Cell-to-cell transport via the lumen of the endoplasmic reticulum


(fax +61 (2) 9351 4771; e-mail robyn.overall@sydney.edu.au).


Plasmodesmata are plasma membrane-lined channels through which cytoplasmic molecules move from cell-to-cell in plants. Most plasmodesmata contain a desmotubule, a central tube of endoplasmic reticulum (ER), that connects the ER of adjacent cells. Here we demonstrate that molecules of up to 10.4 kDa in size can move between the ER lumen of neighbouring leaf trichome or epidermal cells via the desmotubule lumen. Fluorescent molecules of up to 10 kDa, microinjected into the ER of Nicotiana trichome cells, consistently moved into the ER and nuclei of neighbouring trichome cells. This movement occurred more rapidly than movement via the cytoplasmic pathway. A fluorescent 3-kDa dextran microinjected into the ER of a basal trichome cell moved into the ER and nuclei of epidermal cells across a barrier to cytoplasmic movement. We constructed a 10.4-kDa recombinant ER-lumenal reporter protein (LRP) from a fragment of the endogenous ER-lumenal binding protein AtBIP1. Following transient expression of the LRP in the ER of Tradescantia leaf epidermal cells, it often moved into the nuclear envelopes of neighbouring cells. However, green fluorescent protein targeted to the ER lumen (ER-GFP) did not move from cell to cell. We propose that the ER lumen of plant cells is continuous with that of their neighbours, and allows movement of small ER-lumenal molecules between cells.


Plant cells traffic macromolecules through plasmodesmata, the membrane-lined channels that traverse the walls between neighbouring cells. These channels not only connect the plasma membrane and cytoplasm of adjacent cells, but also contain a central tube of endoplasmic reticulum (ER), the desmotubule, that links the ER networks of adjacent cells to form an ER-continuum throughout the tissue. Intercellular movement of small molecules <1 kDa (Tucker, 1982; Erwee and Goodwin, 1983; Wolf et al., 1989) and large proteins up to 67 kDa (Oparka et al., 1999; Stadler et al., 2005) through the cytoplasmic annulus is well documented, as is the cell-to-cell movement of large molecules along the membranes of the ER (Grabski et al., 1993; Martens et al., 2006). This study investigates whether intercellular transport of small molecules occurs between the ER lumen of neighbouring plant cells.

Ultrastructural images of desmotubules, showing them as closed, rod-like structures, led to the conclusion that they are incapable of transport through their lumenal space (Hepler, 1982; Overall et al., 1982). Adding weight to this argument are the numerous observations that the 27-kDa green fluorescent protein targeted to the ER lumen (ER-GFP) does not to move from cell to cell (Crawford and Zambryski, 2000; Stadler et al., 2005; Martens et al., 2006). Although ER-GFP has been reported to move very occasionally from cell to cell in wild-type Nicotiana benthamiana (tobacco) plants (Guenoune-Gelbart et al., 2008), the overwhelming consensus is that the intercellular transport of molecules >27 kDa through the desmotubule is not a regular occurrence. The question remains whether this is true for smaller molecules. The ER has been proposed as a pathway for the movement of water (Zhang and Tyerman, 1997) and photosynthates (Sawidis et al., 1989; Gamalei et al., 1994) between plant cells, and there are indications that small fluorescent molecules can move directly between putative endomembrane systems of neighbouring cells (Lazzaro and Thomson, 1996; Cantrill et al., 1999). However, as yet there is no definitive evidence for intercellular transport of small molecules via the ER.

In this study, we used two complementary approaches to investigate whether the ER is a conduit for cell-to-cell transport of small lumenal molecules. First, we employed the Nicotiana leaf secretory trichome system previously used for studying directed cell-to-cell movement (Oparka et al., 1991; Waigmann and Zambryski, 2000). Using Nicotiana plants expressing GFP as an ER marker, we microinjected red fluorescent dyes directly into the ER of trichome cells and observed their movement into the ER and nuclei of neighbouring cells. Recently, Christensen et al. (2009) demonstrated a barrier for movement of cytoplasmic molecules from basal Nicotiana trichome cells into leaf epidermal cells. After microinjection of a red-fluorescent dextran into the ER of a basal trichome cell, we observed movement across this barrier into the ER and nuclei of underlying epidermal cells. Our second approach was to transiently express a small ER-targeted recombinant protein in Tradescantia virginiana leaf epidermal cells. This lumenal reporter protein (LRP) consisted of a signal peptide, a fragment of the endogenous Arabidopsis ER lumenal binding protein (AtBIP1), an HDEL retention sequence and an haemagglutinin (HA) tag. Following transient expression in the ER of leaf epidermal cells, we used immunofluorescence to observe LRP movement into neighbouring cells. We propose that intercellular transport of small ER-lumenal molecules occurs routinely between the ER of adjacent cells, and that this may be an important transport pathway for plants.


ER morphology in Nicotiana leaf trichomes

Secretory trichomes of Nicotiana emerge from a single epidermal cell, extend perpendicularly from leaf margins and are composed of four large, vacuolate cells (Figure 1a,b) that transport metabolites from the leaf to a densely cytoplasmic gland cell at the trichome’s apex via plasmodesmata (Waigmann and Zambryski, 2000). Using plants constitutively expressing ER-GFP, we observed the three distinct configurations of plant ER, the nuclear envelope, the highly mobile ER in transvacuolar strands and the relatively stable cortical network (Figure 1c).

Figure 1.

Nicotiana trichome cells expressing GFP targeted to the endoplasmic reticulum (ER-GFP).
(a) Nicotiana tabacum leaf margin trichomes in a maximum projection of a confocal z-series. Prostate (arrowhead) and large secretory trichomes (arrows) coexist on the leaves.
(b) A typical secretory trichome consists of a terminal gland cell supported by four large vacuolate cells arising from a solitary supporting epidermal cell.
(c) Each vacuolate trichome cell contains a fine meshwork of cortical ER tubules (arrowheads), the nuclear envelope (arrow) and ER in transvacuolar strands (double-headed arrow).
Scale bars: 200 μm (a); 10 μm (c).

ER-based cell-to-cell movement of small molecules in Nicotiana leaf trichomes

To investigate whether the ER is a conduit for unrestricted cell-to-cell transport of molecules smaller than GFP, we microinjected red-fluorescent dyes into the ER of trichome cells and examined whether they moved into neighbouring cells. The injected dyes ranged in size from 577 Da (lissamine rhodamine B) and 731 Da (Alexa 568 hydrazide), up to a 3-kDa biotinylated tetramethylrhodamine (TMR) conjugated-dextran and a 10-kDa Alexa 568-conjugated dextran. Iontophoretically injecting the dyes into trichome cells did not alter cytoplasmic streaming. Non-injected cells, imaged under similar conditions to injected cells, showed limited bleed-through from the green emission channel (GFP) into the red emission channel (Figure 2a). The image-capture settings were modified to reduce this bleed-through so that any fluorescence in subsequent images would be from the injected dyes only. In the red emission channel of Figure 2a, fluorescence was limited to chloroplasts the size, brightness and morphology of which made them identifiable in all cells. Sixty-eight injections were performed, and of these 16% were vacuolar (Figure 2b) and 59% were cytoplasmic injections where the dyes concentrated in the nuclei as well as diffusely labelling the cell cortex (Figure 2c). In 25% of injections, dyes entered the ER lumen and were visible in a reticulate configuration that co-localized with ER-GFP in the cortical ER network (Figure 2e). At higher magnification (Figure 2f), the similarity between the tubular networks containing ER-GFP and a 3-kDa TMR-dextran is clearly apparent. The dextran also concentrated in a subset of tubules devoid of ER-GFP, as well as small round structures, approximately 1.7 μm in diameter (Figure 2f). Given the size of these round structures (Dupree and Sherrier, 1998) and their proximity to the cortical ER network (Brandizzi et al., 2002; Sparkes et al., 2009b), they could be Golgi complexes or ER-derived vesicles containing dye from the ER. Occasionally, the fluorescent dyes leaked into the cytoplasm during injection, but the ER, containing a higher concentration of the dye, remained significantly brighter. This demonstrates that it is possible to inject small fluorescent dyes into the ER of trichome cells.

Figure 2.

 Simultaneous dual-channel confocal imaging of red fluorescent dyes microinjected into Nicotiana leaf trichomes expressing GFP targeted to the endoplasmic reticulum (ER-GFP). The left image represents GFP, the centre displays red fluorescence from injected dyes, and the right image shows the green and red channels merged together. All images are single optical sections.
(a) Non-injected control. Limited GFP bleed-through is visible in the red channel where the bright organelles are autofluorescent chloroplasts (arrowheads).
(b) Injection of lissamine rhodamine B into a vacuole. ER-GFP is in the nuclear envelope, cortical regions and transvacuolar strands. In the merged image, dark regions between the ER-GFP fluorescence and the dye-filled vacuole indicate the cytoplasm (arrowheads).
(c–d) Microinjection of a 3-kDa biotinylated tetramethylrhodamine (TMR)-dextran into the cytoplasm of a basal trichome cell and its movement into the upper sub-basal cell.
(c) Concentration of the TMR-dextran in the nucleus (arrows) and cortical cytoplasm of the injected cell. Its localisation does not match the reticulate pattern of the ER-GFP (arrowheads).
(d) TMR-dextran moved into the cytoplasm of the cell above the injected cell. It is at the cortex, in cytoplasmic strands (arrowheads) and the nucleus (arrows). The cytoplasm of the injected cell is marked by a double-headed arrow.
(e–h) Microinjection of the 3-kDa biotinylated TMR-dextran into the ER of a sub-basal trichome cell and its movement into the neighbouring basal and sub-apical cells.
(e) The reticulate pattern observed following injection of the dextran into a trichome cell is almost identical to the lace-like ER-GFP meshwork.
(f) Enlargement of the boxed region in (e) illustrates the co-localization of the GFP and TMR markers in cortical ER tubules (arrowheads), indicating that the injected TMR dextran moved throughout the ER network. The dextran also accumulated in a subset of tubules lacking ER-GFP (double-headed arrows) as well as in small endomembrane-like structures (arrows) that may be Golgi.
(g) Movement of the TMR-dextran into the sub-apical cell above the injected cell. The dextran is in the nucleus (arrows) and is also in thicker ER strands (arrowheads) and faint lace-like configurations similar to those of the ER-GFP fluorescing tubules (see insets for more detail).
(h) Movement of the TMR-dextran into the basal cell below the injected cell. The dextran is in the nucleus (arrows) as well as in ER-like tubules (see insets).
Scale bars: 10 μm (a–h); 5 μm for the insets of (g) and (h).

We monitored where the fluorescent dyes moved after microinjection into the ER (14 cells) or cytoplasm (21 cells). In both cases, the dyes consistently moved into neighbouring cells (Table 1). Movement was faster following ER injections, occurring in 2–3 min, whereas cytoplasmic injections took several minutes longer. The pattern of fluorescence in the neighbouring cells differed between cytoplasmic and ER injections. Following the injection of a 3-kDa biotinylated TMR-dextran into the cytoplasm of a trichome cell (Figure 2c), it moved into the cytoplasm and nucleus of its upper neighbour within 11 min (Figure 2d). Conversely, after the dextran was injected into the cortical ER of a sub-basal cell (Figure 2e), it was faintly visible in lace-like patterns corresponding to ER-GFP fluorescing cortical ER tubules of both its upper (Figure 2g) and lower (Figure 2h) neighbours (4 and 9 min, respectively). This was not bleed-through from the green channel, as the bright GFP fluorescence at the top of the cell in Figure 2h was not visible in the red channel. The dextran was also in the nuclei of these neighbouring cells. Whereas the nuclear envelopes of the ER-injected cell and those of its neighbours were occasionally observed to be brighter than the rest of the nuclei in the red channel soon after injection, this was difficult to record as the nuclei rapidly filled with the dye, and the nuclear envelope was obscured before imaging could take place. The size of the dyes (ranging from 577 Da to 10 kDa) was no barrier to movement, as each diffused from the cytoplasm or ER of the injected cells into the cytoplasm or ER of their neighbours. This demonstrates that cell-to-cell transport of molecules <10 kDa can occur between neighbouring trichome cells via an ER lumenal continuum.

Table 1.   Microinjection of fluorescent dyes into Nicotiana trichomes expressing ER-GFP. The number of times dyes moved from cell to cell following an endoplasmic reticulum (ER) or cytoplasmic injection is listed on the left, and the total number of injections is displayed in brackets
Lissamine rhodamine B (577 Da)Alexa 568 hydrazide (731 Da)3-kDa biotinylated TMR dextran10-kDa TMR dextranTotalPercentage of movement (%)
ER5 (5)2 (2)5 (5)2 (2)14 (14)100
Cytoplasmic6 (7)5 (6)7 (7)1 (1)19 (21)90

ER-based cell-to-cell movement across a selective barrier between Nicotiana leaf trichomes and the epidermis

To emphasize that microinjected fluorescent dyes move directly from the ER network of one cell into the ER network of another, we injected a 3-kDa biotinylated TMR-dextran into the ER of a basal trichome cell and monitored its movement into underlying epidermal cells. Following microinjection into a basal trichome cell (Figure 3a), the dextran localized in a net-like pattern similar to that of the cortical ER (Figure 3b). Dye moved apically, and accumulated in the nuclei of three tipward cells, and also moved basally into the epidermis (Figure 3c), where it appeared within the nuclei of the supporting epidermal cell and those of its neighbours (Figure 3c,d). Movement of the dextran into the nucleus of the sub-basal cell was observed within 2 min, and movement into the epidermal cells was observed within 7 min. Although the dextran appears to have leaked into the cytoplasm of the injected cell, it was only faintly, if at all, visible in the cortical regions of the epidermal cells, which appeared predominantly green (Figure 3c,d). This suggests that the trichome was typical of those described by Christensen et al. (2009), with a cytoplasmic barrier to the epidermis, and that transport to epidermal nuclei was via the ER.

Figure 3.

 Simultaneous dual-channel confocal imaging of a 3-kDa biotinylated tetramethylrhodamine (TMR)-dextran microinjected into the endoplasmic reticulum (ER) of a Nicotiana tabacum trichome cell, and its movement into adjacent trichome and epidermal cells. Single optical sections represent ER-GFP (left), TMR-dextran (centre) and merged images (right).
(a) Low-magnification image of the trichome soon after the basal cell (arrow) was microinjected. The chloroplasts and vacuoles from the gland cells are autofluorescent.
(b) The TMR-dextran is in an ER-like configuration in the cortex of the injected cell that matches the organisation of the cortical ER and subcortical strands (arrows). It is also in a subset of tubule-like structures that contain lower levels of ER-GFP (double-headed arrows). Small, round, endomembrane-like compartments could be Golgi complexes (arrowheads).
(c) Montage of a sequence of images taken after microinjection indicating the bi-directional movement of the TMR-dextran from the injected cell into nuclei of the apical, sub-apical, sub-basal and three epidermal cells (arrows).
(d) Injected TMR-dextran moved into the supporting epidermal cell and its neighbouring epidermal cells. The nucleus of the epidermal cell to the left, surrounded by ER-GFP in the nuclear envelope, contained a low concentration of TMR dextran (arrow).
(e) For correlative microscopy, the trichome was fixed, embedded in resin and oriented into the correct plane for sectioning. The nuclei of the supporting epidermal cell and neighbouring epidermal cell (arrows) were targeted for immunolabelling and high-resolution analysis.
Scale bars: 200 μm (a); 10 μm (b); 20 μm (c–e).

To confirm this, the cells were immunogold-labelled for the biotin tag conjugated to the injected dextran, and imaged using transmission electron microscopy (TEM; Figure 3e). We compared the anti-biotin labelling of the epidermal cells with cells of a non-injected control and a control in which the primary antibody had been omitted (secondary control) (Figure 4a–l). The nanogold particles of the secondary antibody were enhanced in size for easy viewing. Few gold particles were found in the nuclei (Figure 4a), nuclear envelopes (Figure 4e) or ER (Figure 4i) of secondary control cells. Endogenous biotin was found in nuclei and chloroplasts of the non-injected control (Figure 4b), and at lower levels in the ER (Figure 4j), but was not observed in the nuclear envelope (Figure 4f). In contrast, biotin was in the nuclear envelope (Figure 4g,h) and ER (Figure 4k,l) of both the supporting epidermal cell and the epidermal cell to the right, below the injected trichome of Figure 3. A high concentration of biotin was found in the nucleus of the supporting epidermal cell (Figure 4c), corresponding to the red nucleus containing the biotinylated TMR dextran visible in Figure 3c. Biotin was also present, albeit at a lower concentration, in the nucleus of the epidermal cell to the right (Figure 4d). In all cells, biotin was rarely in cytoplasmic regions devoid of ER tubules.

Figure 4.

 Immunogold labelling of the biotinylated dextran injected into the basal trichome cell of Figure 3, and that moved into the epidermis below.
(a–l) Labelling in the supporting epidermal cell and epidermal cell to the right is compared with a secondary antibody control and non-injected control. The presence of biotin is indicated by large, black, gold-enhanced particles.
(a–d) Biotin levels in nuclei. (a) Few gold particles were found in the nuclei of the secondary antibody control. (b) Endogenous biotin was found in the nuclei of the non-injected control. (c) The nucleus of the supporting epidermal cell contained a high concentration of biotin. (d) This was reduced in its neighbour to the right. Biotin was in the chloroplasts (*) of all cells.
(e–h) The boxed regions of (a–d) were magnified to show details of biotin labelling (arrowheads) in the nuclear envelope (double-headed arrows mark the position of the nuclear envelope). (e) No gold particles were found in the nuclear envelope of the secondary antibody control. (f) Biotin was absent from the nuclear envelopes of the non-injected control cells. (g) Biotin was at low levels in the nuclear envelopes of the supporting epidermal cell and (h) the cell to the right of the supporting epidermal cell.
(i–l) Biotin labelling in the endoplasmic reticulum (ER; arrows mark position of the ER). (i) In the secondary antibody control, no gold particles were found in the ER lumen. (j) Endogenous biotin of the non-injected control was occasionally located in the ER. (k) Biotin was found in the ER of both the supporting epidermal cell and (l) its neighbour to the right.
(m–p) Colloidal gold labelling (arrowheads) in the ER and nuclear envelope (double-headed arrows mark the position of the nuclear envelope) of the epidermal cell to the left of the supporting epidermal cell (m, o, p) or a control cell (n).
(m) Anti-GFP labelling of ER-GFP: 10-nm colloidal gold particles are found in the nuclear envelope and ER.
(n) Endogenous biotin in the ER of control cells is marked by 5-nm gold particles.
(o) The 5-nm colloidal gold particles indicate biotin in the lumen of the ER.
(p) Biotin in the nuclear envelope is indicated by 5-nm colloidal gold particles.
Scale bars: 2 μm (a–d); 300 nm (e–h); 200 nm (i–l); 100 nm (m–p).

The density of gold particles could not be accurately determined in gold-enhanced samples, so we analysed sections immunolabelled with secondary antibodies conjugated to 5- or 10-nm colloidal gold particles. These sections were of the epidermal cell to the left of the supporting epidermal cell into which the dextran had moved following microinjection (Figure 3d). As a positive control, GFP labelled with a 10-nm gold-conjugated secondary antibody was located almost exclusively within the nuclear envelope and ER (Figure 4m; Table 2). Biotin (5-nm gold particles) was in high densities in the ER of the epidermal cell (Figure 4o; Table 2). It was also in the nuclear envelope (Figure 4p) at lower densities than in the ER, and at even lower densities in the nucleus (Table 2). This density gradient suggests that the dextran was moving from the ER, with a low volume and high dextran density, through the nuclear envelope and into the large volume of the nucleus, where it was comparatively dilute. This correlates with the faint red fluorescence of the nucleus, as seen in the single confocal section (Figure 3d).

Table 2.   Immunogold labelling of a 3-kDa biotinylated tetramethylrhodamine (TMR)-dextran in an epidermal cell, two cells from the injected basal trichome cell, as well as a control mesophyll cell
TreatmentNucleusNuclear envelopeERCytoplasm
  1. Average densities of gold particles per square micrometre of cellular compartment ± standard errors for at least four replicates.

GFP in epidermal cell2.9 ± 0.8524.1 ± 85.5446.8 ± 63.48.0 ± 2.5
Biotin in epidermal cell27.6 ± 5.7211.3 ± 67.5494.6 ± 109.96.6 ± 2.5
Biotin in control cell8.0 ± 4.061.9 ± 27.4266.9 ± 71.14.3 ± 2.5

As a control for the presence of endogenous biotin within the cells, we imaged a cell on the other side of the leaf that had been labelled with the biotin antibody (Figure 4n). Biotin was in the ER and, to a lesser extent, in the nuclear envelope and nucleus of this cell. Biotin densities were significantly lower than in the epidermal cell (< 0.05), and most likely indicate endogenous biotin levels within Nicotiana leaf cells. Altogether, our data indicate that the 3-kDa biotinylated TMR-dextran moved from the ER of the injected basal trichome cell into the ER, nuclear envelope and nucleus of the epidermal cells via an ER transport pathway.

Expression and intercellular movement of a lumenal reporter protein in Tradescantia leaf epidermal cells

To investigate whether small ER-lumenal proteins move into neighbouring cells via this ER transport pathway, we constructed a 10.4-kDa LRP, transiently expressed it in T. virginiana leaf epidermal cells and assessed whether it moved from cell to cell by immunofluorescence with antibodies against its HA tag. Following co-bombardment of the LRP and ER-GFP, transformed cells that contained gold particles lodged in their nuclei expressed both proteins in the cortical ER (Figure 5a), transvacuolar strands and nuclear envelope (Figure 5b). Expression of the LRP was consistently strong in nuclear envelopes when co-expressed with ER-GFP (Figure 5b,c) or expressed individually (Figure 5d). However, it was often difficult to detect in the cortical ER, which sometimes appeared patchy as a result of the fixation and detergent extraction treatments. Therefore, the presence of LRP in cells was determined by scrutinising the fluorescence levels of their nuclear envelopes.

Figure 5.

 Expression and movement of the lumenal reporter protein (LRP) in Tradescantia virginiana leaf epidermal cells. Images in the left column show anti-haemagglutinin (anti-HA) labelling of LRP, whereas the second column shows either fixed GFP targeted to the endoplasmic reticulum (ER-GFP) (a–c) or anti-HDEL labelling (d, e). The third column is the merged image of the first two columns. Transmitted light images in the column to the far right show the position of gold particles in the nuclei (arrows). All images represent single focal planes.
(a) Co-expression of LRP and ER-GFP in the cortical ER of a T. virginiana leaf epidermal cell. A crack in the cell (arrowhead), made by plunging the material into liquid nitrogen, facilitated antibody access.
(b) The cell in (a) at a lower focal plane. Expression of the LRP and ER-GFP was brightest in the nuclear envelope. A 1-μm gold particle was found in the nucleus (arrow).
(c) Co-expression of the LRP and ER-GFP in two cells with gold particles in their nuclei (arrows). A third cell (double-headed arrowhead), which did not express ER-GFP and which lacked a gold particle, contained the LRP in its nuclear envelope, indicating that movement had occurred from an expressing cell. Other cells that were cracked open (*) did not contain anti-HA labelling, indicating that movement had not occurred into their ER.
(d) Movement of the LRP from cells not co-expressing ER-GFP. Two cells contained gold particles in their nuclei (arrows) and strongly expressed LRP in their nuclear envelopes. Two neighbouring cells (double-headed arrows) lacking gold particles in their nuclei contained a lower concentration of LRP in their nuclear envelopes, indicating movement from the expressing cell(s). Anti-HDEL labelling, present at varying concentrations in the nuclear envelopes of these four nuclei, was also in a nucleus with no anti-HA labelling (arrowhead), indicating that the LRP had not moved into it.
(e) A control cell bombarded with a gold particle lacking plasmid DNA (arrow) and its neighbours. Although anti-HA labelling was not observed, anti-HDEL labelling was found in nuclear envelopes.
Scale bars: 20 μm.

We then assessed whether the LRP moved from cells expressing the protein into neighbouring cells that lacked bombarded nuclear gold particles. Following co-expression of ER-GFP and LRP in epidermal cells, the LRP moved from 44% of expressing cells (16 from 36) into at least one neighbouring cell, whereas ER-GFP remained solely in bombarded cells (Figure 5c). The fluorescence intensity of LRP in adjacent cells was noticeably lower than in the expressing cells. However, the absence of LRP labelling in a cell could reflect a problem with antibody access, rather than a lack of movement. It may also reflect a difference in movement patterns following co-expression of the LRP with the larger, non-endogenous, ER-GFP. To test this, we expressed the LRP in cells without ER-GFP and compared anti-HA labelling with anti-HDEL labelling of endogenous ER-lumenal proteins. In gold-only controls, anti-HA labelling was negligible in cells containing anti-HDEL labelling (Figure 5e). In 56% of LRP-expressing cells (19 from 34), both HDEL and weak HA labelling were observed in nuclear envelopes of at least one adjacent cell (Figure 5d), implying that movement had occurred. In Figure 5d the LRP moved into two cells between LRP-expressing cells, but not into the cell to the very right, in which only anti-HDEL labelling was evident. This suggests that intercellular transport of ER-lumenal proteins in T. virginiana leaf epidermal cells may not be even in all directions.


The ER lumen is an intercellular transport pathway

This study has demonstrated that the ER lumen is a conduit for intercellular transport of small molecules, up to 10.4 kDa, from the lumen of one cell directly into that of neighbouring cells. Targeted microinjection of small fluorescent dyes into the ER, followed by correlative microscopy, revealed that ER-based cell-to-cell movement can even occur across a barrier to cytoplasmic movement. Intercellular movement of a 10.4-kDa LRP indicates that the ER is an important pathway for the movement of small endogenous lumenal proteins from cell to cell. This counters the uncertainty over the decades as to the function of the demsmotubule in plasmodesmata (e.g. Robards, 1976). It now appears that there are three functional pathways for intercellular transport through plasmodesmata: the cytoplasmic annulus, desmotubule membrane (Grabski et al., 1993; Martens et al., 2006) and desmotubule lumen.

Our finding that the desmotubule is an intercellular transport route directly contrasts with the conclusions of Hepler (1982) and Overall et al. (1982), whose electron microscopy images showed that desmotubules were closed. This may reflect differences in desmotubule physiology in the tissues examined, differential responses to tissue preparation or just the intermittent opening of the desmotubule. Indeed, the ultrastructure of plasmodesmata in N. clevelandii trichomes (Waigmann et al., 1997) revealed desmotubules that appeared open, suggesting that they could permit the unrestricted movement of small ER-lumenal molecules from cell to cell. Dilated desmotubules have also been observed in bundle sheath cells of Saccharum sp. (sugarcane; Robinson-Beers and Evert, 1991) and Zea mays (Evert et al., 1977), the floral nectaries of Gossypium sp. (cotton; Eleftheriou and Hall, 1983) and Berberis sp. (Overall and Blackman, 1996), and the secondary plasmodesmata of Metasequoia needles (Glockmann and Kollmann, 1996), suggesting that ER-based cell to cell movement may be widespread among plants.

Microinjecting into the ER

In this study, we specifically aimed to microinject the ER, and 25% of injections resulted in fluorescent dyes entering the ER lumen. In our previous studies, by chance alone, up to 10% of injections entered an endomembrane-like network (Cantrill et al., 1999), which we now suggest was the ER. Given the many studies that have microinjected fluorescent tracers into plant cells, it would be expected that others would also have inadvertently injected into the ER. During previous microinjection studies, between 10 and 20% of cells thought to have been injected into the cytoplasm behaved differently to the majority, and it is tempting to speculate that these were actually microinjections into the ER. For example, it could be argued that the 9.4-kDa dextran that moved from 12% of injected maize mesophyll cells, with a known size exclusion limit (SEL) of 0.1–0.8 kDa (Lucas et al., 1995), or the 10- and 20-kDa fluorescent dextrans that moved from 10.5% of injected tobacco mesophyll cells (Ding et al., 1992) could have been within the ER. ER injections might be missed if the cell cortex was not specifically observed after microinjection.

Following microinjection of fluorescent dyes into the ER of trichome cells, they rapidly diffused throughout the ER network, including the cortical tubules. This is similar to the diffusion of ER-GFP, which occurs through the cortical ER within seconds (Tolley et al., 2008). Interestingly, on several occasions we observed that the injected dyes entered tubules that contained little or no ER-GFP, raising the question as to whether all ER tubules are permeable to the diffusion of large lumenal proteins. Recently, two reticulon isoforms have been discovered in the ER membrane (Nziengui et al., 2007). Overexpression of one of these proteins, RTNB13, in Nicotiana leaf epidermal cells resulted in the constriction of cortical ER tubules, such that the normally diffuse ER-GFP (Tolley et al., 2008), or RFP-HDEL (Sparkes et al., 2009a), were restricted to discrete subdomains. Perhaps smaller molecules, like our fluorescently-tagged injected dyes, permeate through such constricted ER-tubules, just as they do through the desmotubule.

Desmotubule size exclusion limit

In T. virginiana leaves, the 10.4-kDa LRP often moved from transiently expressing epidermal cells into neighbouring cells. However, the 27-kDa ER-GFP did not move, implying that the desmotubule SEL of these cells lies between 10.4 and 27 kDa. Other studies have observed that ER-GFP does not move from cell to cell in Nicotiana leaf epidermal cells (Oparka et al., 1999; Crawford and Zambryski, 2000), and both Nicotiana (Martens et al., 2006) and Arabidopsis (Stadler et al., 2005) phloem companion cells. It is possible, then, that many desmotubules have an SEL of <27 kDa. However, ER-GFP is a foreign protein and it may well be that endogenous proteins interact specifically with other proteins to facilitate their intercellular trafficking through the desmotubule. Trafficking of larger molecules may require an active increase in the desmotubule SEL, and it is possible that this process is highjacked by the tobacco mosaic virus movement protein, which has been shown to substanitally increase the movement of the 72-kDa GFP-calreticulin from cell to cell (Guenoune-Gelbart et al., 2008). In this study, microinjecting fluorescent 10-kDa dextrans into the ER of Nicotiana trichomes resulted in their diffusion into neighbouring cells. Given that the LRP is approximately the same size as these dextrans, it is likely that it moved into neighbouring cells via diffusion.

ER cell-to-cell transport occurs when the cytoplasmic pathway is blocked

Recently, Christensen et al. (2009) demonstrated that an active barrier prevents small molecules moving from the cytoplasm of Nicotiana basal trichome cells into underlying epidermal cells. We exploited this barrier to demonstrate that a 3-kDa dextran, microinjected into the ER of a basal trichome cell, moved directly into the ER of the epidermal cells beneath. This indicates that the active barrier blocking the cytoplasmic annulus does not prevent ER-based cell-to-cell transport. This ER pathway could be an important conduit for intercellular transport or signalling when the cytoplasmic pathway is closed, such as occurs during wounding or stress. It may be that the ER is the intercellular transport pathway for calcium ions, as high cytoplasmic calcium concentrations are found to close the cytoplasmic pathway (Tucker and Boss, 1996; Holdaway-Clarke et al., 2000).

Nuclei may be linked by an ER, desmotubule and nucleoplasmic continuum

Following the microinjection of fluorescent dyes into the ER, the first obvious indication of intercellular movement was their appearance in the nuclei of neighbouring cells. Yet this also occurred following cytoplasmic injections. It is possible that in some cases, the dyes leaked from the ER of a microinjected cell into the cytoplasm, and that movement into the nuclei of neighbouring cells was via the cytoplasmic pathway. However, we carefully imaged neighbouring cells to show that following ER injections, the dyes predominantly entered the cortical ER of neighbouring cells. Of particular significance is the observation that, following microinjection of a dye into the ER of a basal trichome cell, the immunolabelled dye was found within the ER and the nuclear envelope of the epidermal cells below at the ultrastructural level. The dye was not found predominately in the cytoplasm, as would be expected if it were moving via the cytoplasmic pathway. The dye concentration in the nucleoplasm was also substantially more than that of the cytoplasm. Without a mechanism for concentrating the dye into the nucleus from the cytoplasm, the only feasible explanation is that the dye moves from the ER directly into the nucleoplasm.

Given that the nuclear envelope and nucleus share the inner nuclear membrane, there may be complexes or channels that allow the movement of molecules between the perinuclear space of the nuclear envelope and the nucleoplasm, across this membrane. However, the full protein composition of the inner nuclear membrane is not yet known (Evans et al., 2009), and such complexes are still to be identified. Alternatively, endocytosis of the inner nuclear membrane may transport molecules into the nucleoplasm. For example, Popłonska et al. (2009) suggest that protamine-type proteins, synthesized in the ER of Chara vulgaris cells undergoing spermiogenesis, move into the nucleoplasm via endocytosis of the inner nuclear membrane. Invaginations of the nuclear envelope into the nucleoplasm (Collings et al., 2000) may increase the surface area available for this transport. It is possible that the fluorescent dyes move into the nucleoplasm directly from the perinuclear space via one of these mechanisms. However, although the 10-kDa dextran entered the nucleoplasm of cells adjacent to ER-injected cells, the 10.4-kDa LRP moved into the nuclear envelopes of neighbouring cells, but did not enter their nuclei. This is likely to reflect a difference between the unregulated movement of the dextran across the inner nuclear membrane and the regulated movement of the LRP that was designed to remain in the ER. Interestingly, the LRP appeared brightest in the nuclear envelopes of cells, possibly because of the greater volume of lumenal space at the nuclear envelope, or as a result of fixation and detergent extraction treatments that could perturbe the cortical ER. As expression of LRP has proved to be effective in demonstrating intercellular ER-lumenal transport, the challenge now is to extend this approach to investigate the transport of other small ER-lumenal molecules and the regulation of this communication pathway in plants.

Experimental Procedures


Unless otherwise stated, chemicals were sourced from Sigma-Aldrich (http://www.sigmaaldrich.com).

Plant material

Nicotiana tabacum L. and Nicotiana benthamiana Domin. plants stably expressing an ER-GFP construct under the control of the CaMV 35S promoter (Haseloff et al., 1997; supplied by David Baulcombe, John Innes Centre, and David McCurdy and Ray Rose, Newcastle University, NSW Australia, respectively) and wild-type controls were used. Seeds were sterilized in dilute hypochlorite (15 min), were washed and grown on 1.2% (w/v) agar plates containing Hoagland’s solution supplemented by 3% (w/v) sucrose, 530 μm inositol and 50 μm thiamine hydrochloride for 2 weeks in a growth cabinet at 25°C with 17 h of light per day. Seedlings were transferred to potting mix and returned to the cabinet. Young leaves were used for experiments. Young, glasshouse-grown T. virginiana plants were excised above their roots, their outer leaves were removed and the basal 2 cm of the second youngest leaf was used for experiments.

Iontophoretic microinjection of tobacco trichomes

Micropipette needles were drawn from 1.2-mm borosilicate glass capillaries with inner filaments (World Precision Instruments, http://www.wpiinc.com) on a P87 puller (Sutter, http://www.sutter.com). Dye-filled needle tips were secured in a microelectrode holder and moved with a hydraulic micromanipulator (WR-30; Narishige, http://narishige-group.com) mounted on a coarse manipulator (M-35; Narishige) attached to a Zeiss Axiophot microscope (Carl Zeiss, http://www.zeiss.com). Tobacco leaf segments were immobilized at the base of coverslip-affixed drilled slides using 1% (w/v) type-VII low melting point agarose. Wells were filled with half-strength Hoagland’s solution and the cells were allowed to recover for >30 min before microinjection. Trichome cells were impaled at 45° by pushing the needle tip against the cell wall and tapping the microscope to penetrate the cell. With the micropipette connected to the active probe of the electrometer (model Intra 767; World Precision Instruments), and with a chloridized silver wire in the bathing solution as a reference electrode, dye was microinjected iontophoretically. A current pulse (1 nA, 15–30 s) was generated as a square wave from a function generator (CFG253; Tektronix, http://www.tek.com). The cells were imaged using a Bio-Rad MRC600 confocal system (Bio-Rad, http://www.bio-rad.com) attached to the Zeiss Axiophot with a 40× long working distance water-dipping lens. Using excitation at 488 and 568 nm, dual-channel images of green and red fluorescence were recorded using Kalman averaging (n = 3). Under the imaging conditions used, non-injected controls showed minimal bleed-through of GFP into the red emission channel.

Dye solutions

The following membrane-impermeable, red fluorescent dyes (all from Molecular Probes, now Invitrogen, http://www.invitrogen.com) were prepared in distilled water and microtinjected into trichome cells: lissamine rhodamine B (100 μm, 577 Da), Alexa 568 hydrazide (1 mm, 731 Da), tetramethylrhodamine conjugated to a 3-kDa dextran (1 mm), tetramethylrhodamine conjugated to biotinylated 3-kDa dextran (1 mm) and AlexaFluor 568 conjugated to a 10-kDa dextran (1 mm). The cytoplasm of selected non-injected trichomes was labelled with carboxy-SNARF-1 AM acetate (1 mm, 1 h).

Immunogold labelling of microinjected tobacco trichomes

Leaf segments with microinjected trichomes were fixed in phosphate-buffered saline (PBS) containing 3% (v/v) paraformaldehyde (ProSciTech, http://www.proscitech.com.au). After washing in PBS, material was dehydrated in 10% (v/v) ethanol at room temperature (22°C), continuing with 30% ethanol at 0°C, and concluding with 50, 70, 95 and 100% ethanol at −20°C in an EM AFS freeze substitution apparatus (Leica, http://www.leica.com). Material was flat-dish embedded and polymerized in LR Gold acrylic resin (ProSciTech) (−20°C, 24 h) under UV light. Samples were viewed with a light microscope to identify the microinjected cells, excised, remounted onto blank stubs, and trimmed and oriented so that ultrathin sections could be cut in similar planes to the confocal optical sections.

For immunolabelling, sections on EM grids were floated on droplets (20 μl) of the following solutions. Sections were blocked with PBS containing 1% (w/v) bovine serum albumin (BSA) and 0.1% (w/v) Tween-20 (30 min), rinsed briefly in PBS containing Tween and incubated in either mouse anti-biotin or rabbit anti-GFP (both from Molecular Probes), diluted to 1/100 in PBS/BSA (1 h). After washing in PBS, sections were incubated (1 h) in Alexa 594 FluoroNanogold anti-mouse Fab′ (Nanoprobes, http://www.nanoprobes.com) diluted to 1/25 or AuroProbe goat anti-rabbit IgG conjugated to 10 nm colloidal gold or AuroProbe goat anti-mouse IgG + IgM conjugated to 5-nm colloidal gold (Amersham Biosciences, now GE Healthcare, http://www.gelifesciences.com), both diluted to 1/20. FluoroNanogold-labelled sections were gold-enhanced (3 min) to enlarge the size of the 1.4-nm particles for transmission electron microscopy (TEM). Sections were post-stained with uranyl acetate and lead citrate, viewed with a CM12 transmission electron microscope (Philips, http://www.philips.com) and images were collected with a Morada CCD camera (Olympus SIS, http://www.olympus-sis.com).

To quantify the density of colloidal gold labelling, diagrams representing the positions of gold particles and organelles were traced from images (photoshop; Adobe, http://www.adobe.com). Organelle areas were measured in NIH Image (v1.63, US National Institutes of Health, http://www.nih.gov) and the number of gold particles were counted in each organelle area. If a gold particle was entirely within an area, it was included in the count for that area, but because of the size of the antibodies used, gold particles within 10 nm of the nuclear envelope or the ER were also included in the count for these regions. Average densities (± standard errors) were calculated and analysis of variance was used to analyse differences in the density of gold labelling.

Cloning of the LRP

The LRP was constructed with a signal peptide (MKTNLFLFLIFSLLLSLSSAE), originally cloned from a plasmid encoding ER-GFP (Haseloff et al., 1997), and a stable fragment (D568-Y643) of the Arabidopsis ER-luminal AtBIP1 (At5g28540, GenBank NM_122737), cloned from a cDNA library (CD4-10; Arabidopsis Biological Resource Centre, http://www.arabidopsis.org). This ubiquitous protein was selected because the fragment is predicted by PsiPred (McGuffin et al., 2000) and Sable II (Adamczak et al., 2005) to have a stable triple α-helix structure with a hydrophobic core and few exposed hydrophilic residues. The C terminus of the protein contained a linker (GGAGGA) followed by an HA epitope tag (YPYDVPDYA) and an HDEL ER retention signal, both of which are detectable by antibodies (refer to Table S1 for primer sequences and cloning strategy). The gene encoding LRP was cloned using the Gateway system (Invitrogen) into p2GW7 (Karimi et al., 2002) for expression under the cauliflower mosaic virus 35S promoter.

Biolistic bombardment

Plasmids encoding the LRP and/or ER-GFP were adsorbed onto 1-μm gold microcarriers (Bio-Rad) at 0.37 or 0.33 ng DNA per mg microcarriers, respectively, and non-DNA controls were prepared in the absence of plasmid DNA. The abaxial surfaces of T. virginiana leaf segments were bombarded using 400 kPa of pressurized helium with a custom-made gene gun (CSIRO Plant Industry, http://www.pi.csiro.au). Leaf segments were kept in the dark for two nights on moistened filter paper.

Fixation and immunofluorescence labelling

Leaf epidermal peels were prepared from bombarded segments (Barton et al., 2008), then fixed in 2% (v/v) paraformaldehyde and 0.5% (v/v) glutaraldehyde (ProSciTech) in 10 mm piperazine-N,N′-bis(2-ethane sulfonic acid) and 5 mm CaCl2 for 30 min. After rinsing in PBS, the peels were plunged into liquid nitrogen and thawed in PBS. They were extracted with 0.5% Triton-X-100 for 20 min, rinsed in PBS and blocked for non-specific antibody binding with 10% (v/v) normal goat serum (IMVS, http://www.imvs.sa.gov.au) and 3% (w/v) BSA in PBS. The primary antibodies, rabbit anti-HA tag (1/3000; ChIP Grade; Abcam, http://www.abcam.com) and mouse anti-HDEL (1/200; clone 2E7; Santa Cruz Biotechnology, http://www.scbt.com) were applied to the peels for 2 h at 37°C while the secondary antibodies goat anti-rabbit Cy3 or goat anti-mouse fluorescein isothiocyanate (FITC) were incubated with the peels at 1/250 for 1.5 h at 37°C. Peels were mounted onto slides in 1:1 (v/v) PBS:Citifluor (Leica Microsystems) with the inclusion of 20 mm ascorbic acid and imaged with an LSM 5 Pascal (Carl Zeiss) attached to an Axiovert 200M microscope using a Plan-Neofluar 40 × /0.75 objective lens, and laser excitation at 488 and 543 nm.


This research was funded by an ARC (Australian Research Council) large grant, number A00106405, to RLO. We thank Ian Kaplin and Ellie Kable at the Australian Centre for Microscopy and Microanalysis for their support with TEM and confocal microscopy.

Accession number: GenBank NM_122737 is the accession number for the Arabidopsis ER-luminal binding protein (AtBIP1).