Expression of GFP-fusions in Arabidopsis companion cells reveals non-specific protein trafficking into sieve elements and identifies a novel post-phloem domain in roots


  • Ruth Stadler,

    1. Molekulare Pflanzenphysiologie, Universität Erlangen-Nürnberg, Staudtstraße 5, D-91058 Erlangen, Germany
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Kathryn M. Wright,

    1. Cell-cell communication programme, Scottish Crop Research Institute, Invergowrie, Dundee DD2 5DA, UK
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Christian Lauterbach,

    1. Molekulare Pflanzenphysiologie, Universität Erlangen-Nürnberg, Staudtstraße 5, D-91058 Erlangen, Germany
    Search for more papers by this author
  • Gabi Amon,

    1. Molekulare Pflanzenphysiologie, Universität Erlangen-Nürnberg, Staudtstraße 5, D-91058 Erlangen, Germany
    Search for more papers by this author
  • Manfred Gahrtz,

    1. Molekulare Pflanzenphysiologie, Universität Erlangen-Nürnberg, Staudtstraße 5, D-91058 Erlangen, Germany
    Search for more papers by this author
    • Present address: Institute of General Botany, Centre for Applied Plant Molecular Biology (AMPII), University of Hamburg, Ohnhorststr. 18, D-22609 Hamburg, Germany.

  • Andrea Feuerstein,

    1. Molekulare Pflanzenphysiologie, Universität Erlangen-Nürnberg, Staudtstraße 5, D-91058 Erlangen, Germany
    Search for more papers by this author
  • Karl J. Oparka,

    1. Cell-cell communication programme, Scottish Crop Research Institute, Invergowrie, Dundee DD2 5DA, UK
    Search for more papers by this author
  • Norbert Sauer

    Corresponding author
    1. Molekulare Pflanzenphysiologie, Universität Erlangen-Nürnberg, Staudtstraße 5, D-91058 Erlangen, Germany
      For correspondence (fax +49 9131 8528751; e-mail
    Search for more papers by this author

For correspondence (fax +49 9131 8528751; e-mail


Transgenic Arabidopsis plants were constructed to express a range of GFP-fusion proteins (36–67 kDa) under the companion cell (CC)-specific AtSUC2 promoter. These plants were used to monitor the trafficking of these GFP-fusion proteins from the CCs into the sieve elements (SEs) and their subsequent translocation within and out of the phloem. The results revealed a large size exclusion limit (SEL) (>67 kDa) for the plasmodesmata connecting SEs and CCs in the loading phloem. Membrane-anchored GFP-fusions and a GFP variant targeted to the endoplasmic reticulum (ER) remained inside the CCs and were used as ‘zero trafficking’ controls. In contrast, free GFP and all soluble GFP-fusions, moved from the CCs into the SEs and were subsequently translocated through the phloem. Phloem unloading and post-phloem transport of these mobile GFP-fusions were studied in root tips, where post-phloem transport occurred only for the free form of GFP. All of the other soluble GFP-fusion variants were unloaded and restricted to a narrow zone of cells immediately adjacent to the mature protophloem. It appears that this domain of cells, which has a peripheral SEL of about 27–36 kDa, allows protein exchange between protophloem SEs and surrounding cells, but restricts general access of large proteins into the root tip. The presented data provide additional information on phloem development in Arabidopsis in relation to the formation of symplasmic domains.


The phloem of higher plants facilitates the long-distance allocation and partitioning not only of organic carbon and nitrogen compounds, such as sugars, sugar alcohols and amino acids, but also of macromolecules, including proteins and RNAs (Balachandran et al., 1997; Sjolund, 1997). In the phloem of angiosperms, the sieve element–companion cell (SE–CC) complex represents a functional unit, within which the individual cells are descendents of a common mother cell (Esau, 1969a; Oparka and Turgeon, 1999). After the division of this cell the newly formed daughter cells undergo different developmental programs, eventually producing individuals with highly specialized anatomies and functions. During maturation, SEs lose their nuclei, ribosomes and vacuoles and possess only reduced numbers or specialized forms of most other cellular organelles (Behnke, 1989). Consequently, these cells depend during their entire lifespan on the continuous supply not only of energy, but also of macromolecules, such as enzymes, structural proteins and membrane transporters. All these components seem to be provided by the closely linked CCs, which have small vacuoles and are densely packed with mitochondria and ribosomes. SEs and CCs are intimately linked via specialized plasmodesmata that function as intercellular conduits between these cell types (Behnke, 1989; van Bel and Kempers, 1996; van Bel et al., 2002; Oparka and Turgeon, 1999).

In many plants, phloem loading occurs from the apoplast via plasma membrane-localized transporters located around the SE–CC complex (Barth et al., 2003; Williams et al., 2000). In Arabidopsis the gene for the plasma membrane-localized AtSUC2 H+-sucrose symporter is expressed specifically, and exclusively in the CCs of the phloem (Stadler and Sauer, 1996; Truernit and Sauer, 1995). Similar expression data were obtained for genes encoding transporters involved in the phloem loading in other plants [e.g. Plantago major (common plantain) sucrose transporter PmSUC2 (Stadler et al., 1995)] or in the phloem loading of other substrates [e.g. common plantain sorbitol transporters PmPLT1 and PmPLT2 (Ramsperger-Gleixner et al., 2004)]. Therefore, the promoters of these genes represent excellent tools for analyses of cell-to-cell trafficking between SEs and CCs (Ayre et al., 2003; Imlau et al., 1999; Oparka et al., 1999).

Studies in Arabidopsis plants expressing GFP under the control of the CC-specific AtSUC2 promoter revealed, for the first time, CC–SE trafficking of GFP and its subsequent unloading in sink tissues (Imlau et al., 1999). This study showed that plasmodesmata between these two cell types have a potentially large size exclusion limit (SEL), allowing the passage of proteins up to 27 kDa, and confirmed earlier results obtained using microinjection of fluorescent dextrans (Kempers and van Bel, 1997). In addition, Imlau et al. (1999) showed that GFP was unloaded from terminal SEs, and that unloaded GFP underwent extensive post-phloem transport within sink tissues. In contrast, no unloading was observed in Arabidopsis source tissues. Subsequently, Oparka et al. (1999) demonstrated that plasmodesmata of sink-leaf mesophyll cells underwent a downregulation in their SEL during the sink-to-source transition. These authors detected a transition from simple (unbranched) plasmodesmata in sink leaf mesophyll cells of tobacco to complex (branched) plasmodesmata in source leaves, which was paralleled by a major decrease in the SEL of these plasmodesmata. Since this study, several papers have reported the movement of free GFP in a range of plant tissues and organs (Crawford and Zambryski, 2000, 2001; Itaya et al., 2000).

Recently AtSUC2 promoter::GFP constructs were used for a functional analyses of different vein classes in developing leaves of transgenic tobacco plants (Wright et al., 2003). In these studies the unloading of phloem-mobile GFP into the sink areas of transition leaves was compared with the loading of 14C-sucrose into the minor veins of the same areas. These analyses showed a clear correlation between the decreasing capacity of veins to unload GFP and their increasing capacity to load sucrose. In the same study, transgenic tobacco plants were analyzed that expressed a GFP variant that was targeted to the endoplasmic reticulum (ER). This ER-GFP could not traffic into tobacco SEs, was not phloem-mobile, and the GFP-dependent fluorescence was detected exclusively in the veins of source leaves. In sink–source transition leaves, the ER-GFP fluorescence correlated precisely with the regions exhibiting apoplastic sucrose loading and lacking GFP-unloading from the phloem in AtSUC2-promoter::GFP plants (Wright et al., 2003). These data confirmed that the sites of AtSUC2 promoter-driven GFP expression represent the sites of active phloem loading into the minor veins. Furthermore, they showed that the movement of free GFP within the phloem, and its unloading into sink leaves, is a clear indicator for the movement and unloading of assimilates.

In the present study, the non-invasive approaches described above were extended to include seven different GFP-variants of increasing molecular mass (36–67 kDa). Transgenic Arabidopsis plants expressing the genes for five cytosolic [GFP-ubiquitin, apparent molecular weight (MWapp): 36 kDa; GFP-sporamin, MWapp: 47 kDa; GFP-aequorin, MWapp: 48 kDa; GFP-patatin, MWapp: 67 kDa], two membrane-anchored (tmGFP9, tmGFP2) and one ER-localized GFP-fusion (ER-GFP) were generated under control of the AtSUC2-promoter. Plants expressing these constructs were used to monitor the CC–SE trafficking of the different fusions, and to study their potential unloading in sink organs (developing leaves and roots). The data show that under non-invasive conditions SEs and CCs are connected by plasmodesmata that exhibit an SEL of up to 67 kDa, allowing the translocation of all the cytosolic GFP-fusion proteins produced in CCs. We demonstrate the presence of a symplasmic domain that encircles the root protophloem, allowing the escape of free GFP but restricting the widespread post-phloem distribution of macromolecules >30 kDa. Our results underline the CC-specificity of the AtSUC2 promoter (Truernit and Sauer, 1995), and provide new information on phloem function that correlates closely with the pattern of phloem development in Arabidopsis roots. Finally, our data underline the potential of the phloem for non-specific protein trafficking.


Expression of GFP and GFP-fusions in Arabidopsis source and sink leaves

Plants of Arabidopsis (ecotype C24) were transformed with the seven GFP fusion constructs (Figure 1), and transgenic lines were selected for growth in the presence of BASTA. The soluble proteins used for these GFP-fusions (patatin, sporamin, aequorin, ubiquitin) were chosen, because they are not expected to interfere with the cellular metabolism. Sporamin and patatin, which are normally localized in storage vacuoles, were amplified without their N-terminal signal sequences responsible for targeting of the proteins to the ER (Hattori et al., 1989; Mignery et al., 1984).

Figure 1.

Eight different constructs were used for the analyses of size exclusion limits.
In all constructs expression is driven by the CC-specific AtSUC2-promoter from Arabidopsis thaliana. In addition to a previously described construct with free GFP (bottom; 27 kDa; Imlau et al., 1999) four GFP-fusion constructs (bottom to top) were generated encoding fusions with ubiquitin (Ubi → fusion is 36 kDa), sporamin (→ fusion is 47 kDa), aequorin (→ fusion is 48 kDa) or patatin (→ fusion is 67 kDa) to the C-terminus of GFP, one construct encoding an ER-resident GFP (ER) and two constructs encoding GFP-variants fused to membrane anchors (tmGFPs for transmembrane-GFPs). The membrane-anchored GFP-variants were generated by cloning the GFP cDNA to the 3′-end of a truncated AtSTP9 gene (tmGFP9) or to the 3′-end of the AtSUC2 gene (tmGFP2). Dashed regions in the constructs for membrane-bound GFPs represent intron sequences.

The soluble fusion proteins have increased molecular masses, and these molecular mass values (kDa) are generally used to describe the SEL of plasmodesmata. Obviously, the molecular mass of GFP or of GFP-fusions is not the ideal unit to measure the SEL. The fusion product of globular GFP (Stokes radius = 2.82 nm; Terry et al., 1995) with another globular protein, such as ubiquitin (Stokes radius about 1.2 nm) or sporamin (Stokes radius about 2.2 nm), will not result in one larger globule, but rather in a dimer of two globules. Therefore, it is not only the increased molecular mass, but also the altered shape that influences the movement of GFP-fusions through plasmodesmata. However, in the absence of structural information on these fusion proteins their molecular masses (in kDa) are a crude but generally accepted unit for describing the SELs of plasmodesmata.

From each transformation plants from at least 30 independent BASTA-resistant lines were screened in the T2 generation for GFP fluorescence in their rosette leaves. Plants transformed with the same construct showed slight differences with respect to the fluorescence intensities, but no differences in the expression patterns were detected. Figure 2 (top row) shows the fluorescence in source leaves from the different transgenic lines in comparison with a leaf from the previously generated AtSUC2-promoter GFP plants (Imlau et al., 1999). GFP-fluorescence was detected in all lines analyzed, although at different intensities. Plants carrying the GFP-ubiquitin, GFP-sporamin and tmGFP9 fusions [AtSTP9 encodes an Arabidopsis monosaccharide transporter (Schneidereit et al., 2003)] showed similar GFP fluorescence, although the fluorescence in these plants was slightly lower than in plants with free GFP. Plants of the other lines showed decreasing fluorescence intensities (tmGFP2 > GFP-ER > GFP-aequorin > GFP-patatin).

Figure 2.

GFP fluorescence in source and sink leaves from transgenic plants expressing constructs for GFP or GFP-fusion proteins.
Source leaves (top row) and sink leaves (bottom row) from transgenic Arabidopsis plants expressing constructs shown in Figure 1 were photographed under a stereomicroscope with an excitation wavelength of 460–500 nm. Emitted fluorescence of GFP (green) and chlorophyll (red) was monitored at detection wavelengths longer than 510 nm. Bars = 2 mm in the top row and 0.1 mm in the bottom row.

Clear differences were observed in the cell-to-cell movement of the different GFP-fusions in the sink leaves of these transgenic lines. The previously described influx of source leaf-synthesized free GFP into Arabidopsis sink leaves (Figure 2; first leaf of second row) was not detected in sink leaves from lines expressing constructs for membrane-anchored GFPs (tmGFP9 or tmGFP2 lines), which is explained by the source-leaf specific activity of the AtSUC2-promoter and which confirms data from AtSUC2-promoter::GUS plants (Truernit and Sauer, 1995). No synthesis of membrane-bound GFP variants was expected to occur in the sink leaves of these plants, and the resulting GFP-fusion protein was not expected to be phloem-mobile. An identical result was obtained in sink leaves of ER-GFP plants. This soluble GFP variant made in the source leaves of these plants was targeted to the ER and could not move from the CCs into the SEs.

In contrast, GFP fluorescence was detected in the sink leaves of GFP-ubiquitin and of GFP-sporamin plants (Figure 2; bottom row) showing that these GFP-fusions were able to traffic from their site of synthesis (the CCs) into the SEs. Inside the SEs these GFP-fusions moved with the mass flow of assimilates and were eventually imported into the sink leaves. It is not clear if these GFP-fusions, like the free form of GFP, were unloaded into the surrounding mesophyll cells, or if post-phloem transport had occurred. Certainly, the extent of unloading was much lower than for free GFP (Figure 2, bottom row). Due to the even lower expression of GFP-aequorin and the GFP-patatin in transgenic plants, no data are presented relating to these transformants.

Trafficking of GFP and GFP-fusions in Arabidopsis roots

Analyses of sink-specific GFP-fluorescence is easier in Arabidopsis roots than in leaves, which are less accessible to non-invasive imaging (Oparka et al., 1994). Immunohistochemical studies (Stadler and Sauer, 1996) and GUS-histochemical analyses of AtSUC2-promoter::GUS plants had previously shown that the AtSUC2-promoter is active in the root phloem (Truernit and Sauer, 1995). Moreover, analyses of roots from AtSUC2-promoter::GFP plants had revealed that GFP synthesized under the control of this promoter is symplastically released via plasmodesmata from the vascular bundles at the root tips (Imlau et al., 1999), predominantly from the protophloem sieve tubes. It appears that these protophloem files undertake the bulk of phloem unloading in the root tip (Oparka et al., 1994; Schulz, 1994; Zhu et al., 1998).

We investigated, which of the GFP variants could be unloaded from the protophloem SEs into the sink tissues at the root tips and subsequently undergo post-phloem transport. All root analyses shown in Figure 3 were performed by confocal laser scanning microscopy (CLSM). As shown before by epifluorescence microscopy (Imlau et al., 1999), free GFP is unloaded at the tip of Arabidopsis roots from the two protophloem files (Figure 3a). Reproducibly, we observed symplastic unloading of free GFP also along the transport phloem (Figure 3b), although the extent of this unloading was significantly lower than in the root tips. This was unexpected, because in previous papers (Oparka et al., 1994; Wright and Oparka, 1997) no unloading of the small fluorescent probe 5(6) carboxyfluorescein (CF) had been observed from the root transport phloem. The fluorescence resulting from symplastically unloaded GFP was observed in the nuclei of all cell layers of the root cortex and in the nuclei of the root epidermal cell layer. This unloading from the root transport phloem is also seen in an optical cross section (Z-axis), where GFP-fluorescence is seen in the two strands of root vascular bundles, but with decreasing intensities also in cells of the root cortex and of the root epidermis (Figure 3c).

Figure 3.

Analysis of GFP-fluorescence in roots of transgenic Arabidopsis plants.
Green fluorescence resulting from symplastic phloem unloading of GFP in AtSUC2-promoter::GFP plants is seen in all cells of the root tip (a). AtSUC2-promoter::GFP plants show symplastic unloading of free GFP also into the cortical and rhizodermal cell layers in the more proximal regions of the root; arrows mark some of the labeled nuclei (b). The cross-section (optical Z-section) through the root section shown in (b) is presented in (c). Unloading of GFP is seen into several cells adjacent to the vascular strands (arrow with asterisk) and low amounts of GFP can even be detected in epidermal and subepidermal cells (arrows). A similar optical Z-section section as in (c) but from the root of an AtSUC2-promoter::GFP-sporamin plant shows no unloading of GFP-sporamin and fluorescence is restricted in the two phloem files (d). Longitudinal section through the root of an AtSUC2-promoter::GFP-sporamin plant (e). As in (d) and in contrast to (a) and (b) no unloading of GFP-sporamin is observed in the transport phloem and into the root tip of (e). Longitudinal section through the root of an AtSUC2-promoter::GFP-ubiquitin plant with no detectable unloading of GFP-ubiquitin from the phloem (f). Labeling of intracellular structures in the root of an AtSUC2-promoter::ER-GFP plant. No fluorescence is detected outside the phloem (g). Phloem tissue (longitudinal) close to the root tip of an AtSUC2-promoter::GFP-ubiquitin plant is shown in (h). The contact sites between two of the GFP-labeled cells (arrow) characterize these cells as SEs. The other fluorescent cells are SEs or CCs. In (i) individual CCs are labeled in the longitudinal section through the root of an AtSUC2-promoter::AtSUC2-GFP plant. All photos were taken under the CLSM and represent Z-stacks. Red fluorescence of cell walls results from staining with propidium iodide [not in (b), (e) and (h)]. For (b) and (e) several Z-stacks were assembled. Bars = 50 μm for (a), 40 μm for (b) and (d), 25 μm for (c), (f) and (g), 100 μm for (e), 10 μm for (h) and 20 μm for (i).

No symplastic unloading was seen from the transport phloem (Figure 3d) or from the more distal unloading phloem in the roots of GFP-sporamin plants (Figure 3e). In these plants the fluorescence is confined to the vascular strands, and nuclei in the cortex or the root epidermis do not show any GFP-labeling. The identical distribution of fluorescence was seen in GFP-ubiquitin plants (Figure 3f) and in the other plants expressing constructs for soluble GFP-fusions (data not shown). The sloping contact sites between the labeled cells shown at higher magnification identify these cells as SEs (Figure 3h; Stadler and Sauer, 1996) confirming that the GFP-fusions moved from CCs into SEs.

In the GFP-sporamin plants shown in Figure 3(e), fluorescence terminated at the ends of the two mature protophloem poles, and there was no evidence of post-phloem transport toward the root tip. However, closer examination of the root tips from these plants revealed the GFP-fusion protein also in a domain of cells surrounding the mature protophloem files (Figure 4a–c). To more clearly delineate this GFP-containing domain from the conducting protophloem SEs, we first imaged the root tip for GFP (Figure 4a), and subsequently for aniline blue staining to reveal the sieve plates of the protophloem SEs (Figure 4b). Clearly, the fluorescence in these root tips was confined to the protophloem and to this ‘protophloem domain’ (Figure 4c and insert). In contrast to plants synthesizing free GFP, no unloading beyond this domain and no GFP-labeling in the nuclei of the cortex or the root epidermis was observed (Figures 3e and 4). An identical distribution of fluorescence was seen in all other GFP-fusion plants examined (data not shown).

Figure 4.

Unloading of GFP-sporamin from the protophloem into a specific domain of cells.
A confocal GFP image of the root tip from a GFP-sporamin plant is shown (a). This root was also stained with aniline blue (for callose detection) and an epifluorescence image of the aniline blue-derived fluorescence is presented in (b). Aniline-stained sieve plates in the single protophloem pole are stained pink [arrow; false color representation for better differentiation of aniline-blue staining (bluish green) and GFP fluorescence (green) in the merged picture]. For the image shown in (c) and for the magnification shown in the insert of (c) images (a) and (b) were merged (Adobe Photoshop; Adobe Systems Inc., San Jose, CA, USA). GFP-sporamin is unloaded from the very thin protophloem sieve tube (identified by the callose stained sieve plates; arrow) into a discrete domain of surrounding cells.
One of the two differentiating protophloem SEs (marked with an asterisk) is seen in the center of the longitudinal section from an Arabidopsis root tip shown in (d). Vacuoles (V) and other organelles are still seen in the early SEs of the protophloem at the lower end of the SE-file, but are absent in the mature SEs in the upper end. In contrast, sieve plates (SP), which are clearly visible in the fully differentiated SEs (upper end of the file), are just being formed in the SEs of the protophloem (lower end). Bars = 25 μm for (a) to (c), 10 μm for the insert in (c) and 4 μm for (d).

As expected from the sink-leaf analyses (Figure 2), no trafficking of GFP out of the CCs was detected in the roots of ER-GFP plants, where the label was restricted to intracellular structures of the CCs, most likely the ER, which is suggested by the ring-like structures labeled inside these cells (possibly the labeled nuclear envelopes; Figure 3g). Similarly, no trafficking of GFP fluorescence was seen in tmGFP2 plants (Figure 3i), where individual CCs but no SEs were labeled as a consequence of the membrane anchor fused to this GFP-variant.

GFP translocation correlates with root phloem development

The trafficking behavior of soluble GFP-fusion proteins, and the lack of trafficking of the two membrane-bound GFPs, provides an excellent tool for the analysis of functional phloem development, and of the SELs between the cells involved. In Figure 5, typical confocal images of the tips of main roots from all eight transgenic lines expressing free GFP or GFP-variants under the control of the AtSUC2 promoter are presented. Note that only the free form of GFP showed extensive post-phloem transport throughout the root tip. In contrast, the fluorescence of all other phloem mobile GFP-variants (GFP-ubiquitin, GFP-sporamin, GFP-aequorin and GFP-patatin) ended at a similar distance from the root tip (approximately 250 μm). This distance was significantly greater in the root tips of the lines expressing the genes for the two membrane-bound GFPs (STP9-GFP or AtSUC2-GFP) and the ER-GFP (approximately 500 μm). Because of the CC-specific localization of the membrane-anchored GFP probes (Figure 3g,i), these GFP-variants should be trapped inside the CCs, suggesting that the extended GFP/tip-distances in these three lines is due to the onset of expression of the AtSUC2 promoter in the most distal CCs of the metaphloem.

Figure 5.

Comparison of cell-to-cell trafficking of free GFP and of different GFP-fusions in Arabidopsis root tips.
The three images on the top show the fluorescence emitted from the root tip of a tmGFP9 plant, of a GFP-sporamin plant and of a plant expressing free GFP. Fluorescence was monitored with a CLSM (cell walls stained with propidium iodide; maximal projection of 20 scans). Fluorescence was observed in the CCs of the transport phloem (tmGFP9), in the SE–CCs of the transport phloem plus in the protophloem (GFP-sporamin) or in the SE–CCs of the transport phloem, in the protophloem plus in all cells of the root tip (free GFP).
The bottom row of pictures shows that similar differences in the distribution of fluorescence are detected with all eight transgenic lines. These confocal images were taken without the propidium iodide staining that was used for the top row images. The horizontal insert shows the GFP-fluorescence in the two vascular strands of the main root of a GFP-ubiquitin plant (merged presentation of GFP-fluorescence and transmitted light picture). Scale bars are 100 μm for the root tips in the top row, 50 μm for the root tips in the bottom row and 40 μm for the horizontal insert.

Figure 6 shows a quantitative analysis of this observation in the main roots of up to 12 plants from each of the different lines. The average GFP/tip-distances are presented, and confirm the qualitative observation made in Figure 5 that the transgenic plants fall into three classes: the membrane-anchored class (STP9-GFP, AtSUC2-GFP, ER-GFP) which are expressed in the first CCs of the metaphloem, the soluble GFP-fusions (GFP-ubiquitin, GFP-sporamin, GFP-aequorin, GFP-patatin) which are translocated to the ends of the protophloem files, and the free (unfused) GFP which is transported out of the protophloem domain into the extreme root tip.

Figure 6.

Quantitative analysis of the distance between the most distal GFP-fluorescence and the surface of the tip in the main root of the different transgenic Arabidopsis lines.
Distances were determined under the confocal microscope in up to 12 plants from each transgenic line. Plants were analyzed as shown in the bottom row in Figure 5. The individual results (a) or the average distances (b) between the most distal GFP-fluorescence and the root tips (mean ± SD) are presented.


The approach presented in this paper uses free GFP, GFP-fusions similar to those previously used by Oparka et al. (1999; GFP-sporamin and GFP-patatin) and new GFP-fusions in a non-destructive approach for analyses of symplastic domains along the phloem path of Arabidopsis and in the terminal sink of the root tip. To this end cDNAs encoding free GFP (27 kDa) or one of seven different GFP-fusions was expressed under the control of the CC-specific AtSUC2-promoter (Imlau et al., 1999; Truernit and Sauer, 1995).

The obtained results demonstrate that ER-GFP, STP9-GFP and AtSUC2-GFP do not traffic into the SEs and are restricted to the CCs (Figures 2, 3g,i and 5) confirming the cell-specificity of the AtSUC2 promoter (Stadler and Sauer, 1996). These plants represent ‘zero trafficking controls’ for the analyses of the established cell-to-cell trafficking of free GFP, and the potential cell-to-cell trafficking of soluble GFP-variants.

All soluble GFP-fusions can move from CCs into SEs

Analyses of leaves (Figure 2) and/or roots (Figure 3) from the different transgenic Arabidopsis lines showed that all soluble GFP-fusions synthesized in the cytoplasm of the CCs were able to traffic into the phloem SEs. Green fluorescence resulting from GFP-ubiquitin or GFP-sporamin fusions was detected in sink leaves, where the AtSUC2-promoter is not active (Stadler and Sauer, 1996; Truernit and Sauer, 1995; see also Figure 2). Green fluorescence resulting from the GFP-ubiquitin, GFP-sporamin, GFP-aequorin and GFP-patatin fusions was also seen in the root phloem. The root tips analyzed in this study represent terminal sinks, where fluorescent proteins released from the CCs into the SEs accumulate with time, and can thus be detected even in plants with lower expression levels. Our data demonstrate that in Arabidopsis plasmodesmata connecting CCs and SEs have a SEL of at least 67 kDa. A recent report describing the complete lack of GFP movement out of the tomato CCs (Lalonde et al., 2003) may be explained with a different SEL of the plasmodesmata connecting CCs and SEs in this species, where phloem loading occurs into the SEs (Kühn et al., 1997) and not into the CCs as shown for Arabidopsis (Stadler and Sauer, 1996; Truernit and Sauer, 1995). Alternatively, this observation in tomato may result from the extremely low activity of the rolC promoter used in these analyses. Data supporting a different SEL in solanaceous plants were also obtained by Itaya et al. (2002). These authors found no movement of a GFP–GFP-fusion from the CCs into the SEs of transgenic tobacco plants.

GFP-fusions show restricted movement out of the protophloem domain

In none of the transgenic lines expressing soluble GFP-fusions were we able to detect significant transport of the fusions beyond the protophloem terminus. In Arabidopsis, the protophloem is formed by two separate files of SEs, each composed of a single, narrow sieve tube (Bonke et al., 2003; Dolan et al., 1993). Figure 4(d) shows a longitudinal section of one of these two protophloem SE files. Its diameter of approximately 2–3 μm fits well to the diameter of the aniline blue-stained SEs shown in Figure 4(b,c). Comparison of these protophloem files and the fluorescent, GFP-labeled cells in this part of the root (Figure 4b,c) revealed that the regions showing GFP fluorescence were too wide to represent just this single protophloem SE file. All of the soluble GFP-fusions had exchanged laterally between the protophloem SEs and one or two surrounding cell layers (Figure 4c). This ‘protophloem domain’ is represented diagrammatically in Figure 7.

Figure 7.

Model illustrating the unloading zone of an Arabidopsis root.
The model summarizes the data obtained on the cell-to-cell trafficking of GFP or GFP-fusions and relates the expression pattern (tmGFP2, tmGFP9, ER-GFP) and the trafficking distances (all other constructs) to the anatomy of protophloem and metaphloem in the developing Arabidopsis root tip.
Plants with non-mobile GFP-fusions (tmGFP2, tmGFP9, ER-GFP) label exclusively the CCs of the mature metaphloem that ends about 500 μm behind the root tip. All soluble GFP-fusions (GFP-ubiquitin, etc.) as well as free GFP can move from these metaphloem CCs into the metaphloem SEs and enter eventually the two single protophloem SE files that end about 250 μm behind the root tip. From there they can be unloaded into one or two cell layers forming the protophloem unloading domain, which allows further movement into all cells of the root tip only for free GFP. In contrast, all soluble GFP-fusions are retained in this unloading domain indicating a smaller SEL between these cells and the adjacent cells of the root tip.

It appears that the protophloem of many plant species is devoid of true CCs (Eleftheriou and Tsekos, 1982), a feature that may be related to its short lifespan (see discussion in Sjolund, 1997). In Lemna roots, however, protophloem SEs and CCs form from the division of a common phloem mother cell (Melaragno and Walsh, 1976), although other species appear to lack such an obvious ontogeny (Esau and Gill, 1973; Schulz, 1994). Our observation that AtSUC2 promoter activity is absent from the root protophloem files in Arabidopsis supports observations that protophloem SEs are not accompanied by CCs. Nevertheless, cells adjoining these protophloem SEs might perform CC-like functions, such as the exchange of macromolecules radially out of the protophloem SEs. Clearly, GFP and GFP-fusions can move into these adjoining cells, but all GFP-fusions are restricted from passing beyond this nucleate cell layer. We, therefore, suggest that this protophloem domain may play a role in phloem unloading, limiting the passage of macromolecules into the main body of the root. It may also play a role as a ‘checkpoint’ for macromolecular trafficking (see also Foster et al., 2002). As mature protophloem SEs are enucleate (van Bel and Knoblauch, 2000; Esau, 1969b; Oparka and Turgeon, 1999; Sjolund, 1997), such a role would be best served by living cells surrounding the phloem.

GFP is unloaded from the transport phloem

An unexpected observation was the limited but reproducible unloading of free GFP from the transport phloem (Figure 3b,c), although this occurred to a much lower extent than the unloading of free GFP from the SEs in the root protophloem (Figure 3a). In contrast, no unloading from the transport phloem was observed for GFP-ubiquitin or for any of the larger GFP-fusions (Figure 3d,e). In previous analyses, Oparka et al. (1994) studied the unloading of CF in Arabidopsis roots after application of the dye to the cotyledons. In these studies, unloading of CF was detected mainly from the root tip but not from the transport phloem, suggesting that the number of plasmodesmata in the root transport phloem is low, or that these plasmodesmata are non-functional. Wright and Oparka (1997) also showed that in Arabidopsis the root transport phloem functions as an isolated domain that is connected to surrounding cells by plasmodesmata that are ‘held’ in a closed configuration. These plasmodesmata opened only after treatment of the root with inhibitors of transmembrane ion fluxes [cyanide-m-chlorophenylhydrazone (CCCP) or probenicid (Cole et al., 1991)], suggesting that they may function for transient periods in the symplasmic supply of assimilates to the root stele and cortex (Wright and Oparka, 1997). In a similar approach Itaya et al. (2002) showed that the fluorescent tracer fluorescein, which was loaded into the phloem of a mature leaf, was confined to the SE–CC complex in stems and source leaves of tobacco. From these data they suggested that the SE–CC in the transport phloem represents a symplastically isolated domain.

The plants used in the present study were transgenic, rather than pulse labeled [as in Wright and Oparka (1997) or in Itaya et al. (2002)], and the observed limited movement of free GFP out of the transport phloem of mature roots is a clear indication for the existence of a symplastic path for the lateral unloading of assimilates. In addition, photoassimilates may be unloaded into the apoplast of the transport phloem by plasma membrane-localized transport proteins. This process has previously been postulated to be catalyzed by the AtSUC2 sucrose transporter (Truernit and Sauer, 1995).

Phloem differentiation in the root tip

Quantitative analyses of the trafficking of GFP and GFP-fusions into the root tip (Figure 6) demonstrated that GFP and all GFP-fusions passing from the CCs into the SEs trafficked toward the very end of the protophloem. The protophloem, by definition, is the ‘first formed’ phloem (Esau, 1977), and in the roots of most species occurs as vertical files of SEs. In Arabidopsis the most mature SEs are found at about 250 μm from the root tip (Zhu et al., 1998). This mature protophloem sieve tube appears to conduct the bulk of phloem unloading in root tips (Giaquinta et al., 1983; Oparka et al., 1994; Schulz, 1994; Zhu et al., 1998) and, although short lived, appears to be specialized for this function. Proximal to the protophloem files, and internal to them, the metaphloem (‘late-formed’ phloem) develops and contains the first true CCs (Esau, 1969a). Our observations of GFP movement into the root protophloem, and the appearance of AtSUC2 expression in metaphloem CCs, correlates perfectly with the progressive development of protophloem and metaphloem (see Figure 7).

The root tip represents a large symplastic domain, where non-targeted cell-to-cell movement of proteins can occur

Our results on the cell-to-cell transport of free GFP in roots confirm a recent report of Meyer et al. (2004), where free GFP (together with a membrane-anchored GFP-variant; tm-GFP for transmembrane-GFP) was expressed under the control of the AtSUC3-promoter. This promoter is very active in several Arabidopsis sink tissues, one of them being the epidermal cell layer of the root tip. This cell layer was labeled with high specificity in AtSUC3-promoter::tm-GFP plants. In contrast, a massive movement of free GFP out of this cell layer into the other cells of the root tip was observed in AtSUC3-promoter::GFP plants. These data and the result from the present paper show that GFP can undergo cell-to-cell transport in roots. Clearly, this transport is not unidirectional (e.g. from the phloem toward the root surface). It rather follows the concentration gradient of GFP, can occur in either direction and is, therefore, driven by diffusion.

Implications for macromolecular signaling and transport in the phloem

Another important feature of the presented work is the demonstration that CC-synthesized proteins of up to 67 kDa can enter the translocation stream non-specifically. The fusion proteins used here are ‘xenobiotic’ in the sense that they are not normal constituents of the CC cytoplasm. Therefore, it is reasonable to ask whether such macromolecular trafficking reflects the behavior of CC-derived proteins that enter the phloem for long-distance signaling purposes. Several publications have drawn attention to the entry of macromolecular signals into SEs. Such signals include mRNAs, gene silencing signals and transcription factors that have been suggested to pass selectively through the plasmodesmata connecting CC and SE by a specific, chaperone-mediated transport process (Citovsky and Zambryski, 2000; Crawford and Zambryski, 1999; Lucas et al., 2001; Ruiz-Medrano et al., 2001).

An alternative view is that the transport of many proteins and solutes in SEs is regulated not at the level of CC–SE plasmodesmata, but rather within the conducting phloem itself (Ayre et al., 2003; Fisher et al., 1992; Oparka and Santa Cruz, 2000). In an early study, Fisher et al. (1992) radiolabeled amino acids in wheat leaves and detected a large range of radiolabeled proteins (10–79 kDa) in sieve tube exudate. They proposed a highly selective regulation of protein removal from SEs of the pathway phloem and non-selective protein removal from the SEs in sink tissues. As a third model, Imlau et al. (1999) suggested that movement of soluble CC proteins into the SE may represent the ‘default’ pathway, unless a given protein has a retention signal that locates it to a given domain within the CC or, alternatively, targets it to a specific site within the SE. The implication of this hypothesis is that molecules synthesized within CCs may be continually lost to the translocation stream.

This kind of passive, diffusional cell-to-cell trafficking through plasmodesmata has recently been shown for the non-cell-autonomous Arabidopsis transcription factor LEAFY (LFY) (Sessions et al., 2000; Wu et al., 2003) and for transcription factors from maize (Lucas et al., 1995) or Antirrhinum majus (Perbal et al., 1996). Wu et al. (2003) provided strong evidence that the movement of LFY and LFY-GFP fusions from L1 into underlying cell layers of the Arabidopsis apex is non-targeted. Specific movement signals could not be identified and the observed cell-to-cell movement of LFY is thus likely to be driven by diffusion (Wu et al., 2003). Based on parallel analyses of the trafficking of APETALA1 (AP), of AP-GFP fusions and of GFP–GFP dimers these authors suggested that diffusion-driven, non-targeted cell-to-cell movement of proteins may represent a general mechanism and that the extent of this trafficking may only be reduced by subcellular trapping, e.g. in the nuclear compartment or by the formation of large complexes.

In contrast, plasmodesmata-trafficking of the maize transcription factor KNOTTED1 (KN1) seems to be targeted, and it has been shown that KN1 can increase the SEL of plasmodesmata and enables the transport of its own mRNA (Lucas et al., 1995). Similarly, the Arabidopsis SHORT-ROOT protein (SHR) moves from the stele into a single layer of adjacent cells, where it enters the nucleus (Nakajima et al., 2001). Analyzes with in frame-fusions of GFP to the coding region of SHR showed that SHR-GFP fusion does traffic from the stele into this single, adjacent cell layer representing the endodermis. The authors concluded that SHR (60 kDa) might need a special transit signal that widens the plasmodesmata or initiates a transient unfolding/refolding of the protein.

Also some of the macromolecules shown to traffic from CC to SE appear to depend on targeted trafficking and to be transported no further than into the SE parietal layer. Examples are specific enzymes of the alkaloid biosynthetic pathway (Bird et al., 2003). Moreover, the phloem exudate of Cucurbits appears to be replete with proteins that can ‘gate’ the plasmodesmata present in mesophyll cells to a higher-than-normal SEL (Balachandran et al., 1997), suggesting that many phloem proteins have plasmodesmata-modifying functions. However, most of the proteins translocated in the phloem (Fisher et al., 1992) appear to be smaller than the passive SEL ‘cutoff’ (67 kDa) demonstrated here for Arabidopsis, suggesting that they may enter the SE from the CC by diffusion. A clear challenge for the future will be to demonstrate whether or not or to what extent the specialized plasmodesmata that connect SE and CC (Mezitt and Lucas, 1996; Oparka and Turgeon, 1999; Schulz, 1998; Sjolund, 1997) require to be gated in order to permit macromolecular trafficking in the phloem.

Our analyses demonstrate that GFP and GFP-fusions are powerful tools for studying phloem transport and symplasmic domains under non-invasive conditions in intact plants. In the future, it will be interesting to induce biotic and abiotic stresses in such transgenic plants and to examine the effects on macromolecular trafficking via the phloem.

Experimental procedures

Strains and growth conditions

If not otherwise indicated, Arabidopsis thaliana plants (ecotype C24) were grown in potting soil in the greenhouse. For root tip analyses plants were grown under sterile conditions on vertical Petri plates on Murashige–Skoog medium (pH 5.8) containing 1% phyto agar (Duchefa, Haarlem, the Netherlands; Murashige and Skoog, 1962). Roots were grown on the agar surface by incubating the plates in a near-vertical position under long-day conditions in 16 h light/8 h dark regime at 22°C and 70% relative humidity for 14 days. For microscopy of leaves, 3-week-old seedlings were transferred to soil and grown for two more weeks in a growth chamber (long-day conditions, 22°C, 70% humidity). Agrobacterium tumefaciens strain GV3101 (Holsters et al., 1980) was used for plant transformation. All cloning steps were performed in Escherichia coli strain DH5α (Hanahan, 1983).

Construction of vectors for stable plant transformation

Plants expressing free GFP under the control of the 900-bp AtSUC2 promoter (pEPS1) have been described previously (Imlau et al., 1999). For construction of GFP-ubiquitin, GFP-sporamin and GFP-patatin fusions we generated the pUC19-based vector pGA04 that contained the 900-bp AtSUC2 promoter fragment from pEPS1 (Imlau et al., 1999) as an SphI/NcoI fragment, followed by GFP (NcoI/SacI) and the nopaline synthase terminator (SacI/EcoRI). At the very 3′-end of the GFP open-reading frame (ORF) BglII and XbaI cloning sites were introduced by a polymerase chain reaction (PCR).

For construction of the GFP-ubiquitin fusion the complete ORF of a Plantago major ubiquitin 1 (PmUBI1; accession no.: AJ841743) was PCR-amplified from the vector pPP35 (N. Sauer and M. Gahrtz, unpublished data) and BamHI (5′-end) and XbaI (3′-end) cloning sites were introduced. Using these sites the ubiquitin ORF was cloned into the BglII/XbaI sites at the 3′-end of the GFP-ORF in pGA04, yielding the plasmid pGA04-Ubi. From this plasmid a HindIII/SacI fragment was excised and cloned into pGPTV-Bar (Becker et al., 1992), yielding pGPTV-Ubi.

For construction of the GFP-sporamin fusion the ORF of sporamin (accession no.: P14715) lacking the N-terminal 37 amino acids signal sequence responsible for targeting to the ER (Hattori et al., 1989) was PCR-amplified from the vector pIM023 (Hattori et al., 1985) and BglII (5′-end) and XbaI (3′-end) cloning sites were introduced. Using these sites the sporamin ORF was cloned into the BglII/XbaI sites at the 3′-end of the GFP-ORF in pGA04, yielding the plasmid pGA04-Spor. From this plasmid a HindIII/SacI fragment was excised and cloned into pGPTV-Bar (Becker et al., 1992), yielding pGPTV-Spor.

For construction of the GFP-patatin fusion the ORF of patatin (accession no.: A24142) lacking the N-terminal 23 amino acids signal sequence responsible for targeting to the ER (Mignery et al., 1984) was PCR-amplified from the vector pPATB2 (Stiekema et al., 1988) and BglII (5′-end) and XbaI (3′-end) cloning sites were introduced. Using these sites the patatin ORF was cloned into the BglII/XbaI sites at the 3′-end of the GFP-ORF in pGA04, yielding the plasmid pGA04-pat. From this plasmid the C-terminal XbaI site was removed and a new XbaI site was introduced at the 5′-end of the AtSUC2-promoter yielding plasmid pGA05-Pat. From this plasmid an XbaI/SacI fragment was excised and cloned into pGPTV-Bar (Becker et al., 1992), yielding pGPTV-Pat.

For construction of the GFP-aequorin (accession no. for aequorin: P07164) fusion under the control of the AtSUC2 promoter an already existing fusion was PCR-amplified from the plasmid p7rolB-GFP-AEQ (C. Plieth, Zentrum für Biochemie und Molekularbiologie, University of Kiel, Kiel, Germany) and BspHI (5′-end) and SacI (3′-end) cloning sites were introduced. Using these sites the GFP-aequorin fusion was cloned into pGA05-pat after removal of the GFP-patatin insert by a NcoI/SacI-digest. From the resulting plasmid pGA05-Aeq an XbaI/SacI fragment was excised and cloned into pGPTV-Bar (Becker et al., 1992), yielding pGPTV-Aeq.

For construction of the ER-resident GFP-variant an already existing sequence was PCR-amplified from the vector pBIN-m-gfp5-ER [contains the N-terminal signal sequence (21 amino acids) of an Arabidopsis basic chitinase (accession no.: AAM10081)] and a C-terminal HDEL-retention signal; J. Haseloff, Department of Plant Sciences, University of Cambridge, Cambridge, UK) and a BspHI cloning site was introduced into the start ATG of the resulting PCR-fragment, which was cloned into pGEM-T Easy (Promega, Madison, WI, USA), yielding plasmid pMG002. From this plasmid the modified GFP-ORF was excised with BspHI/SacI and cloned into NcoI/SacI-digested pEP/pUC (Imlau et al., 1999). A BamHI/SacI fragment harboring 2200 bp of AtSUC2-promoter sequence and the modified GFP-ORF was excised from the resulting plasmid and cloned into pGPTV-Bar (Becker et al., 1992), yielding the plasmid pMG004.

For construction of the AtSTP9-GFP fusion (tmGFP9) a genomic AtSTP9 fragment [encoding the 232 N-terminal amino acids of AtSTP9 (Schneidereit et al., 2003) and harboring the first two introns of At1g50310] was PCR-amplified and NcoI cloning sites were introduced on both ends. Using these NcoI sites the genomic AtSTP9 fragment was cloned into the unique NcoI site of plasmid pAF12 yielding the plasmid pMH4. pAF12 represents a pUC19-based plasmid that harbors the 900-bp AtSUC2-promoter of pEPS1 (Imlau et al., 1999) followed by GFP with a unique NcoI site in its start ATG. The AtSUC2-promoter/AtSTP9-GFP fragment was excised from pMH4 with HindIII/SacI and cloned into the respective sites of pAF16, which represents a modified version of pGPTV-bar (Becker et al., 1992), where the GUS-reporter gene has been removed. The resulting plasmid was named pMH5a.

For construction of the AtSUC2-GFP fusion (tmGFP2) the entire AtSUC2 gene (At1g22710; including all three introns) was PCR-amplified and BspHI cloning sites were introduced on both ends. Using these BspHI sites the AtSUC2 gene was cloned into the compatible NcoI site in the start ATG of GFP in pEPS/pUC (Imlau et al., 1999), yielding plasmid pTF5004. In this plasmid the unique SphI site at the 5′-end of the AtSUC2-promoter was replaced by SacI, yielding plasmid pTF5008. From this plasmid the AtSUC2-promoter/AtSUC2-GFP fragment was excised by SacI and cloned into pAF16, yielding the plasmid pTF5010.

Plant transformation

The seven plasmids (pGPTV-Ubi, pGPTV-Spor, pGPTV-Aeq, pGPTV-Pat, pMG004, pMH5a and pTF5010) were used to transform Agrobacterium tumefaciens strain GV3101 (Holsters et al., 1980) and finally for transformation of Arabidopsis thaliana C24 WT by dipping (Clough and Bent, 1998).

Epifluorescence and confocal laser scanning microscopy

Green fluorescent protein in leaves was detected using a stereofluorescence microscope (SV11; Carl Zeiss, Jena, Germany) after excitation with light of 460–500 nm wavelengths. Emitted fluorescence was monitored using a filter permeable for wavelengths >510 nm. Photos were taken with a Sony 3CCD color video camera and a Carl Zeiss Vision KS200, 3.0 imaging software.

Roots were imaged in situ using a confocal laser scanning microscope (CLSM, Leica TCS SP II; Leica Microsystems, Bensheim, Germany). For cell wall staining, roots grown on the agar surface were covered with a drop of 0.5% propidium iodide and incubated for 10 min at room temperature. After two washes with water, roots were imaged while still in their Petri dishes. GFP was excited by 488 nm light produced by an Argon laser and observed using a detection window from 497 to 526 nm. Propidium iodide-stained cell walls were detected with the argon laser 488 nm line and a detection window of 595–640 nm.

Callose was detected by staining roots with 0.01% aniline blue made up in 0.07-m phosphate buffer (pH 7.5) and imaging using the epifluorescence microscope by excitation at 330–380 nm, with emitted light being monitored above 420 nm.

For GFP/tip-distance measurements, seedlings were grown on plates for 10 days and main root tips were imaged as described above. Distances of the GFP signals from the root tip were calculated using the Leica Confocal Software.


We thank Marina Henneberg and Anja Schillinger for excellent technical assistance. This work was supported by the Deutsche Forschungsgemeinschaft (Grant Sa 382/8 to N.S.) and the Scottish Executive Environment and Rural Affairs Department (SEERAD; grant-in-aid to K.J.O.).