OPEN ACCESS ARTICLE

This Article Free via Open Access: OA OA Abstract Full Text (PDF) Supplemental Data OA All Versions of this Article: 18/11/3182    most recent tpc.105.036400v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (6) Citing Articles via Google Scholar Google Scholar Articles by Saint-Jore-Dupas, C. Articles by Gomord, V. Search for Related Content PubMed PubMed Citation Articles by Saint-Jore-Dupas, C. Articles by Gomord, V. Agricola Articles by Saint-Jore-Dupas, C. Articles by Gomord, V.

Plant N-Glycan Processing Enzymes Employ Different Targeting Mechanisms for Their Spatial Arrangement along the Secretory Pathway^[W],[OA]

^a Centre National de la Recherche Scientifique, Unité Mixte de Recherche 6037, IFRMP 23, GDR 2590, UFR des Sciences, Université de Rouen, 76821 Mont-Saint-Aignan Cedex, France
^b Department of Biochemistry, Cellular, and Molecular Biology, University of Tennessee, Knoxville, Tennessee 37996-0840
^c School of Biological and Molecular Sciences, Oxford Brookes University, Oxford OX3 0BP, United Kingdom

¹ To whom correspondence should be addressed. E-mail vgomord{at}crihan.fr; fax 33-2-35-14-67-87.

	ABSTRACT

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

 
The processing of N-linked oligosaccharides in the secretory pathway requires the sequential action of a number of glycosidases and glycosyltransferases. We studied the spatial distribution of several type II membrane-bound enzymes from Glycine max, Arabidopsis thaliana, and Nicotiana tabacum. Glucosidase I (GCSI) localized to the endoplasmic reticulum (ER), -1,2 mannosidase I (ManI) and N-acetylglucosaminyltransferase I (GNTI) both targeted to the ER and Golgi, and ß-1,2 xylosyltransferase localized exclusively to Golgi stacks, corresponding to the order of expected function. ManI deletion constructs revealed that the ManI transmembrane domain (TMD) contains all necessary targeting information. Likewise, GNTI truncations showed that this could apply to other type II enzymes. A green fluorescent protein chimera with ManI TMD, lengthened by duplicating its last seven amino acids, localized exclusively to the Golgi and colocalized with a trans-Golgi marker (ST52-mRFP), suggesting roles for protein–lipid interactions in ManI targeting. However, the TMD lengths of other plant glycosylation enzymes indicate that this mechanism cannot apply to all enzymes in the pathway. In fact, removal of the first 11 amino acids of the GCSI cytoplasmic tail resulted in relocalization from the ER to the Golgi, suggesting a targeting mechanism relying on protein–protein interactions. We conclude that the localization of N-glycan processing enzymes corresponds to an assembly line in the early secretory pathway and depends on both TMD length and signals in the cytoplasmic tail.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

 
The plant Golgi apparatus consists of individual stacks of membrane bound flattened cisternae surrounded by small vesicles that are dispersed throughout the cytoplasm (Staehelin and Moore, 1995; Driouich and Staehelin, 1997; Andreeva et al., 1998; Dupree and Sherrier, 1998). The number of Golgi stacks per cell and the number of cisternae per stack vary with the plant species and the cell type, reflecting the physiological conditions, the developmental stage, and the functional requirements of a plant cell (reviewed in Staehelin and Moore, 1995; Andreeva et al., 1998). Despite these variations, each individual Golgi stack can be described as a polarized structure with its cisternal morphology and its enzymatic activities changing gradually from the endoplasmic reticulum (ER)-adjacent cis-face to the trans-face (Fitchette et al., 1999).

The Golgi apparatus is not only a central sorting point of the secretory pathway (Nebenführ, 2002; Hawes, 2005; Hawes and Satiat-Jeunemaitre, 2005), but this compartment also provides the central biosynthetic function of a complex carbohydrate factory (Driouich and Staehelin, 1997; Nebenführ and Staehelin, 2001). Thus, the Golgi is the site of maturation of N-glycans on glycoproteins, O-glycosylation of Hyp-rich glycoproteins and arabinogalactan proteins, and de novo synthesis of cell wall matrix polysaccharides, such as hemicelluloses and pectins. These biosynthetic processes are performed by glycosidases and glycosyltransferases (Lerouge et al., 1998; Saint-Jore-Dupas et al., 2004).

During the maturation of N-glycans, ER and Golgi-localized glycosidases trim sugar residues from an oligosaccharide precursor of N-glycans to a Man₅GlcNAc₂ structure. Subsequently, Golgi-localized glycosyltransferases mediate the transfer of sugar residues from nucleotide sugar donors onto the N-glycans. In mammalian cells, it has been shown that glycosidases and glycosyltransferases are distributed along the Golgi from the cis- to the trans-regions in the order in which they process N-glycans; however, there is some overlap and variation in their distribution depending on the cell type (Roth, 1991; Nilsson et al., 1993a; Velasco et al., 1993; Rabouille et al., 1995; Idgoura et al., 1999). With the exception of -glucosidase II (Trombetta et al., 1996, 2001; P. Soussilane, C. Saint-Jore-Dupas, and V. Gomord, unpublished data), which is a soluble heterodimer, glycosidases and glycosyltransferases responsible for the N-glycan maturation are type II membrane proteins with a short N-terminal cytoplasmic tail (CT), a single transmembrane domain (TMD), a stem region (S), and a large catalytic domain associated with the enzyme activity. In mammalian cells, Golgi retention of glycosyltransferases has been attributed to the TMD with variable contributions of the cytoplasmic and lumenal domains depending on the protein (reviewed in Colley, 1997; Gleeson, 1998; Munro, 1998; Zerfaoui et al., 2002). Various models have been proposed to explain the retention of glycosyltransferases in the mammalian Golgi. The kin recognition model proposes that the resident proteins form large hetero-oligomers by protein–protein interaction in the Golgi membrane and are consequently prevented from entering the budding vesicles destined to another compartment (Nilsson et al., 1993b, 1994, 1996; Opat et al., 2000; Qian et al., 2001). According to the lipid bilayer model, the fit between the length of TMD and the thickness of the lipid bilayer of each organelle membrane determines the localization because the membrane of each organelle has its specific lipid components (Munro, 1991, 1995a; Bretscher and Munro, 1993). The third model proposes that the CT of glycosyltransferases plays a key role in their specific localization to the Golgi (Osman et al., 1996; Milland et al., 2001, 2002).

This article illustrates an attempt at elucidating the targeting domain(s), allowing the selective retention of glycosyltransferases in certain cisternae of the plant Golgi apparatus. The main reason for the absence of data available in this field is that the first plant Golgi enzymes were cloned only recently (Essl et al., 1999; Nebenführ et al., 1999; Strasser et al., 1999, 2006; Boisson et al., 2001; Dirnberger et al., 2002; Pagny et al., 2003). To date, only 10 cDNA encoding Arabidopsis thaliana enzymes involved in N-glycan maturation and eight cDNA involved in cell wall polysaccharides biosynthesis have been cloned and the corresponding protein characterized, while 420 genes encoding enzymes implicated in glycoconjugate biosynthesis and potentially located in the Golgi have been identified in the Arabidopsis genome (Coutinho et al., 2003). This also explains why the first evidence for a subcompartmentation of N-glycosylation enzymes in the plant Golgi was obtained using an indirect approach based on immunolocalization of glycosyltransferase products (Lainé et al., 1991; Fitchette et al., 1994, 1999). In this way, it was shown that ß-1,2-xylose is added onto plant N-glycans mainly in the medial Golgi, whereas -1,3-fucosylation occurs predominantly in the trans-side of Golgi (Fitchette et al., 1994). Moreover, the biosynthesis of Lewis-a antenna constitutive of plant complex N-glycans is a late Golgi event taking place in the trans-Golgi and the trans-Golgi network (Fitchette et al., 1999). A similar method was also used to show subcompartmentation within the Golgi of glycosyltransferases involved in the assembly of cell wall polysaccharides (Zhang and Staehelin, 1992). Finally, another indirect approach based on the expression of heterologous glycosyltransferases, such as rat -2,6-sialyltransferase (-2,6-ST), in plant cells has provided data suggesting that similar subcompartmentation signals might be conserved between plant and animal Golgi type II membrane proteins (Boevink et al., 1998; Wee et al., 1998).

With the recent cloning of plant glycosyltransferases, data on their localization were obtained by expression of these enzymes as fusions with a green fluorescent protein (GFP). For instance, it was shown that the N-terminal 77 amino acids of the tobacco (Nicotiana tabacum) N-acetylglucosaminyltransferase I (GNTI; Essl et al., 1999) or the N-terminal 92 amino acids of the Arabidopsis mannosidase II (MANII; Strasser et al., 2006), including the CT and TMD of both maturation enzymes, are sufficient to target GFP to the Golgi of Nicotiana benthamiana leaf epidermal cells. Similarly, the N-terminal 35 amino acids of Arabidopsis ß-1,2-xylosyltansferase (XYLT) were fused to GFP and the recombinant protein expressed in tobacco BY-2 suspension-cultured cells (Dirnberger et al., 2002; Pagny et al., 2003). In both studies, it was proposed that the CT and the TMD contain the information sufficient for Golgi targeting, and the fusion protein was shown to be concentrated in the medial Golgi cisternae (Pagny et al., 2003).

All studies mentioned above point to a central role of the CT and TMD for proper localization of N-glycan processing enzymes; however, no systematic study of a type II membrane protein targeting in the early secretory pathway of plant cells has been published to date. Herein, we demonstrate that TMD length is a major determinant for localization of soybean (Glycine max) -1,2 mannosidase I (ManI) to the cis-half of the Golgi and to the ER and that an increase in the length of its TMD relocates this type II membrane protein further downstream to the trans-half of the Golgi consistent with the lipid bilayer thickness model. By contrast, in silico analyses performed on cloned glycosylation enzymes illustrate that the TMD length is not sufficient by itself to explain the compartmentation of all N-glycan processing enzymes as illustrated here for glucosidase I (GCSI) where the CT is crucial for ER retention.

	RESULTS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

 
Early Golgi Type II Membrane Proteins Are Partially Located in the ER
To better understand the mechanisms allowing the selective retention of N-glycan processing enzymes in the early Golgi compartments, the localization of a series of GFP fusions to four different members of the N-glycan processing machinery (GCSI, ManI, GNTI, and XYLT; Figure 1 ) was studied after stable expression in tobacco BY-2 cells. Fluorescence of a full-length ManI-GFP fusion construct was detected by confocal laser scanning microscopy in small bodies (Figures 2A and 2B) that moved through the cytoplasm as it has been described previously for this construct in another independent cell line (Nebenführ et al., 1999). In addition, a substantial fluorescence signal was observed in a reticulate network throughout the cytoplasm that was indistinguishable from the ER network stained by a GFP-HDEL construct (Figures 2D and 2E). To check if ER labeling was due to overexpression of the recombinant proteins, BY-2 cells expressing ManI-GFP were incubated with the protein synthesis inhibitor cycloheximide. After 2 h of treatment, the targeting pattern remained unchanged, showing that the steady state location of ManI-GFP is the Golgi and the ER (cf. Figures 2B and 2C).

View larger version (53K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 1. Schematic Representation of Fusion Proteins Analyzed in This Study. GCSI-GFP: full-length Arabidopsis GCSI fused to GFP. GCS90-GFP: the first 90 N-terminal amino acids of GCSI fused to GFP. 13GCS90-GFP: GCS90-GFP minus the first 13 N-terminal amino acids. ManI-GFP: full-length G. max ManI fused to GFP. 19Man-GFP: ManI-GFP minus the first 19 N-terminal amino acids. Man99-GFP and Man49-GFP: the first 99 and 49 amino acids of ManI, respectively, fused to GFP. 19Man49-GFP: Man49-GFP minus the first 19 N-terminal amino acids. CTMan49-GFP: the whole CT was deleted from Man49-GFP. MAAAMan49-GFP: the CT of Man49-GFP was replaced by an artificial CT containing three Ala residues. ManTMD23-GFP and Man99TMD23-GFP: ManI-GFP and Man99-GFP, respectively, where the TMD was lengthened from 16 to 23 amino acids. GNTI-GFP: full-length N. tabacum GNTI fused to GFP. GNT38-GFP: the first 38 N-terminal amino acids of GNTI fused to GFP. XylT-GFP: full-length Arabidopsis XYLT fused to GFP. XylT35-GFP: the first 35 amino acids of XYLT fused to GFP. ST52-GFP/mRFP: the first 52 amino acids of a rat -2,6-ST fused to mRFP. GFP-HDEL: a GFP version containing the sporamine signal peptide and a C-terminal HDEL ER retention sequence.

View larger version (121K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 2. ManI-GFP and GNTI-GFP Are Located to the Golgi and to the ER, whereas GCSI-GFP Is Exclusively Accumulated in the ER and XYLT-GFP in the Golgi. Transgenic BY-2 tobacco cell lines were analyzed 3 to 4 d after subculturing. Bars = 8 µm. (A) and (B) ManI-GFP is located to the Golgi and to the ER ([A], cortical view; [B], cross section where ER labeling around the nucleus and in the vacuolar strands is characteristic). (C) After a 2-h treatment with the protein synthesis inhibitor cycloheximide, ER and Golgi labeling observed with ManI-GFP fusion remained unchanged, showing that the steady state localization of ManI-GFP is the Golgi and the ER. (D) and (E) GFP-HDEL highlights the ER. (F) After a 2-h treatment with BFA (50 mg·mL–1), fluorescent Golgi stacks have disappeared, while ER fluorescence is increased due to the relocation of ManI-GFP. (G) GCSI-GFP is an ER resident membrane protein and shows a similar fluorescence pattern as GFP-HDEL (D). (H) GNTI-GFP is targeted to the Golgi and to the ER as observed for ManI-GFP (B). (I) XYLT-GFP accumulates exclusively in Golgi stacks.

To compare the location of ManI to one of the other plant N-glycosylation enzymes in the secretory pathway, we analyzed under the same conditions the subcellular localization of N-glycan maturation enzymes acting before, just after ManI, or much later. The first enzyme we studied was Arabidopsis GCSI. This type II membrane protein trims the first sugar residue from the precursor oligosaccharide in the ER immediately after its attachment to the nascent glycoprotein (see a schematic representation of plant N-glycan maturation in Figure 11 and additional data in Supplemental Table 1 online). The full-length protein (Boisson et al., 2001) was fused to GFP and, consistent with what was shown for human GCSI in COS cells (Hardt et al., 2003), the fusion protein was exclusively located in the ER in BY-2 cells (Figure 2G).

View larger version (34K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 11. Comparison of CT and TMD Length for Plant N-Glycosylation Enzymes. (A) Processing of N-linked glycans in the ER and Golgi apparatus of plant cells. A precursor oligosaccharide (Glc3Man9GlcNAc2) assembled onto a lipid carrier is transferred on specific Asn residues of the nascent polypeptide (P). This precursor is then modified by glycosidases and glycosyltransferases mainly in the ER and the Golgi apparatus during the transport of the glycoprotein downstream the secretory pathway. Glycosidases and glycosyltransferases responsible for plant N-glycan maturation and their localization in the biochemical processing pathway are indicated. N-glycan processing enzymes whose intracellular localization has been studied to date (in boldface characters; this study; Pagny et al., 2003; Strasser et al., 2006; C. Saint-Jore-Dupas, M.C. Kiefer-Meyer, and V. Gomord, unpublished data) confirm the assembly line concept in that their position within the organelles of the secretory pathway (at the right) mirrors their position in the biochemical pathway (at the left). (B) CT and TMD length of N-glycan processing enzymes that have been cloned from different plant species. Accession numbers are indicated at the right of each schematic representation. For each membrane protein, the position and the size of the TMD were estimated from TmHMM_v2 software (http://www.cbs.dtu.dk/services/TMHMM/). For some of the proteins, the probability to define the position of the TMD is below 50% (//). Boxes outlined in bold correspond to N-glycan processing enzymes whose intracellular localization has been studied to date by confocal and/or electronic microscopy. GNTII, N-acetylglucosaminyltransferase II; ß1,2XYLT, ß1,2 xylosyltransferase; 1,3 FucT, 1,3 fucosyltransferase; 1,4 FucT, 1,4 fucosyltransferase; ß1,3 GalT, ß1,3 galactosyltransferase.

Finally, the third candidate, XYLT from Arabidopsis, was located in the Golgi only (Figure 2I), confirming the results from Pagny et al. (2003), who demonstrated that the N-terminal end of this enzyme targets GFP to a medial subset of cisternae of Golgi stacks.

To ascertain whether protein expression levels might alter localization of our fusion proteins, we have confirmed these results in different independent cell lines expressing the fusion proteins, at least 3 months after transformation. Imaging of cells was always performed on the third or fourth day after subculturing, which corresponds to the optimal growth phase under our culture conditions. Nevertheless, to further validate that ER labeling was not due to overexpression of the fusion, we have controlled for each fusion that the labeling pattern was unchanged after a 2-h treatment with cycloheximide. Additional data illustrating that translation is indeed blocked under these conditions are available online. Protein gel blots revealed with anti-GFP antibodies and the enhanced chemiluminescence staining have shown a very low signal over background for the recombinant proteins, indicating a low level of expression for all fusion proteins in this study (data not shown). Further evidence for a level of fusion protein expression compatible with a functional nonsaturated secretory pathway was obtained from coexpression experiments when ManI-GFP is located both in the ER and Golgi in the same cell, while a Golgi marker (ST52-mRFP) is found exclusively in the Golgi (Figures 4A to 4C). All together, results obtained under these carefully controlled conditions clearly show that N-glycosylation enzymes are targeted specifically to the ER (GCSI) or to the Golgi (XYLT) exclusively, but some enzymes have a dual steady state location in both organelles, as is the case for ManI, GNTI (this study), and other membrane proteins, such as prolyl 4-hydroxylase (Yuasa et al., 2005) and ERD2 (Boevink et al., 1998; Saint-Jore et al., 2002). In a next step, we decided to investigate which signals were responsible for the targeting of the population of glycosylation enzymes showing a dual steady state distribution Golgi/ER.

View larger version (102K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 4. ManI and GNTI Fusions Accumulate into the Early Golgi. Transgenic cell lines coexpressing ST52-mRFP ([B], [E], [H], and [K]) and ManI-GFP (A), Man99-GFP (D), Man49-GFP (G), or GNT38-GFP (J) in BY-2 suspension-cultured cells and corresponding merged images ([C], [F], [I], and [L]). Insets: magnification of selected Golgi stacks (2.2x zoom). Note that stacks often appear tricolored (arrows in [F] and [I]) with the GFP fusions on one side (green), the ST52-RFP on the other side (red), and a region of overlap between them (yellow). This suggests that all the GFP constructs localize to the cis-Golgi as had been shown previously for ManI-GFP (Nebenführ et al., 1999), whereas ST52-RFP is in the trans-Golgi (Boevink et al., 1998). Bars = 8 µm in (A) to (I) and 16 µm in (J) to (L).

To determine if the portion of ManI located in the Golgi lumen plays a role in the targeting of this glycosidase to the Golgi and the ER membranes, the first 99 amino acids (CT+TMD+S) or the first 49 amino acids (CT+TMD) of ManI were fused to GFP, and the corresponding chimeric proteins were named Man99-GFP and Man49-GFP, respectively (Figure 1). Man99-GFP and Man49-GFP were either stably expressed in BY-2 suspension-cultured cells or transiently expressed in tobacco leaf epidermal cells by leaf infiltration. Both Man99-GFP and Man49-GFP chimeric proteins were observed in the Golgi and in the ER in both expression systems (Figures 3A , 3B, and 3D to 3F), exactly as previously observed for the full-length construct (Figures 2A and 2B). It is important to note that when these truncated fusions were transiently expressed in tobacco leaves, the ER labeling was still observed 5 d after transformation when the overall expression levels are already strongly declining (Figure 3F), whereas XYLT35-GFP was located in the Golgi only (Pagny et al., 2003; Figure 3I). This further confirms that the partial location of the ManI fusions in the ER is not due to overexpression of the chimeric proteins. In addition, when BY-2 cells expressing Man99-GFP were treated with the protein synthesis inhibitor cycloheximide for 2 h, the ER labeling did not disappear (Figure 3C) as had been observed for the full-length fusion (cf. Figure 2C).

View larger version (116K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 3. Luminal Domains Are Not Necessary to Target ManI and GNTI to the Golgi and the ER. (A) and (B) Cells expressing Man99-GFP display a punctuate and network pattern of GFP fluorescence typical of the organization of Golgi stacks and cortical ER in BY-2 cells and similar to the labeling observed for ManI-GFP. (C) The ER and Golgi labeling remains unchanged when cells expressing Man99-GFP were incubated with the protein synthesis inhibitor cycloheximide for 2 h. (D) Man49-GFP was located to the Golgi and to the ER in BY-2 suspension-cultured cells. (E) and (F) Leaf epidermal cells expressing Man99-GFP (E) and Man49-GFP (F) display the same labeling pattern 5 d after agroinfiltration as BY-2 cells expressing the same proteins. (G) and (H) As observed for the truncated forms of ManI fused to GFP, GNT38-GFP was located to the Golgi and to the ER in BY-2 suspension-cultured cells (G) and in Nicotiana leaf epidermal cells (H). (I) XYLT35-GFP highlights only the Golgi in Nicotiana leaf epidermal cells. Bars = 8 µm.

The first 77 N-terminal amino acids of the tobacco GNTI, including the CT, the TMD, and the stem, were previously described to contain the information required to maintain Golgi retention of this glycosyltransferase (Essl et al., 1999). This polypeptide domain fused to GFP has been shown to be preferentially located in the Golgi, but the chimeric protein was also detected in the ER as observed here for the full-length construct (Figure 2H).To determine if the sequence remaining in the Golgi lumen is involved in the Golgi and ER targeting of GNTI, the lumenal part (39 amino acids) was removed and the remaining first 38 N-terminal amino acids (CT+TMD) of this glycosyltransferase were fused to GFP (Figure 1). The fusion protein was named GNT38-GFP and stably expressed in BY-2 suspension-cultured cells or transiently in tobacco leaf epidermal cells. In both expression systems, GNT38-GFP was located in the Golgi and in the ER (Figures 3G and 3H) as previously observed for the full-length construct (GNTI-GFP; Figure 2H). In addition, when GNT38-GFP was stably coexpressed with ST52-mRFP in BY-2 cells, as observed before with Man99-GFP and Man49-GFP, the two fluorescent spots overlapped, but some red fluorescence was distinguishable from the yellow, suggesting that GNT38-GFP is not in the trans-Golgi (Figures 4J to 4L) as previously observed for ManI.

It is clear from these results that the CT and TMD of both ManI and GNTI are sufficient to target these glycosylation enzymes to their steady state location: the ER and the early Golgi compartments. By contrast, the same domain (CT+TMD) targets XYLT35-GFP to the Golgi only, both in BY-2 cells (Pagny et al., 2003) and N. tabacum leaf epidermal cells (Figure 3I).

The CT Is Not Necessary for the Retention of ManI in the Early Compartments of the Secretory Pathway
The N-terminal cytosolic region of many membrane-bound proteins residing in the mammalian and yeast ER and/or in the Golgi apparatus contains signals that facilitate either their retrieval from the Golgi back to the ER (Teasdale and Jackson, 1996; Zerangue et al., 1999) or their export from the ER to the Golgi (Giraudo and Maccioni, 2003). In plants, the length of CTs can vary widely between the different glycosidases and glycosyltransferases (Figure 11). For instance, GCSI and ManI contain a long CT (51 and 29 amino acids, respectively). In comparison, the CTs of GNTI and XYLT are only composed of 11 amino acids.

To define more precisely the targeting signal of ManI and to investigate the role of the relatively long cytoplasmic domain (29 amino acids) of this glycosidase in this targeting, we generated two fusion proteins (19Man-GFP and 19Man49-GFP). In the latter two proteins, 19 amino acids were removed so that the CT was shortened down to 10 amino acids, like the CTs of XYLT and GNTI. This truncation removed a potential dibasic motif (KxR) that might function in ER-to-Golgi transport (Giraudo and Maccioni, 2003), although another potential ER export signal remained. A complete removal of the CT was attempted (CTMan49-GFP, Figure 1), but it was not possible to get this fusion protein expressed in tobacco cells. To look for a targeting determinant in the remaining 10 amino acids of the CT, the CT sequence was substituted by an artificial sequence (MAAA) (MAAAMan49-GFP, Figure 1). This artificial sequence does not contain any known targeting sequences and does not affect the length of the hydrophobic TMD.

The three constructs (MAAAMan49-GFP, 19Man-GFP, and 19Man49-GFP) were expressed in tobacco cells. The two latter labeled the ER and the Golgi (Figures 5A to 5D) just like the constructs from which they originate (Figures 2A, 2B, 3D, and 3F). In addition, the fusion protein containing an artificial MAAA CT was located in the same compartments (Figure 5E), indicating that the N-terminal cytosolic region is not necessary for ManI targeting; consequently, all information required for its steady state localization to both the ER and the Golgi apparatus is contained within the 20 amino acids of the MAAAMan49-GFP construct (i.e., in the TMD and C-terminal flanking amino acids).

View larger version (83K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 5. The CT Is Not Needed for ManI Localization to the Golgi and the ER. (A) to (C) 19Man-GFP (A) and 19Man49-GFP ([B] and [C]) are located to the Golgi and to the ER in BY-2 suspension-cultured cells. Bars = 16 µm. (D) 19Man49-GFP also highlights the Golgi and the ER in Nicotiana leaf epidermal cells. (E) A similar distribution pattern was observed for MAAAMan49-GFP after expression of this fusion in leaf epidermal cells. Bars = 8 µm.

In this study, we have shown that the information required for ManI targeting is contained within a 20–amino acid sequence, including the 16–amino acid TMD. To investigate whether the length of the TMD could play a key role in the targeting of this type II membrane protein in the early plant secretory pathway, we designed two fusion proteins, ManTMD23-GFP and Man99TMD23-GFP, where the TMD of ManI was lengthened from 16 to 23 amino acids by duplication of its last seven amino acids (Figure 1). ManTMD23-GFP and Man99TMD23-GFP were expressed in BY-2 suspension-cultured cells and in tobacco leaf epidermal cells. In both plant expression systems, ManTMD23-GFP and Man99TMD23-GFP were exclusively located in bright spots (Figures 6A , 6B, and 6D) and were sensitive to the fungal toxin BFA (50 µg·mL^–1, 2 h; Figure 6C). The expression patterns of ManTMD23-GFP and Man99TMD23-GFP were similar to either the XYLT-GFP fusion (Figures 3I and 6E) or the ST52-mRFP fusion (Figure 6F), both located exclusively in the Golgi in BY-2 suspension-cultured cells and tobacco leaf epidermal cells, as has been confirmed previously by electron microscopy (Boevink et al., 1998; Pagny et al., 2003).

View larger version (30K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 6. An Increase in TMD Length Displaces ManI from the ER to the Golgi. Modification of TMD length in ManTMD23-GFP (A) and Man99TMD23-GFP (B) leads to the accumulation of fusion proteins in the Golgi of BY-2 tobacco cells, as observed for XYLT35-GFP (E) and ST52-mRFP (F). No ER labeling was observed in cells expressing these fusion proteins. The same results were obtained after transient expression in Nicotiana (D), G. max ([G] and [H]), or S. lycopersicum ([I] and [J]) leaf epidermal cells. ManTMD23-GFP is redistributed to the ER after a 2-h treatment with BFA (C). Bars = 16 µm.

To further investigate the subcompartmentation of ManTMD23-GFP and Man99TMD23-GFP, stable BY-2 suspension-cultured cells coexpressing one or the other of these GFP fusions and ST52-mRFP were established. In the merged images, it was impossible to separate green spots from red spots, suggesting that the GFP fusions containing a 23–amino acid TMD have moved forward within the Golgi toward the trans-face so that they colocalize with ST52-mRFP at the confocal level (cf. Figures 4 and 7 ). Interestingly, the spot patterns were similar in cortical images (Figures 7D to 7F) compared with cross sections (Figures 7G to 7I), reinforcing the assumption that the Man-GFP fusions with a longer TMD and the trans-Golgi marker ST52-mRFP perfectly colocalize. By contrast, the medial Golgi marker (XYLT35-GFP) and the trans-Golgi marker (ST52-mRFP) resulted in fluorescent spots that did not overlap perfectly in the merged image (Figures 7J to 7L).

View larger version (72K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 7. An increase in TMD Length Displaces ManI and Man99 to Late Golgi Compartments: Evidence from Coexpression with the trans-Golgi Marker ST52-mRFP. (A) to (C) When ManTMD23-GFP (A) and ST52-mRFP (B) were coexpressed in BY-2 suspension-cultured tobacco cells, they both targeted to the Golgi and perfectly colocalized, as illustrated in the zoomed insets (2.2x) where the spots are yellow. (D) to (I) When Man99TMD23-GFP and ST52-mRFP were coexpressed in BY-2 suspension-cultured cells, they both targeted to the Golgi and perfectly colocalized (2.2x zoomed insets). In the cross sections ([G] to [I]), some red fluorescence can be distinguished from the yellow, whereas it is almost impossible in the cortical section, suggesting that the Golgi stacks are oriented toward the plasma membrane in the cortical cytoplasm. (J) to (L) XYLT35-GFP and ST52-mRFP were coexpressed in BY-2 suspension-cultured cells. Both fusion proteins targeted to the Golgi, but some green fluorescence can be distinguished from the yellow in some cases, consistent with their localization mainly in the medial Golgi (XYLT35-GFP) or in the trans-Golgi (ST52-mRFP). Bars = 8 µm.

View larger version (60K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 8. Subcompartmentation of Man99-GFP and Man99TMD23-GFP in the Golgi Apparatus. Electron micrographs of Golgi stacks were realized from suspension-cultured wild-type BY-2 tobacco cells (A) or from BY-2 cells expressing Man99-GFP (B) or Man99TMD23-GFP (C). Immunogold labeling was performed with anti-GFP antibodies. Distribution of fusion proteins in the Golgi was expressed as a percentage of label observed for 15 individual stacks analyzed in different cells (D), with counts of gold particles on the cis-side and trans-side after having divided Golgi in two domains. SV, secretory vesicle.

Late and Early Golgi Proteins Redistribute in the ER in the Presence of BFA
Taking advantage of the large panel of Golgi marker generated during this study, we have investigated the possibility that Golgi proteins located in different Golgi subcompartments may behave differently after BFA treatment. Cells expressing ER-soluble or membrane markers (GFP-HDEL or GCS90-GFP, Figures 9A to 9D), the ER/early Golgi marker (19Man49-GFP, Figures 9E and 9F), the medial Golgi marker (XYLT35-GFP, Figures 9G and 9H), or the late Golgi marker (Man99TMD23-GFP or ST52-mRFP, Figures 9I to 9L) were treated with BFA (50 µg·mL^–1) for 2 h. At the end of the BFA treatment, all the fusion proteins accumulated in bright aggregates and in the ER (Figures 9E to 9L) except for the ER markers that were never found in aggregates (Figures 9A to 9D). We illustrate here that all markers relocated in the ER in presence of BFA. In cells coexpressing ER/early Golgi or late Golgi proteins with ST52-mRFP, BFA induces the redistribution of both markers into the ER and into Golgi aggregates (Figure 10 ), except for GFP-HDEL (Figures 10A to 10F) and GCS90-GFP (Figure 9D) that were not found in the aggregates. In some cases, there are subtle differences in timing, but these are not trivial to detect and also not informative with respect to intra-Golgi localization. Furthermore, no significant difference was observed in the fluorescence patterns observed after BFA treatment of cells expressing either a soluble (GFP-HDEL) or a membrane protein (GCS90-GFP) marker (Figures 9A to 9D).

View larger version (42K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 9. Effects of BFA on ER and/or Golgi Proteins in BY-2 Cells. BY-2 cells expressing a soluble ER marker (GFP-HDEL; [A] and [B]) or a membrane ER marker (GCS90-GFP; [C] and [D]) display a typical fluorescence of the ER in the presence or absence of BFA. An ER and early Golgi marker (19Man49-GFP; [E] and [F]), a medial Golgi marker (XYLT35-GFP; [G] and [H]), or late Golgi markers (Man99TMD23-GFP and ST52-GFP; [I] to [L]) highlight the ER and aggregates assimilated to Golgi clusters in the presence of BFA. Note that in all cases, the ER network often turns into fenestrated sheets of fluorescence. Bars = 8 µm.

View larger version (61K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 10. BFA Induces the Simultaneous Redistribution of Both Early and Late Golgi Markers into the ER and Golgi Clusters. BY-2 cells coexpressing GFP-HDEL ([A] to [F]), 19Man49-GFP ([G] to [L]), or Man99TMD23-GFP ([M] to [R]) with ST52-mRFP were analyzed in the absence ([A] to [C], [G], [H], and [M] to [O]) or presence ([D] to [F], [J] to [L], and [P] to [R]) of BFA (50 µg·mL–1).

Based on its short 18–amino acid TMD that could perfectly fit with the lipid bilayer model to explain its localization in the ER membrane, we have selected GCSI to check for general applicability of this model. To define whether the TMD of GCSI was sufficient for its targeting and retention in the ER, we first deleted most of the luminal part of this glycosidase (containing the catalytic domain) and fused its first N-terminal 90 amino acids (CT+TMD+S) to GFP to get the fusion protein GCS90-GFP (Figure 1). When this fusion was expressed in tobacco cells, the ER was highlighted (Figures 12A and 12B) in a pattern very similar to the one obtained with the full-length construct GCSI-GFP and the GFP-HDEL construct (cf. micrographs in Figures 12A and 12B to Figures 2D, 2E, 2G, and 10C). This result clearly shows that GCSI targeting to the ER depends on signals located within the CT, the TMD, and/or the 21 luminal amino acids remaining in this truncated protein. In a further attempt at defining the minimal protein sequence required for localization of GCSI in the ER, we deleted the first N-terminal 13 amino acids from the GCS90-GFP construct to obtain 13GCS90-GFP (Figure 1). When this fusion was expressed in tobacco suspension-cultured or leaf epidermal cells, the chimeric protein was located exclusively in Golgi-like spots (Figures 12D and 12E), as observed for XYLT-GFP (Figures 3I and 6E) and ST52-GFP (Figure 12F). In conclusion, the 18–amino acid long TMD of GCSI is not sufficient to target this glycosidase in the ER membrane, and additional information contained in the first 13 amino acids of the CT is required for the normal localization of this glycosylation enzyme in the secretory system. This result provides experimental proof that factors other than TMD length can influence the positioning of glycosylation enzymes in the early secretory pathway.

View larger version (46K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 12. The TMD Does Not Always Carry the Major Targeting Information: Evidence from GCSI Targeting to the ER. (A) to (C) In BY-2 cells, GCS90-GFP (A) highlighted the ER like the full-length GFP fusion GCSI-GFP (Figure 2G) and the ER marker GFP-HDEL (Figures 2D and 2E). In Nicotiana leaf epidermal cells, GCS90-GFP (B) also shows the same network pattern as GFP-HDEL (C). Bars = 8 µm. (D) to (F) After deletion of the first 13 amino acids of the CT of GCS90-GFP, 13GCS90-GFP localized in the Golgi in BY-2 cells (D) and Nicotiana leaf epidermal cells (E) as observed for XYLT or ST52-GFP/mRFP (Figures 6E and 6F, respectively) . Bars = 16 µm in (D) and (E) and 8 µm in (F).

	DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

The model that glycosidases and glycosyltransferases constituting the plant N-glycan processing machinery are organized in an assembly line is illustrated in Figure 11. With the exception of glucosidase II (GCSII), all glycosidases and glycosyltransferases involved in this assembly line have the same type II membrane protein topology with a short N-terminal CT, a TMD, and lumenal stem and catalytic domains. Even though the relative position of glycosyltransferases in the assembly line is known to govern in part the structure of N-glycans produced by the cell, the mechanisms allowing their selective retention in certain Golgi cisternae are still poorly understood, especially in plant cells.

In this study, we have analyzed the targeting of several N-glycosylation enzymes and found that they are located in the ER and/or in the Golgi in good agreement with their position in the N-glycan maturation pathway. We have also investigated the sequence responsible for the targeting of ManI, a type II membrane protein located in the early Golgi and in the ER, and we have shown that the TMD length contains the essential targeting elements for this glycosidase. However both in silico analyses of plant glycosylation enzymes cloned so far and the characterization of a 13–amino acid N-terminal sequence responsible for GCSI retention in the ER indicate that TMD length is not always the only sorting determinant allowing the subcompartmentation of glycosidases and glycosyltransferases along the secretory system. Our results also exemplify that in silico analysis does not allow for a prediction of the location of type II membrane enzymes within the secretory pathway.

Plant N-Glycan Processing Enzymes Are Arranged along an Assembly Line in the Early Secretory Pathway
While there is general agreement that enzymes involved in the maturation of N-linked oligosaccharides are arranged in the correct sequence along the individual subcompartments of the animal secretory pathway (de Graffenried and Bertozzi, 2004; Young, 2004), very little data on the distribution of the homologous proteins from plants exist to date. This study represents a systematic elucidation of resident enzyme localization in the secretory pathway of plant cells. Although the results presented here are in general agreement with those obtained in animal systems, some differences between secretory systems in these two eukaryotes are noteworthy.

Arabidopsis GCSI is located in the ER exclusively like its human counterpart (Hardt et al., 2003), and the CT in both proteins contains targeting information. In contrast with GCSI, we have found ManI and GNTI to be localized to both the Golgi and the ER. These two N-glycan maturation enzymes act sequentially during the production of mature N-glycans in plant and mammalian cells after a first trimming of the oligosaccharide precursor by GCSI and GCSII. The dual ER and Golgi localization of ManI and GNTI in plant cells contrasts with results obtained in mammals where these glycosylation enzymes are concentrated in the early Golgi compartments only. This difference could be a result of the different organization of the early secretory pathway in these organisms. Mammalian cells contain an intermediate compartment between ER and Golgi (known as ERGIC or VTC) that performs essential recycling functions (Ben-Tekaya et al., 2004). It is conceivable that this additional compartment allows animal cells to maintain a more specific distribution of ManI or GNTI.

Our results illustrate that there are at least two different classes of type II membrane proteins in the membranes of the early secretory pathway. The first class is exemplified in this study by ManI. In this case, we have shown that the TMD acts as a targeting sequence holding the protein in the ER and cis-Golgi compartments based on the length of hydrophobic TMDs. In a second population illustrated by GCSI, the TMD acts as a general anchor, and additional mechanisms, depending on signals contained in the CT, would locate these type II membrane proteins preferentially in the ER. In the latter case, we suggest, as it has been proposed by Hanton et al. (2005), that additional sequences, preferentially located in the CT on some type II membrane protein, are dominant over TMD length in determining protein targeting.

XYLT, a glycosyltransferase acting later in the maturation of N-glycans, has been localized almost exclusively to the medial Golgi apparatus (Pagny et al., 2003), again in keeping with the assembly line model. Interestingly, under the conditions used in this study, we cannot detect any of the XYLT-GFP fusion protein in the ER, which contrasts with the results obtained with ManI-GFP and GNTI-GFP. Assuming that XYLT-GFP continuously recycles through the ER, as has been shown for the trans-Golgi marker ST52-GFP (Brandizzi et al., 2002b), we have to postulate that the resident time of the protein in the ER is too short to allow for the accumulation of significant fluorescence signal in this compartment or that only a small fraction of XYLT-GFP actually recycles back to the ER.

In summary, all N-glycan processing enzymes of plants whose intracellular localization has been studied to date conform to the assembly line model in that their position within the organelles of the secretory pathway mirrors their position in the biochemical pathway that results in complex N-linked oligosaccharides. While the precise distribution of enzyme activities in plants may not always match that found in animal systems, the general sequential arrangement of the enzymes is the same. As a corollary, it is now possible to use enzymes of the plant N-glycan maturation pathway as markers for the different subcompartments in the secretory pathway of plant cells. These new tools should also allow us to study the transport processes involved in movement of secretory cargo from the ER to the plasma membrane or to the vacuoles in greater detail. Furthermore, the differential localization of plant glycosyltransferases within the Golgi and the possibility to resolve at the light microscopy level the intra-Golgi targeting of these new markers offer a simple way to identify where the subset of cisternae a protein of interest goes.

Subcompartmentation of ManI Is Related to Its TMD Length
For mammalian cells, several models have been proposed to explain how type II membrane proteins are retained at different levels within the Golgi. According to the kin recognition model (Nilsson et al., 1993b), aggregation of N-glycan maturation enzymes by homo/hetero-oligomerization would prevent the resulting large complexes from being delivered to secretory vesicles and ongoing forward transport downstream in the secretory pathway. One of the first reported cases of this type of association involved ManII and GNTI, two glycosylation enzymes located in the medial-Golgi and acting sequentially in mammalian N-glycan maturation (Nilsson et al., 1994). It should be noted that this model originally assumes the presence of stable Golgi cisternae and the anterograde flow of secretory cargo via vesicular shuttles (Nilsson et al., 1993b). To fit with the cisternal progression/maturation concept, the kin recognition model would have to be modified to allow for the oligomeric complexes to be preferentially packaged into retrograde vesicles (Füllekrug and Nilsson, 1998). A second model, the lipid bilayer model (Bretscher and Munro, 1993) proposes that the fit between the length of TMD of glycan maturation enzymes and the thickness of the lipid bilayer of each organelle membrane determines the localization because each organelle has its specific membrane lipid composition and consequently its own thickness (Morré and Mollenhauer, 1974; Hartmann and Benveniste, 1987; Lynch, 1993; Moreau et al., 1998). Comparisons revealed that the length of Golgi protein TMD were on average five amino acids shorter than those of plasma membrane proteins (Masibay et al., 1993; Munro, 1995a). Several examples are in favor of this model. An increase of the length of the TMD of rat -2,6-ST and bovine ß-1,4-galactosyltransferase reduced the Golgi retention of these glycosyltransferases. In addition, a synthetic type I TMD made of 17 Leu residues resulted in Golgi retention of the lymphocyte surface antigen CD8 extracellular domain, whereas a 23-Leu TMD was found in increased amounts at the cell surface (Munro, 1991, 1995b; Masibay et al., 1993). Furthermore, incremental increases in the length of the 18–amino acid -2,6-ST TMD by insertion of one to nine hydrophobic amino acids also resulted in increased cell surface expression of similar -2,6-ST lysosyme chimeras, while the decrease in the length of a plasma membrane protein TMD led to its increased retention in the Golgi (Munro, 1995b).

In plant cells, preliminary data in favor of a subcompartmentation of membrane proteins along the endomembrane system related to the TMD length was obtained by varying the length of TMDs in two type I membrane proteins fused to GFP (Brandizzi et al., 2002a). First, the human lysosomal membrane protein LAMP1 containing a 23–amino acid TMD was fused to GFP and was expressed in tobacco leaves. The fusion was located in the plasma membrane. By contrast, when the TMD was shortened to 20 and 17 amino acids, the GFP chimeras were localized to the Golgi and ER membranes, respectively. Secondly the 19–amino acid TMD of the vacuolar sorting receptor BP80 targeted GFP to the Golgi, whereas a lengthened TMD of 22 amino acids targeted GFP to the plasma membrane.

In this study, we have investigated whether the TMD length of the type II membrane protein ManI could affect its subcompartmentation in the Golgi. In particular, we have increased the TMD length from 16 to 23 amino acids by duplicating the last seven amino acids of this domain. In contrast with their homologues containing a 16–amino acid TMD, which were located in the ER and the cis-half of the Golgi apparatus, chimeric proteins with a 23–amino acid TMD were localized exclusively to the Golgi and more precisely in the trans-half of the Golgi stacks. Together with our results demonstrating that the TMD is sufficient to confer an identical localization as the full-length protein ManI, our data suggest that the length of the TMD is a crucial factor for precise positioning of this type II membrane protein within the Golgi stacks and the ER. Thus, protein–lipid interactions are expected to play a key role in ManI targeting within the secretory system.

Interestingly, these results also clearly point out differences in TMD length requirements in the targeting of type I and type II membrane proteins in the plant secretory system. Indeed, the 23–amino acid TMD of XYLT (Dirnberger et al., 2002; Pagny et al., 2003) or the 22–amino acid TMD of ManII (Strasser et al., 2006) targets GFP to the Golgi only. In addition, here, ManI with a lengthened TMD (23 amino acids) is also detected exclusively in the Golgi. By contrast, the 23–amino acid TMD of a type I membrane protein and the lengthened 22–amino acid TMD of BP80 target GFP to the plasma membrane in Brandizzi et al. (2002a). The TMD length requirements for a membrane protein to stay in a membrane with a given thickness might depend on the topology of the protein (type I or type II). While these experiments clearly demonstrate the role of TMD length in ManI protein targeting, we could also show that other enzymes require other signals for proper localization. For example, contradicting the trend to longer TMDs in the later parts of the Golgi, the ST52-GFP fusion with an 18–amino acid TMD is found further downstream in the trans-Golgi (Boevink et al., 1998; Wee et al., 1998) than the XYLT35-GFP fusion with a 23–amino acid TMD (Pagny et al., 2003).

Even in the situation illustrated here with GCSI, whose TMD is one of the shortest identified so far for a plant glycosylation enzyme and allows for a localization in the ER, we now show that additional information contained in the CT are required for proper targeting. Thus, in silico analyses and mutagenesis studies performed on GCSI are not consistent with TMD length as the only signal for compartmentation of glycosylation enzymes in the plant secretory system. In other words, while the TMD length plays a key role for ManI targeting in the ER and the cis-Golgi, results obtained with GCSI illustrate that specific localization of some membrane proteins in the ER or Golgi membranes could also depend on both protein–lipid (via the TMD) and protein–protein (via special sorting motifs) interactions. The identification of cytosolic partners, such as Golgi matrix proteins or cytoplasmic regulators, should help us to explain mechanisms involved in this second model for partitioning the N-glycan maturation enzymes along the plant secretory pathway.

The large collection of enzymes localizing to different levels in the Golgi has allowed us to test the question whether all cisternae within the Golgi stack fuse with the ER in response to treatment with the fungal toxin BFA. Indeed the Man-GFP fusions containing either a 16– or 23–amino acid TMD and ST52-mRFP all moved back to the ER or in Golgi clusters over a 2-h time-course experiment with BFA. These conclusions are consistent with previously published results (Nebenführ et al., 2002; Ritzenthaler et al., 2002).

Conclusions
We have demonstrated that the enzymes of the N-glycan processing pathway are spatially arranged so that they form an assembly line where individual enzymes act sequentially on their glycoprotein substrates. The length of the TMD can act as a key signal in targeting some glycosylation enzymes to their proper location within the secretory pathway, as exemplified here with ManI. However, we have also discovered instances where additional targeting signals located in the CT of type II membrane proteins must be operating, for example, to prevent loss of GCSI from the ER. Further studies will use GCSI as a model to investigate the mechanisms that plant cells employ to translate these signals into a specific subcellular localization.

	METHODS

TOP ABSTRACT INTRODUCTION RESULTS DISCUSSION METHODS REFERENCES

 
Constructs
All ManI fusion constructs were derived from the full-length GFP fusion (here called ManI-GFP) originally described by Nebenführ et al. (1999). First, a linker containing an AatII restriction site was introduced between the ManI and the GFP coding regions. In combination with the native AatII site near the end of the predicted stem region, this allowed for simple removal of the catalytic domain to yield Man99-GFP.

Second, to facilitate the removal of specific segments of the N-terminal region, three new restriction sites were introduced by PCR mutagenesis: one NheI site immediately behind the start codon, one SpeI site at codons 21 and 22, and another AatII site at codons 50 and 51. The integrity of the modified construct was confirmed by sequencing. In this new construct, the CT could be removed with NheI and SpeI to give 19CTManI-GFP, while an AatII digest would remove the entire lumenal part of ManI to yield Man49-GFP. Combination of the two procedures resulted in 19CTMan49-GFP. Finally, forward (5'-GATCCTTGGGAATGCTTGCTCTGCTCTTCATCGTTTTCGTTTGTGTCTCTTTCGTTTTCTGGGACCGTCAAA-3') and reverse (5'-CTAGTTTGACGGTCCCAGAAAACGAAAGAGACACAAACGAAAACGATGAAGAGCAGAGCAAGCATTCCCAAG-3') oligonucleotides encoding the 18 amino acids of the TMD domain with a start codon were synthesized, fused, and subcloned into the pBLTI121 binary vector containing the GFP without any start codon (M.C. Kiefer-Meyer and V. Gomord, unpublished data) to give CTMan49-GFP. The same strategy has been used to generated MAAMan49-GFP using forward (5'-GATCCTTGGGAATGGCTGCTGCTCTTGCTCTGCTCTTCATCGTTTTCGTTTGTGTCTCTTTCGTTTTCTGGGACCGTCAAA-3') and reverse (5'-CTAGTTTGACGGTCCCAG