Several vacuolar sorting determinants (VSDs) have been described for protein trafficking to the vacuoles in plant cells. Because of the variety in plant models, cell types and experimental approaches used to decipher vacuolar targeting processes, it is not clear whether the three well-known groups of VSDs identified so far exhaust all the targeting mechanisms, nor if they reflect certain protein types or families. The vacuolar targeting mechanisms of the aspartic proteinases family, for instance, are not yet fully understood. In previous studies, cardosin A has proven to be a good reporter for studying the vacuolar sorting of aspartic proteinases. We therefore propose to explore the roles of two different cardosin A domains, common to several aspartic proteinases [i.e. the plant-specific insert (PSI) and the C–terminal peptide VGFAEAA] in vacuolar sorting. Several truncated versions of the protein conjugated with fluorescent protein were made, with and without these putative sorting determinants. These domains were also tested independently, for their ability to sort other proteins, rather than cardosin A, to the vacuole. Fluorescent chimaeras were tracked in vivo, by confocal laser scanning microscopy, in Nicotiana tabacum cells. Results demonstrate that either the PSI or the C terminal was necessary and sufficient to direct fluorescent proteins to the vacuole, confirming that they are indeed vacuolar sorting determinants. Further analysis using blockage experiments of the secretory pathway revealed that these two VSDs mediate two different trafficking pathways.
In eukaryotic cells, protein traffic from the Golgi apparatus (GA) to vacuoles or lysosomes requires information embedded in the protein. In plants, the targeting signal is part of the protein sequence. Three major groups of plant vacuolar sorting determinants (VSDs) have been identified so far.
The sequence-specific VSDs (ssVSDs) consist of conserved motifs, such as NPIRL or similar, and do not tolerate any major modifications. Usually, this type of signal is located in the N terminus of a protein (Nakamura and Matsuoka, 1993).
The second group of VSDs comprises the C-terminal VSDs (ctVSDs) (Nakamura and Matsuoka, 1993; Neuhaus and Rogers, 1998). These signals are not conserved, nor do they have a defined size; however, common to all C-terminal VSDs is that they are rich in hydrophobic amino acid residues, and they need to be surface-exposed (Zouhar and Rojo, 2009).
A third group of VSDs are those dependent on the physical structure of proteins (psVSDs). These signals are located in the centre of the protein and involve one or more motifs, which become exposed only when the protein acquires its folded conformation (Neuhaus and Rogers, 1998; Jolliffe et al., 2003).
As a result of the variety in plant models, cell types and experimental approaches used to decipher vacuolar targeting processes, it is not clear whether these three groups of VSDs exhaust all the vacuolar targeting mechanisms in plants, nor if they reflect certain protein types or families. Furthermore, some storage proteins carry two types of VSDs, questioning the sorting efficiency of a single VSD (Nishizawa et al., 2006).
The vacuolar sorting of aspartic proteinases (APs) (for a review, see Simões and Faro, 2004) is expected to follow yet another mechanism, as no common VSDs have been identified so far for this protein family. Moreover, the APs are an intriguing group of proteins: different APs accumulate in distinct compartments, and the same protein can be either secreted to the apoplast or directed to the vacuole, depending on the cell type and developmental stage, and suggesting a tight mechanism of regulation of trafficking (da Costa et al., 2010; Oliveira et al., 2010).
Cardosins, an AP family extracted from cardoons, illustrate such plurality and complexity in sorting mechanisms. Despite their high degree of similarity in protein sequence, cardosin A and cardosin B may accumulate in different cell compartments during plant development (Vieira et al., 2001). Cardosin A has mostly a vacuolar localization, in lytic or storage vacuoles, depending on the organ it is expressed in and on specific cell needs (Ramalho-Santos et al., 1997; Pissarra et al., 2007; Pereira et al., 2008; Oliveira et al., 2010). Cardosin A was also localized in vacuoles when expressed in heterologous systems, such as Arabidopsis thaliana or Nicotiana tabacum (Duarte et al., 2008). These features make cardosins, and particularly cardosin A, good reporters for the study of vacuolar trafficking of aspartic proteinases in plant cells.
Little is known about the intracellular route followed by cardosin A to reach its final destination, or about the sorting signal(s) responsible for its vacuolar targeting. Three main working hypotheses are proposed.
Firstly, the role of the plant-specific insert (PSI) domain. Some plant APs contain specific sequences known as PSIs, which are protein domains of 100 amino acids between the heavy and light chains of the enzyme. This PSI domain may be involved in vacuolar targeting. Some mammalian APs, such as cathepsin D, are transported to the lysosome as complexes with so-called saposins (Fusek and Vetvicka, 2005). PSI domains show significant sequence similarity with saposins, and are sometimes referred to as saposin-like proteins (SAPLIPs) (Simões and Faro, 2004). It has been suggested that PSI domains interact with the endoplasmic reticulum (ER) or GA membranes (Faro et al., 1999; Kervinen et al., 1999). This was based on sequence analysis and the crystallographic structure of the plant AP phytepsin, suggesting an interaction with a receptor in the Golgi membrane. However, the role of PSI domains in vacuolar targeting is not clearly defined, and some published data are even contradictory. Terauchi et al. (2006) suggested that the PSI function as a vacuolar determinant may actually be dependent on the type of vacuole in which the AP resides. Therefore, the exact role of the PSI in vacuolar targeting of APs remains to be elucidated.
The second hypothesis proposes that the C-terminal regions of plant APs contain the VSD. This is based on the fact that the C-terminal peptide of barley lectin (VFAEAIA), which is very similar to the C-terminal sequence of cardosin A, and of several other plant APs, is responsible for the protein targeting to the lytic vacuole (Ramalho-Santos et al., 1998).
A third hypothesis, based on studies on barley aspartic proteinase phytepsin, proposes that the sorting of APs to the vacuole is dependent on the three-dimensional structure of the protein, which implies that small motifs all over the protein sequences are involved.
The objective of the work presented here was to explore these hypotheses to better understand the diversity of vacuolar targeting processes in aspartic proteinase, concentrating on the role of PSIs and the C-terminal domain of the proteins. To do so, we used cardosin A as a model protein and undertook a systematic analysis of the intracellular localization of cardosin A and its truncated forms in plant cells. Fluorescent chimeric proteins were designed to decipher the role of the C-terminal regions and PSI domains in vacuolar targeting, using the fluorescent reporter protein mCherry. Agrobacterium-mediated transient expression of these proteins in tobacco leaves was used to test the fluorescent proteins. Their localization was then monitored by confocal microscopy.
Our results show that the presence of either the PSI region or the C-terminal peptide is sufficient to target cardosin A, or mCherry alone, to the vacuole, suggesting that the two domains act as VSDs. Moreover, our data show that these two VSDs may mediate distinct routes to reach the final vacuolar compartment. Further investigations demonstrate that the short peptide FAEA is a key sequence for the efficiency of the C-terminal VSD, meanwhile the absence of a glycosylation site may confer specific properties to the cardosin-A PSI domain. These data are discussed along with the regulation of cardosin A and the trafficking of aspartic proteinases to the vacuoles.
A whole set of chimeric proteins fused with the fluorescent reporter protein mCherry were made (Figure 1), first to validate the use of cardosin A as the reporter protein (Figure 1a, positive control, CdA-mCherry), secondly to address the putative role of the C-terminal and PSI domains of cardosin A as putative VSDs (Figure 1b), and thirdly to further decipher their efficiency and specificities in vacuolar sorting (Figure 1c,d). Each chimeric protein was then infiltrated in leaf epidermis; 3 days post-infiltration, cells were imaged by confocal laser scanning microscopy to assess the intracellular localization of the proteins. Quantitative analyses of the observed fluorescent patterns were performed in order to gain insight into the efficiency of protein targeting to the vacuolar compartments (resulting from the observation of 80–150 cells from three independent experiments).
PSIs or C-terminus domains are sufficient for vacuolar targeting
It has previously been shown that the non-tagged version of cardosin A accumulated in the vacuoles when transiently expressed in N. tabacum leaf epidermis cells (Duarte et al., 2008). Concordantly, our results clearly show that the expression of the fusion protein CdA-mCherry was detected inside the vacuoles in 100% of the observed cells (Figure 2a,b,i; see also Figure 4a). In some cases (10%), in addition to the vacuolar labelling, a punctate pattern may also be seen at the cell periphery, suggesting the presence of the fluorescent proteins in endomembrane compartments upstream of the vacuole, such as ER/GA complexes or pre-vacuolar compartments (PVCs).
To test the ability of C-terminal and PSI domains to achieve this vacuolar targeting, deletions of either the PSI and/or the C-terminal peptide were performed on the native sequence fused with m-Cherry [i.e. CdAΔPSI-mCherry, CdAΔC-ter-mCherry and CdAΔ(PSI+C-ter)-mCherry, see Figure 1b].
The removal of the PSI alone (CdAΔPSI-mCherry) still allowed the vacuolar localization of cardosin A, as the fluorescent chimera was observed in the central lytic vacuole of all the cells expressing the fluorescent chimaerae (100%; Figure 2c,i; see also Figure 4e). In 60% of the observed cells, fluorescence was also detected in small compartments within the cytoplasm (Figure 2d), which may represent protein accumulation in ER/GA complexes or PVC, prior to final accumulation in the vacuole. Being known to play a role in the maturation process of cardosins, it is possible that the lack of PSI delays protein trafficking to the vacuole. Alternatively, it may usually facilitate vacuolar traffic. In all cases, the PSI does not seem to be needed for vacuolar targeting. These results suggest that this CdAΔPSI-mCherry protein accumulating in the vacuole was targeted via another putative VSD than PSI, leaving the C-terminal domain as a potential candidate to test.
Surprisingly, the removal of the C-terminal peptide of cardosin A-mCherry (CdA-ΔC-ter-mCherry) also allowed the vacuolar accumulation of cardosin A in 100% of the observed cells, as a major accumulation of the fusion protein in the vacuole was still clearly observed (Figure 2e,f,i; see also Figure 4i). In 30% of the observed cells, fluorescence was also detected in small compartments within the cytosol, which again may represent an accumulation of the proteins in ER/GA or PVC compartments, because of a slight delay of protein processing/targeting to the vacuoles. Time-lapse experiments would be needed to confirm this delay. Still, these results show that deletion of the C-terminal peptide alone was not sufficient to inhibit vacuolar trafficking, and suggest that another protein domain was acting as a VSD.
When both C-terminal and PSI domains were deleted [CdAΔ(PSI+C-ter)-mCherry], the protein was never observed in the vacuole. The fluorescence was either contained in small compartments within the cytoplasm, suggesting again an accumulation in ER/GA/PVC complexes, or secreted towards the apoplast (Figure 2g–i).
Quantitative analysis of the fluorescence distribution (Figure 2i) clearly suggested a delay in the trafficking of the mutated fluorescent chimaera to the vacuole, compared with non-mutated cardosin A. Fluorescent proteins were retained in compartments within the cytoplasm before reaching their final location. This delay is more accentuated in cardosin AΔPSI-mCh. We cannot exclude that the observed accumulation in intermediate compartments may result from protein misfolding caused by the removal of the protein's domains, therefore causing protein retention along the secretory pathway. Nevertheless, from these results, we postulated that actually the presence of only one domain, either the PSI or the C-terminal peptide, may be necessary and sufficient for vacuolar accumulation of the reporter protein. Deletion of one domain would then not impair the vacuolar targeting, and would mask the VSD function of the lacking domain. The role of the PSI or the C-terminal peptide as VSD was then further analysed to test this hypothesis.
Both the PSI domain and the C-terminal peptide act as vacuolar sorting determinants
To validate our hypothesis that both the PSI domain and the C-terminal peptide VGFAEAA are capable of targeting proteins to the vacuole, we proposed to test the functionality of these putative VSDs on mCherry trafficking by fusing each domain with SP-mCherry (with SP representing the signal peptide needed to enter the secretory pathway; cf. Figure 1c, SP-mCherry, SP-mCherry-C-ter, SP-PSI-mCherry).
First, to validate our experimental approach, two control experiments had to be performed. The localization of SP-mCherry had to be checked: as it has no vacuolar sorting signal, SP-mCherry was expected to be secreted to the apoplast by the default pathway, which was the case here, as none of the cells expressing fluorescent chimera displayed vacuolar labelling, meanwhile almost all of the cells exhibited apoplast labelling (80%; Figure 3a,b,e). Secondly, the ability to redirect the SP-mCherry protein using a well-known VSD was also investigated using the Barley lectin VSD (Dombrowski et al., 1993). The chimeric protein SP-mCherry-Bl_VSD (see Figure 1c) was therefore expressed, and its vacuolar localization was clearly observed in 100% of the cells expressing the fluorescent proteins (Figure 3c–e). In the two controls, 20% of the cells also exhibited a punctate pattern within the cytoplasm, confirming the entry of the protein in the secretory pathway and validating our functional approach.
To investigate the putative function of the PSI domain as a VSD, the PSI region of cardosin A was placed between the signal peptide and mCherry (SP-PSI-mCherry, see Figure 1c). The fluorescence pattern of this SP-PSI-mCherry protein was again exclusively observed in the vacuole (Figure 3f,g), describing the PSI domain as a VSD. Less than 2% of those cells displayed a cytosolic punctate pattern (ER/GA/PVC labelling), suggesting an efficient (i.e. a more rapid and/or complete) vacuolar targeting, with almost no time spent in the endomembrane compartments upstream of the vacuole.
Similarly, the C-terminal sequence of cardosin A, VGFAEAA, was added to SP-mCherry. The fluorescent protein (SP-mCherry-C-ter, Figure 1c) was clearly targeted to the vacuole, as 80% of the cells were labelled (Figure 3e,h,d,i). In 20% of the cells observed, fluorescence was also detected in association with the apoplast, probably resulting from saturation of the system (Figure 3e). This result strongly indicates that cardosin A C-ter peptide VGFAEAA is a vacuolar sorting domain, which may share the characteristics of a C-ter VSD group. Moreover, the small number of cells exhibiting ER/GA labelling (7%) suggests efficient transport between the ER and the vacuole.
When carrying the two VSD domains, as expected the proteins reached the vacuoles in all of the cells observed (Figure 3j,k). Interestingly, however, distinct fluorescent patterns were observed besides the vacuolar labelling, combining the trafficking characteristics of the two former fluorescent chimaera: a small retention in ER/GA/PVC compartments, as seen in SP-PSI-mCherry, and a secretion to the apoplast, as seen in SP-mCherry-C-ter, suggesting the occurrence of two distinct sorting mechanisms.
As a whole, these data show that both the PSI and the C-terminal peptide VGFAEAA are indeed vacuolar sorting determinants, with one of them being sufficient for sorting mCherry protein to the vacuole. The presence of these two VSDs on one protein raises several questions on the functional relations between the two domains, however, and their efficiency in vacuolar targeting. The lack of one domain may provoke a delay in protein targeting. Does it result from a defect in protein processing? Or from a loss of targeting efficiency because of a synergy between the two signals? Do these distinct VSDs cover different vacuolar trafficking pathways? Is one VSD dominant over the other? Are they specific for cardosin A or ubiquitous?
The PSI and the C terminus may specify different trafficking pathways
The occurrence of two vacuolar sorting determinants on the same protein has previously been reported for a storage protein from Glycine max (Nishizawa et al., 2006). It may be related either to a functional redundancy in a unique vacuolar targeting pathway or to distinct targeting mechanisms to the vacuole, which may even implicate different routes. To address the hypothesis that PSI and C-terminal domains may mediate distinct vacuolar pathways, we postulated that reporter proteins for each of these two putative pathways (i.e. XXX-∆PSI-Cter or XXX–PSI-∆Cter) would react differently to specific blockages of the ER–Golgi pathway or of the post-Golgi–vacuole pathway. Dominant-negative mutant proteins known to alter such pathways were therefore co-expressed with our set of fluorescent chimaera, and the induced fluorescent patterns were imaged. Aleurain-GFP was used to control blockage efficiency, as previously described (da Costa et al., 2010). Aleurain-GFP labels the PVC (Figure S1a), as revealed by a characteristic punctate pattern around the lytic vacuole. This labelling is significant of a protein vacuolar pathway. When co-infiltrated with the dominant-negative mutants RabD2aN121I, SarI H74L, or RabF2b S24N, this characteristic pattern was altered: replaced by an ER-like pattern in the cytosol and around the nuclei in the two former cases (Figure S1b,c) or directed to the apoplast in the latter case (Figure S1d). Those induced modifications of aleurain-GFP transport to PVC validate the experimental approach described below. The blockage experiments with dominant-negative proteins were highly convincing, as all the cells expressing the fluorescent proteins showed the same fluorescence accumulation pattern.
In control experiments, and as previously shown, any protein chimaera carrying either a PSI or a C-terminal, or both domains, exhibit a vacuolar localization (Figure 4, first column). Co-expressed with the dominant-negative mutant RabD2a N121I, which specifically blocks traffic between ER and the Golgi by inhibiting the activity of a GEF co-factor (Batoko et al., 2000), a clear difference occurred between the fluorescent proteins guided to the vacuole by a C-ter sequence, and those driven by a PSI domain (Figure 4, column 2). This suggests a differential sensitivity of these two domains to the ER–GA blockage. Indeed, the vacuolar fluorescent proteins containing only the C-ter determinant (i.e. cardosin AΔPSI-mCherry and SP-mCherry-C-ter) never reached the vacuole when co-expressed with RabD2a N121I (Figure 4f,r). Meanwhile, the two fluorescent vacuolar proteins that just contained the PSI domain (i.e. CdA-∆c-ter-mCherry and SP-PSI-mCherry) always reached the vacuoles (Figure 4j,n). These observations strongly suggest that the PSI domain is insensitive to the blockage induced by the defective RabD2a N121I machinery, and that it mediates an additional ER to vacuole route.
Interestingly, any chimeric protein carrying both the C-ter and PSI domains (i.e. CdA-mCherry and SP-PSI-mCherry-Cter) never reached the vacuoles when co-expressed with the RabD2a N121I mutant (Figure 4b,v, respectively). This suggests that the C-terminal VSD is dominant over the PSI.
To confirm these results, similar experiments were performed using a co-expression with Sar1 H74L, which also blocks the ER–Golgi trafficking by impairing COPII vesicle formation (Andreeva et al., 2000). The results obtained (Figure 4, column 3) were similar to those described above, showing a clear inhibition of vacuolar accumulation when the C-ter sequence is present alone in the reporter protein sequence (Figure 4g,s) or appears together with a PSI domain (Figure 4c,w). There is, however, no efficient blockage when only the PSI sequence was driving the protein (Figure 4k,o). This set of results strongly suggests that the additional ER–vacuole route, taken by the PSI domain, is actually a COPII-independent pathway, and possibly bypasses the Golgi.
This hypothesis was further reinforced by results obtained following a third blockage of the ER–GA–vacuolar pathway, this time affecting post-GA routes. The dominant-negative mutant RabF2b S24N (Kotzer et al., 2004) impairs the vacuolar targeting of proteins by affecting the trafficking between the GA and the PVC, causing the vacuolar marker Aleu-GFP to be secreted to the apoplast (Figure 4, column 4, and S1d). Repeating the previous co-expression experiments with this negative mutant, it was clear that the C-ter-containing proteins (cardosin A-mCherry, cardosin AΔPSI-mCherry, SP-mCherry-C-ter and SP-PSI-mCherry-C-ter) were secreted to the apoplast (Figure 4h,t,x), and may also be retained in dot-shaped compartments (Figure 4d), but it never reached the vacuole. By contrast, proteins driven by PSI alone (i.e. cardosin AΔC-ter and SP-PSI-mCherry, respectively; Figure 4l,p) were still accumulating in the vacuole, meaning that the route to the vacuole was independent from the Golgi–PVC route. Again, in these later experiments, the C-ter VSD appears to be dominant over the PSI, as a chimaera bearing the two VSDs always behaves as a protein carrying the C-terminal alone.
These results confirmed that cardosin A carries two VSDs, which may specify two different trafficking pathways, with the C-terminal VSD being dominant over the PSI VSD. Therefore, we can conclude that the cardosin A-mCherry protein in this particular heterologous system and in this type of tissue will use a ‘conventional’ ER–GA pathway to reach the vacuole. The next step was to question whether the cardosin A VSDs are ubiquitous for vacuolar proteins and aspartic proteinases. Further insight into the sequence of each determinant may help in classification, and in understanding their specificities or mechanisms of action.
Is the C-terminal domain of cardosin A a ubiquitous sequence for vacuolar proteins?
When comparing the C-terminal sequence VGFAEAA of cardosin A with the same sequence from other APs, and with Hordeum vulgare (barley) lectin VSD (Figure 5), we recognized that four amino acids, FAEA, were conserved among them. We wondered whether this conserved short peptide was enough to target the protein to the vacuole. Two novel chimeric proteins were made, one with the short peptide FAEA at the C terminus of mCherry, and the other with the same peptide at the N terminus (Figure 1d, SP-mCherry-FAEA and SP-FAEA-mCherry).
The short peptide FAEA was indeed sufficient to target mCherry to the vacuole (Figure 6a,b). Interestingly, when the C-terminal peptide or the short peptide FAEA were placed between the SP and mCherry (SP-C-ter-mCherry and SP-FAEA-mCherry, see Figure 1d), the fluorescent proteins were not delivered to the vacuole (Figure 6c–f, respectively), confirming that the C-ter peptide or its shortened version acts as true ctVSD, and belongs to the ctVSDs group described for several vacuolar proteins.
Cardosin A PSI domain: a ubiquitous VSD for aspartic proteinases?
The PSI domains are embedded in several (but not all) aspartic proteinase sequences. To go further in the understanding of PSI domains as key structures in AP biology, we compared the vacuolar targeting efficiency of two PSI domains issued from two different proteins. Along with cardosin A, cardosin B is another AP from Cynara cardunculus that shares high similarity with cardosin A. Regarding the PSI domains, cardosin B and cardosin A share about 70% similarity in terms of nucleotide sequence, but they have a distinguishable feature: the PSI domain in cardosin B has a putative glycosylation site that is absent from the PSI domain of cardosin A. On the other hand, the C-terminal peptide differs in only one amino acid between the two cardosins. It has also been previously shown that cardosin B accumulates in the vacuole when heterologously expressed in tobacco leaves (da Costa et al., 2010). We took advantage of these characteristics to explore the specificity of PSI and C-terminal domains in the vacuolar targeting of cardosins as a whole, to evaluate whether they share the same sorting characteristics. The PSI and C-terminal domains of cardosin B were therefore isolated and fused with mCherry to obtain the fluorescent chimaera SP-PSIB-mCherry and SP-mCherry-C-terB (Figure 1c).
Both fluorescent proteins were detected in the vacuole of N. tabacum cells (Figure 7a,b,e,f, respectively), indicating that cardosin B PSI and cardosin B C-terminal domains act as vacuolar sorting domains for cardosin B, as observed for cardosin A.
Regarding the two C-terminal domains, they react with the same sensitivity as the blockage experiments. When the dominant-negative versions of RabD2a and Sar1 were used to block the trafficking between the ER and GA, the mCherry protein driven by the C-terminal region of cardosin B was retained in the ER (Figure 7c,d), reminiscent of the observations made using the C-terminal VSD of cardosin A, and suggesting a COPII-dependent vacuolar trafficking pathway. This result reinforces the ubiquity of the C-terminal domain of cardosins acting as VSD, as already mentioned.
By contrast, a strong difference was observed between the two PSI domains. The fluorescent proteins driven by the glycosylated PSI domain of cardosin B remained blocked in the ER when co-expressed with the dominant-negative mutant of RabD2a or Sar1 (Figure 7g,h), suggesting the involvement of an ER–Golgi-mediated pathway for these proteins. These observations contrast with those obtained when using the non-glycosylated PSI domain of cardosin A, where the fluorescent chimaerae continued to reach the vacuole, despite an ER–GA blockage.
These results confirm, therefore, that PSI domains of aspartic proteinases may act as VSDs; however, they highlight that PSI domains may mediate different routes according to their structure. Taking into account the fact that the cardosin-B PSI domain differs from the cardosin-A PSI domain by the presence of a glycosylation site, these results highlight a putative role for glycosylation in determining the route to the vacuole.
Our results: (i) identified two vacuolar sorting determinants for cardosin A, PSI and the C-terminal peptide VGFAEAA/FAEA, (ii) revealed two different vacuolar routes for the CdA-mCherry reporter proteins, and (iii) provided new insight into the PSI domains in APs. It must be kept in mind that the trafficking events described in our experimental model (tobacco leaf epidermal cells with one large lytic vacuole) may not reflect the in planta situation, where different types of vacuoles may co-exist; however, these data still contribute to shedding new light on the AP vacuolar trafficking pathways in plant cells, as discussed below.
Two VSDs on one protein: an original way to regulate vacuolar traffic in plants
The fact that cardosin A carries two types of VSDs raises questions about the sorting efficiency of a single VSD, their ability to act independently or whether there is a need for synergy between the two. In the latter hypothesis, the two regions could be part of a larger vacuolar signal, such as a physical structure vacuolar sorting determinant (psVSD). psVSDs are composed of several motifs distributed along the protein sequence, which are only functional in the mature, correctly folded polypeptide (Jolliffe et al., 2003). Therefore, the PSI and the C terminus of cardosin A could act cooperatively. It is also possible that the vacuolar sorting may benefit from the cumulative effects of the sorting signals, as suggested for two different storage proteins (Holkeri and Vitale, 2001; Nishizawa et al., 2006). Our data, however, do not favour the hypothesis of such a cooperative process. Instead, they suggest a putative redundancy between the two signals, as the presence of only one of them was sufficient for correct vacuolar targeting. Each signal must therefore be considered as a true distinct vacuolar sorting determinant for cardosin A. Our data show that they may act independently. The presence of such VSDs on one protein would then ensure a ‘poly-sorting’ ability, conferring the APs with several functions, according to their final targeting, and controlling vacuolar traffic according to cell constraints or need through a specific PSI signalling pathway, which may operate in parallel with another vacuolar pathway.
Moreover, our results have shown that, in our experimental model, the C-terminal signal is dominant over the PSI signal, as it dictates the route the protein takes to the vacuole. Searching for receptors of the C-terminal VSD signals in different parts of the plants, and related developmental stages, would help us reveal whether such dominance is achieved in all plant cells, or whether PSI acts on its own as soon as the mechanisms for C-terminal recognition are missing or saturated. These data provide exciting possibilities to solve previous contradictions on the role of the PSI as vacuolar sorting determinant (Faro et al., 1999; Kervinen et al., 1999; Brodelius et al., 2005). The structure and function of the PSI domain may indeed be a clue in understanding the complexity of AP targeting in plant cells.
Two VSDs for two pathways: specificity for cardosin-A traffic?
Our data outline a complex regulation for the trafficking of cardosin A-mCherry to the vacuoles: the C terminal drives the cardosin A-mCherry protein to the lytic vacuole via a COPII-mediated ER–Golgi–PVC pathway, whereas PSI-mediated targeting may bypass the Golgi and PVC.
These results are in contrast with those obtained for cardosin B (da Costa et al., 2010) and phytepsin (Törmäkangas et al., 2001), where the PSI domain was essential to enter a COPII pathway. This apparent discrepancy may be related to the different functions of cardosins, and may stem from differences in the composition of the PSI domain: cardosin B and phytepsin have a conserved glycosylation site in the PSI domain, which is absent from cardosin A.
The presence or the absence of a glycosylation site on the PSI domain can therefore be a key structure in determining the route to be taken by PSI-driven targeting.
Glycosylation events may influence sorting mediated by AP PSI domains
The potential role of glycosylation in determining a vacuolar route, bypassing or not the Golgi, has been previously discussed (Rayon et al. 1998; Paris et al. 2010). Previous studies have suggested that the glycosylation process was not required for vacuolar trafficking, but may indirectly impair the regulation of vacuolar trafficking by affecting the processing and transport rates of the cargo proteins (Wilkins et al., 1990). Our results, based on the glycosylation status of the PSI domain, indeed outline such a potential effect of glycosylation on vacuolar transport. In contrast with other plant APs, such as cardosin B, there is no glycosylation site in the cardosin-A PSI domain (Frazão et al. 1999). We have shown that replacing the PSI domain of cardosin A (wich has no glycosylation site) with the PSI domain of cardosin B (which has a glycosylated site) changes the PSI mechanisms of vacuolar sorting, as it shifted from a COPII-independent trafficking pathway to a COPII-dependent trafficking pathway. The presence or the absence of a glycosylation site on the PSI domain therefore appears to be a key structure in determining the route to be taken by the PSI-driven targeting. This hypothesis needs further investigation. The use of point-mutation mutants to remove or add the glycosylation sites in cardosin-A and cardosin-B PSIs will open the debate on the role of glycosylation on protein sorting and trafficking.
A typical C-terminal VSD versus an unconventional PSI signal: new insight into the regulatory function of the PSI domain in aspartic proteinase trafficking
The C-terminal signal of cardosin A or cardosin B falls into the well-known ctVSD category (Neuhaus et al., 1991; Frigerio et al., 1998; Zouhar and Rojo, 2009). The short peptide FAEA, which is common to other protein VSDs (Dombrowski et al., 1993), may even play a key role in the recognition mechanism needed for protein sorting along the secretory pathway to the lytic vacuoles.
The PSI domain, however, does not meet any of the usual criteria defining a VSD. It cannot be considered as a ssVSD in the strict sense of the word, as the PSI amino acid sequence is not conserved; the psVSD definition does not apply either, as only a single region of the protein is involved (Jolliffe et al., 2003; Zouhar and Rojo, 2009). Therefore, the PSI domain would define a new class of VSD. On the other hand, not all PSI domains of other aspartic proteinases have been shown to play an essential role in vacuolar sorting. They have been reported to act as a targeting signal to the lytic vacuoles for several APs [phytepsin, Glycine max (soya bean) and cardosin A; Törmäkangas et al., 2001; Terauchi et al., 2006], but not to the storage vacuole (Terauchi et al., 2006). PSI may therefore act as a vacuolar signal only in specific conditions or developmental stages. Comparison with other PSI domains would help to define its structural characteristics. In all cases, the PSI domain appears to be a key structure for regulating the vacuolar trafficking of proteins according to cell needs.
A working model for cardosin A and AP trafficking
Gathering the data proposed for AP trafficking mechanisms to the vacuole, a working model may be proposed (Figure 8). C-terminal mediated sorting would correspond to a COPII-dependent ER–Golgi pathway to the vacuole, and may be favoured for specific tissues and developmental stages (Figure 8 arrow 1). Similarly, a glycosylated PSI-mediated targeting dependent on COPII vesicles (Figure 8, arrow 1) has already been described for phytepsin (Törmäkangas et al., 2001), and may concern other PSIs containing APs, such as cardosin B. The data presented here point to a PVC–vacuolar route for cardosin A and C-terminal-mediated sorting (Figure 8, arrow 2), but it cannot be discarded that in other tissues/organs cardosin A could travel directly from the GA, in PAC vesicles, to reach the vacuole (Figure 8, arrow 2′). An alternative, COPII-independent, vacuolar route mediated by non-glycosylated PSI domains (Figure 8, arrow 3) must be considered. Furthermore, it was already shown that cardosin A, in cardoon seeds, reaches the protein body (PB) in a GA-independent manner (Figure 8, arrow 3′).
Regarding cardosin A, in planta, the dual localization of cardosin A in protein bodies in cardoon cotyledon cells, and in protein storage vacuoles in flower cells (Pereira et al., 2008), could be related to the existence of a PSI-specific pathway, dependent on the type of tissue. In fact, in cardoon, it was shown that cardosin A accumulates in protein bodies in a Golgi-independent manner (Pereira et al., 2008). Our hypothesis is that the PSI-mediated pathway is more relevant in metabolically active organs (such as flowers and seeds), where protein storage vacuoles are predominant over lytic vacuoles, and cells have different needs.
In conclusion, these data, focusing on the processes of vacuolar trafficking for APs, outline the amazing potential for proteins in plants carrying two VSDs, which may help to regulate distinct vacuolar routes according to cell needs.
Production of cardosin-A fluorescent proteins
Cardosin A and all of the chimeric proteins used in this study [cardosin AΔPSI (a deletion of the PSI region); cardosin AΔC-ter (a deletion of the C-terminal peptide VGFAEAA); and cardosin AΔ[PSI+C-ter] (a deletion of both the PSI region and C-terminal peptide); see Figure 1], were generated by PCR using specific primers containing a XbaI recognition site (forward primer, underlined), and a SacI recognition site (reverse primers, underlined) [fwd primer for cardosin A, cardosin AΔC-ter and cardosin AΔ(PSI+C-ter), 5′-TCTAGAGCCGCCACCATGGGTACCT-3′; fwd primer for cardosin AΔPSI, 5′-TGCTGCAGGTGGTACTTCATCTGAAGAATTAC-3′; rev primer for cardosin A, 5′-CTGAGCTCTCAAGCTGCTTCTGCAAATC-3′; rev primer for cardosin AΔPSI, 5′-ACCTGCAGCACCACCCCCGTTAGCGCCAATTG-3′; rev primer for cardosin AΔC-ter and cardosin AΔ(PSI+C-ter), 5′-GAGCTCTAGTAAATTGCCATAATC-3′]. Unmodified cardosin A (Duarte et al., 2008) was used as the template for all PCR reactions. To generate the double mutant, the cardosin AΔPSI cDNA was used. PCR fragments were cloned using the Zero Blunt® cloning kit (Invitrogen, http://www.invitrogen.com) and analysed by restriction mapping. They were sequenced with M13 primers [M13uni(-21) and M13rev(-29); Eurofins MWG Operon, http://www.eurofinsdna.com/home.html). Fragments were inserted into the XbaI and SacI sites of the binary vector pVKH18-En6 (Sparkes et al., 2006), under the action of a CaMV 35S promoter. To obtain fusions with the red fluorescent protein mCherry (Shaner et al., 2004; Figure 1), mCherry was PCR-amplified using specific primers (fwd primer, 5′-CAGTCGACAATGGTGAGCAAGGGCGAG-3′; rev primer, 5′-CTGAGCTCGGATCCTTACTTGTACAGCTCGTCCATGC-3′; rev primer to introduce cardosin A C-terminal sequence (in bold), 5′-ACGAGCTCTCAAGCTGCTTCTGCAAATCCAACCTTGTACAGCTCGTCCAT-3′), introducing a SalI recognition site into the 5′ end, and SacI (underlined) and BamHI (dashed) sites into the 3′ end. The PCR fragments were sequenced and cloned into the vectors containing the cardosin A chimeric proteins to obtain in-frame fusions.
mCherry chimeric proteins with putative sorting determinants
The PSI domain was isolated from cardosin-A cDNA by PCR using specific primers designed for the PSI borders, and to allow the introduction of SalI recognition sites (underlined) (fwd primer, 5′-ACGTCGACTGTCATGAACCAGCAATGCAA-3′; rev primer, 5′-ACGTCGACGCACCACCTGCAGCACCACCGGATAAGTGTTCACACAACTCG-3′), and was cloned into the pVKH18-En6-SP-mCherry at the 5′ end of mCherry. To isolate cardosin B PSI the strategy was the same and the primers used were fwd, 5′-ACGTCGACTTTAAACCAACAATGCAAAACATTGG-3′, and rev, 5′-ACGTCGACGCACCACCTGCAGCACCACCTTCTGCACTTGAAGTGGGTA-3′. To introduce the cardosin A/B C-terminal peptide sequence (VGFAEAA and FAEA) and the VSD of barley lectin into the 3′ end of mCherry, the motif was amplified by PCR with modified reverse primers [rev primer for VGFAEAA (cardosin A), 5′-ACGAGCTCTCAAGCTGCTTCTGCAAATCCAACCTTGTACAGCTCGTCCAT-3′; rev primer for VGFAEAV (cardosin B): 5′-AAGAGCTCTCAAACTGCTTCTGCAAATCCAACCTTGTACAGCTCGTCCAT-3′; rev primer for FAEA, 5′-AAGAGCTCTCATGCTTCTGCAAACTTGTACAGCTCGTCCAT-3′; rev primer for barley lectin VSD, 5′-AAGAGCTCTCAGGCGATGGCCTCGGCGAAGACCTTGTACAGCTCGTTCCAT-3′], introducing the C-terminal peptide sequences (bold) and a SacI recognition site (underlined). The primer designed for mCherry amplification at the 5′ end (5′-CAGTCGACAATGGTGAGCAAGGGCGAG-3′) was used as the forward primer. For the introduction of C-terminal peptides VGFAEAA and FAEA in the N terminus of mCherry, the strategy was the same, but the C-terminal sequence was introduced in the forward primer (fwd primer for VGFAEAA, 5′-CAGTCGACGTTGGATTTGCAGAAGCAGCGATGGTGAGCAAGGGCGAG-3′; fwd primer for FAEA, 5′-CAGTCGACTTTGCAGAAGCAATGGTGAGCAAGGGCGAG-3′; rev primer, 5′-CTGAGCTCGGATCCTTACTTGTACAGCTCGTCCATGC-3′; SalI recognition sites are underlined and the Bam HI recognition site is represented with a dashed line, with the introduced sequences set in bold). To obtain mCherry with both domains, the isolated PSI was cloned into SalI sites in the fluorescent protein SP-mCherry-C-ter. These modified mCherry cDNAs were cloned into pVKH18-En6-SP-mRFP under the action of a CaMV 35S promoter, as described earlier.
Transient expression in tobacco leaves
All transformation experiments were repeated at least three times using different plants. A selection of representative data is shown. Agrobacterium tumefaciens (strain GV3101::pMP90) was electro-transformed (25 μF, 2.5 kV and 200 Ω; Bio-Rad Gene Pulser Electroporation System, http://www.bio-rad.com) with each fluorescent chimaera, and the clones were screened by restriction mapping. Positive clones were selected for agro-infiltration of N. tabacum L. cv. Petit Havana SR1, and transient expression was carried out as described by Batoko et al. (2000), with the following modification: YEB-medium was replaced by LB medium supplemented with 25 g ml−1 kanamycin. For experiments requiring co-infection of more than one protein, bacterial strains containing the cDNAs were mixed before infiltration, with the titre of each Agrobacterium strain adjusted to the required OD600. Nicotiana tabacum plants were kept under a 16-h light photoperiod at all times.
Unless otherwise stated, transfected leaves were imaged 3 days after infiltration. Experiments were repeated more than three times for each fluorescent chimaera.
Co-expression of RabD2a N121I, SarI H74L and RabF2b N24S
Cardosin A, cardosin A versions and mutants, cardosin B versions and controls were infiltrated alone (control) and with the dominant-negative mutants RabD2a N121I (Batoko et al., 2000), SarI H74L (Andreeva et al., 2000) or RabF2b N24S (Kotzer et al., 2004) in tobacco leaves. Aleurain-GFP was used as a control. Three days after tobacco leaf infiltration, cells were imaged.
Images were acquired with an inverted SP2 Leica laser scanning microscope. Pieces of leaf were sampled from the infiltrated area in a random fashion, and mounted in water. The 561-nm laser line was used for the excitation of mCherry, whereas the 488-nm line was used for exciting GFP. In co-expression experiments, mCherry and GFP were scanned sequentially with the multitrack scanning mode. Images shown in the document are representative of several observations of the same protein/situation. Images were processed with lcs 2.61 build 1538 (Leica, http://www.leica.com). Quantitative analysis was performed on the results from the observation of 80–150 cells from three independent experiments. For quantification it was considered that the total number of cells with fluorescent signal will define a 100% value and, among this population, the different localization patterns were then scored.
RabD2a N121I, SarI H74L, RabF2b N24S, GFP-HDEL, ST-GFP and Aleurain-GFP were kindly provided by Dr Ian Moore (University of Oxford). The authors also acknowledge the support of the Cell Biology Pole facilities of Imagif (http://www.imagif.cnrs.fr, CNRS, France). C.P. was the recipient of a PhD grant (SFRH/BD/37201/2007) supported by the Portuguese Science and Technology Foundation (FCT), and benefitted from a joint PhD programme at the University of Porto (Portugal) and University Paris-Sud (France). The strategic project (PEst-OE/BIA/UI4046/2011) of the Center for Biodiversity, Functional & Integrative Genomics (BioFIG) for 2011–2012 is supported by FCT funding.