the nuclear envelope well within 1 min after bleaching (Figure 8l), reflecting rapid diffusion from the ER lumen, fluorescence in the nucleoplasm did not recover under these conditions.

These results suggested that nuclear targeting of sGFP-P from the cytosol, but not cytosolic GFP itself, was microtubule dependent and illustrated that the two molecules reached the nucleoplasm via different pathways. It could thus be ruled out that the nuclear targeting of sGFP-P was because of an sGFP-intrinsic nuclear localisation mechanism.

A putative nuclear localisation signal has been identified within the P coding region of maize calreticulin (Napier et al., 1994). To test if this region was responsible for our observations, we carried out similar experiments on an ER-targeted GFP fused to the C region of the maize calreticulin (sGFP-C). The latter region did not contain a nuclear targeting sequence, but fusion of this domain caused a similar distribution of fluorescence compared to that of sGFP-P and included the ER, the nuclear envelope, the cytoplasm and the nucleoplasm (Figure 8m–r). In FRAP experiments in the absence of colchicine, bleaching was partial, followed by a quick recovery (Figure 8m–o). The fluorescence in the nuclei of colchicine-treated cells transformed with sGFP-C did not recover after photobleaching (Figure 8p–r). Thus, we concluded that the nuclear targeting of sGFP-P and sGFP-C was independent of any known motif and might be a novel feature of the ERAD pathway.


Visualisation of the ERAD pathway with a GFP fusion

Quality control by the ER is proposed to occur via two different mechanisms, the disposal by the lytic vacuole (Hong et al., 1996) and the ERAD pathway (Ward et al., 1995). Evidence for the former pathway arose from the generation of an invertase fusion protein containing a C-terminal malfolded portion (Hong et al., 1996). It could be argued that this fusion protein mimics a partially unfolded protein as it is composed of a folded portion (invertase) and an additional malfolded portion that is recognised by the quality control machinery. This disposal route was shown to be dependent on the vacuolar sorting receptor VPS10, which could recognise the incompletely folded protein either directly, or perhaps via interaction with an adapter molecule that becomes a VPS10 ligand once it is associated with an incorrectly folded protein. It is also argued that such a mechanism would be energetically economic compared to ERAD, as the unfolding of the entire molecule (including the folded invertase) would cost more energy. In contrast, the proteins that are completely unfolded may follow the ERAD pathway as they remain translocation competent.

The GFP fusion protein used in this study appeared to be sufficiently folded to support fluorescence but yet permitted us to visualise the ERAD pathway. The fact the ER-targeted GFP constructs without H/KDEL retention signals have been shown to be secreted (Batoko et al., 2000; Boevink et al., 1999) suggests that a retention mechanism must be in operation to explain ER staining because of sGFP-P. This is most likely because of the association of sGFP-P with BiP as demonstrated biochemically. As proposed for rice prolamins (Li et al., 1993) and for assembly-defective form of the bean vacuolar storage protein phaseolin (Pedrazzini et al., 1997), proteins can thus be retained in a signal-independent manner within the ER through prolonged interaction with other proteins that contain an HDEL signal. In the case of sGFP-P, the candidate would be the molecular chaperone BiP, as suggested by the biochemical evidence presented here.

Biochemical and genetic analyses have indicated that luminal proteins were selected for proteolysis and exported to the cytosol by the ERAD process. This required initial retention of the substrate protein in the ER by BiP (Brodsky and McCracken, 1999; Pempler et al., 1997) and was followed, sometimes after a considerable lag phase, by rapid degradation without the detection of detectable proteolytic intermediates in the cytosol (Hiller et al., 1996; Ivessa et al., 1999; Sommer and Wolf, 1997). The dynamics of sGFP-P, namely entering the ER lumen, an ATP-dependent interaction with BiP as well as a distribution in the ER, the cytosol and the nucleoplasm, showed features similar to those exhibited by ERAD substrates. Moreover, the distribution of the fluorescence because of sGFP-P expression was not altered by BFA or cold treatment, as would be expected if a transport between ER and Golgi along the secretory pathway prior to the release into the cytosol was involved (Boevink et al., 1999). Our results therefore indicate in vivo evidence of the occurrence of an ER retrograde transport of a protein.

The aberrant location of sGFP-P is not because of bypassing the translocation process itself, but occurs after cleavage of the signal peptide. As higher molecular weight forms of the constructs containing the signal peptide were not detected, it is likely that the protein enters the translocation machinery, although it does not show whether translocation is completed before retrograde transport is initiated. Most importantly, the fact that sGFP-P is first detected in the ER and later in the cytosol and nucleoplasm rules out the possibility that one pool of sGFP-P enters the ER while the other bypasses translocation. If the latter were the case, deviation from the secretory pathway should be independent of time.

There is also significant evidence suggesting malfolding of sGFP-P or portions of it. First, sGFP-P is a much stronger BiP ligand than sGFP-HDEL itself. It is also degraded faster than sGFP-HDEL, albeit slowly compared to the findings on the ricin A chain. We have also shown that sGFP and sGFP-HDEL are proteolytically cleaved into a defined degradation intermediate, also seen in A. thaliana plants expressing secreted GFP and GFP-HDEL (Zheng, Hawes and Moore, unpublished), which is not seen for sGFP-P. This suggests that sGFP-P is not degraded in the same way. We propose that this GFP fusion highlights the ERAD pathway in plant cells. Perhaps, the small size of GFP allows retrograde translocation in spite of folding into a conformation that supports partial fluorescence, a possibility recently suggested by Fiebiger et al. (2002).

An unexpected result was the time-dependent location of sGFP-P in the nucleoplasm. This appeared to be in a manner distinct from that of the well-known nuclear transport of cytosolic GFP. It may be this unexpected targeting step that prevents the protein from degrading quickly in the proteasome, which would explain the lack of inhibition by lactacystin.

Our results indicate that different pathways for protein degradation may exist in protoplasts. Different sensitivities to degradation pathways verified for sGFP-P and ricin A may depend on the exposure to different signals for degradation. For example, de-glycosylation occurring for ricin A may be an essential prerequisite for proteasome-mediated degradation. However, it is possible that sGFP-P and ricin A may share similar signals for retrotranslocation, but not for degradation. The fact that sGFP-P fusion does not seem to be a candidate for proteasomal degradation could explain the longer half-life of this protein in comparison to ricin A.

Relevance of the nuclear targeting of an ERAD substrate

The unexpected nuclear localisation of the fusion protein prompted us to investigate the dynamics of sGFP-P by FRAP experiments. The nuclear localisation of free GFP could be explained by its small size that was below the exclusion limit for passive transport through the nuclear pore complex. In our experiments, free GFP was localised only in the cytosol and nucleoplasm. Reits et al. (1997) proved by FRAP experiments that, in mammalian cells, free GFP movement to the nucleus follows passive diffusion dynamics. Our FRAP experiments confirm these data, as photobleaching the nuclei of cells expressing free GFP in the presence and absence of the microtubule-destabilising agents did not alter the rapid rate of diffusion of the GFP to the nucleus.

Photobleaching the nuclei of cells transformed with PVX.sGFP-P and treated with microtubule inhibitors induced a non-recoverable loss of fluorescence in the nucleoplasm, although the fluorescence because of sGFP-P quickly recovered in the nuclear envelope. The recovery in the nuclear envelope after photobleaching would be expected to occur because the synthesis pathway starts in the ER/nuclear envelope, followed by diffusion within the ER lumen. However, the complete inhibition of protein movement towards the nucleoplasm by colchicine indicates that the transport of the malfolded GFP fusion to the nucleus is microtubule mediated, and occurs in a manner different from that of cytosolic GFP. Moreover, this movement does not seem to depend on the nuclear targeting sequence present on the sGFP-P itself. In FRAP experiments on PVX.sGFP-C expressing cells, which present a similar cellular distribution pattern to sGFP-P producers, despite the absence of a putative nuclear localisation signal in the C-region of calreticulin, showed that in the absence of microtubules there was no recovery of bleached nucleoplasm fluorescence.

ERAD substrates require cytosolic or nucleoplasmic factors such as proteasomes to mediate their degradation (Ivessa et al., 1999). Proteasomes interact continuously with cytosolic and nuclear proteins and perform quality control to degrade proteins that are either tagged by ubiquitination or otherwise already in an appropriate conformation for degradation (Reits et al., 1997). Little data exist on the location of the proteosome in plants. They have been located in the cytoplasm of potato suspension cells by direct immunofluorescence, are possibly associated with elements of the cytoskeleton and may also reside in the nuclei (Vallon and Kull, 1994).

Reichel and Beachy (2000) have demonstrated that inhibition of proteasomal activity results in the location of a tobacco mosaic virus movement protein–GFP fusion on the ER. These data indicate a location of proteasomes similar to that known for animal cells (Arcangeletti et al., 2000; Palmer et al., 1994; Rivett, 1998). Reits et al. (1997) have showed the occurrence of an active transport of the proteasome into the nuclei. It is therefore reasonable to presume that in the system described here sGFP-P and sGFP-C move towards the nuclei along the proteasome pathway. A possible saturation of the proteasome machinery because of overexpression could justify the visualisation of the GFP fusions in our experiments.


We have illuminated a retrograde transport pathway from the ER lumen to the cytosol using a GFP-tagging approach. In addition, our data demonstrate that the ERAD pathway may involve transport of malfolded proteins to the nucleoplasm. In combination with the biochemical assays for signal peptide processing and interaction with the ER chaperone BiP, this system will be invaluable to elucidate the various steps controlling the disposal pathway for malfolded proteins.

Experimental procedures


Amplification products were generated by using Taq/Pwo (HYBAID Ltd, Ashford, UK). All DNA manipulations were performed according to the established procedures (Sambrook et al., 1989). E. coli XL1-Blue strain (Stratagene, La Jolla, USA) was used for the amplification of all the plasmids with the exception of pET23b(+) vector (Novagen, Madison, USA), which was amplified in BL21(DE3).

Virus vector

Potato virus X (pTXS.P3C2) was chosen as an expression vector (Boevink et al., 1996), and the wild-type GFP was used. The following constructs were cloned as ClaI–NsiI fragments under the transcriptional control of duplicated subgenomic RNA promoters: cytosolic GFP (cGFP, i.e. without any signal); secreted GFP (sGFP, i.e. a fusion of the sporamin signal peptide to GFP at the N-terminus); GFP targeted to the ER by the sporamin signal peptide and retained/retrieved by HDEL (sGFP-HDEL); GFP targeted to the ER by the sporamin signal peptide and C-terminally fused in-frame to aa 185–297 of the P region of maize calreticulin (sGFP-P); untargeted GFP and C-terminally fused in-frame to aa 185–297 of the P region of maize calreticulin (GFP-P); GFP targeted to the ER by the sporamin signal peptide and C-terminally fused in-frame to aa 298–392 of the C region of maize calreticulin (sGFP-C) (Napier et al., 1994).

Protoplast expression

For N. tabacum protoplasts, expression of mGFP5 was used, which is a GFP with a codon usage optimised for plant nuclear expression (Haseloff et al., 1997). Substituting the wild-type GFP with mGFP5 produced the constructs described above, with the exception of sGFP-C, which was not used in protoplast expression. The constructs were amplified and cloned as ClaI–BamHI fragments into pLL4, a novel 35S based plant expression vector in which the NcoI codon spanning the ATG initiation codon was replaced by ClaI. The plasmid contained a spacer in between the promoter and the 3′ untranslated end of the nopaline synthase gene, which ended with a BamHI site.

Escherichia coli expression

For E. coli expression, the coding regions of GFP-P and sGFP-P previously used for virus expression were cloned as an NheI–SacI fragment into an expression vector pET23b(+).

Expression systems

Virus expression

The run-off transcripts and inoculation on leaves of N. benthamiana were produced as described (Boevink et al., 1996). For all the experiments described here, plants 10–14 days after transformation were used.

Protoplast transient expression

Tobacco leaf protoplasts were prepared, and electroporation experiments were performed as described previously (Denecke and Vitale, 1995). Protoplasts were harvested for protein analysis 24 h after electroporation. For microscopy, cell suspensions were cultured for 24, 48, or 72 h.

Escherichia coli expression

The GFP-P and sGFP-P constructs were cloned in the pET-23b(+) vector, and protein production was induced in BL21(DE3) lysogens. Positive clones were selected for low-scale protein production. A single colony was inoculated initially into 5 ml of Luria-Bertani (LB) containing ampicillin (100 µl ml−1) and further expanded into a 50 ml shaker culture in 250 ml flasks. The cells were incubated with shaking at 37°C until OD600 was approximately 0.9. Protein production was induced by the addition of 1 mm IPTG and further incubation of the culture for 2.5 h at 37°C. Cells were then pelleted leaving an equal volume of culture medium with the cells, before storage at −70°C. Protein gel analysis was performed after 1 day.

Protein extraction and gel blot analysis

For immunoprecipitation, leaf discs were homogenised in ice-cold buffer composed of 200 mm Tris–HCl (pH 8), 300 mm NaCl, 1% Triton X-100 and 1 mm EDTA (Crofts et al., 1998). Harvesting of protoplasts and culture medium was performed as described (Denecke and Vitale, 1995).

Protoplast and E. coli cellular extracts were obtained by repetitively pipetting the cells in an ice-cold buffer containing 0.1 m Tris–HCl (pH 7.8), 0.2 m NaCl, 1 mm EDTA, 2% β-mercaptoethanol and 0.2% Triton X-100 (Pedrazzini et al., 1997). Protein analysis of the protoplast culture medium was performed after concentration with aqueous ammonium sulphate solution (final concentration 60%) in the presence of BSA as a carrier (200 µg per 600 µl of medium) and re-suspending the protein precipitate in a buffer containing 25 mm Tris and 5 mm EDTA. An equal volume of protein extract was added to 2× SDS loading dye (Crofts et al., 1998), prior to boiling and gel loading. Protein analysis was performed by SDS–PAGE on 10–12% acrilamide gels. Immunodetection of GFP in protein gel blots was performed using the enhanced chemiluminescence (ECL) system (Amersham, Life Sciences, UK), with rabbit polyclonal antiserum raised against GFP (Molecular Probes, Inc., USA) at 1 : 3000 dilution.

Immunoprecipitation and pulse chase

For sGFP-P immunoprecipitation, untransformed and transformed leaf disks of the same size were labelled for 3 h with 100 µCi ml−1 Pro-mix (Amersham Life Science, UK; containing 70% 35S-methionine and 30% 35S-cysteine) in MS medium. Homogenisation of the leaf tissue and subsequent immunoprecipitation were performed essentially as described (Crofts et al., 1998) with a purified anti-BiP serum. ATP washes of the immunoprecipitates were performed as described previously (Vitale et al., 1993). Re-immunoprecipitation was performed with anti-maize calreticulin. Proteins were analysed by SDS–PAGE (12% gel), and labelled proteins were visualised with a phosphorimager SI. Pulse-chase experiments used a 1 h labelling and chase buffer containing 10 mm methionine and 5 mm cysteine.

Clasto-lactacystin β-lactone (Calbiochem, CN Biosciences, UK) was used as the proteasome inhibitor at a concentration of 100 µm (stock solution 20 mm in DMSO) after a 1 h pulse. Samples were then chased at different time points. Chased protoplasts were pelleted, and proteins were extracted in four times volume of Tris–EDTA buffer (1 mm EDTA, 1 mm Tris).

Confocal microscopy

Leaf segments were imaged with a Zeiss LSM 410 confocal microscope using a 488 nm Argon ion laser plus a narrow bandpass filter set (BP 505–515 nm) and photobleached using the region of interest software. For propidium iodide observations, a 543 nm Helium ion laser with the LP 570 nm emission filter configuration was used. The analysis of the fluorescence intensity was performed with the line measurement software package of the microscope.


The BBSRC is acknowledged for a grant (P08503) supporting this research, and we thank Drs J. Boyce and R. Napier for allowing us the use of the calreticulin DNA and antibodies. Luis L. Pinto daSilva is indebted to the British Council and the Brazilian CNPQ for supporting a scientific exchange to the University of Leeds.