Live Imaging of Companion Cells and Sieve Elements in Arabidopsis Leaves

The phloem is a complex tissue composed of highly specialized cells with unique subcellular structures and a compact organization that is challenging to study in vivo at cellular resolution. We used confocal scanning laser microscopy and subcellular fluorescent markers in companion cells and sieve elements, for live imaging of the phloem in Arabidopsis leaves. This approach provided a simple framework for identifying phloem cell types unambiguously. It highlighted the compactness of the meshed network of organelles within companion cells. By contrast, within the sieve elements, unknown bodies were observed in association with the PP2-A1:GFP, GFP:RTM1 and RTM2:GFP markers at the cell periphery. The phloem lectin PP2-A1:GFP marker was found in the parietal ground matrix. Its location differed from that of the P-protein filaments, which were visualized with SEOR1:GFP and SEOR2:GFP. PP2-A1:GFP surrounded two types of bodies, one of which was identified as mitochondria. This location suggested that it was embedded within the sieve element clamps, specific structures that may fix the organelles to each another or to the plasma membrane in the sieve tubes. GFP:RTM1 was associated with a class of larger bodies, potentially corresponding to plastids. PP2-A1:GFP was soluble in the cytosol of immature sieve elements. The changes in its subcellular localization during differentiation provide an in vivo blueprint for monitoring this process. The subcellular features obtained with these companion cell and sieve element markers can be used as landmarks for exploring the organization and dynamics of phloem cells in vivo.


Introduction
The cellular biology of phloem cells, especially the sieve element-companion cell complex (SE-CC), has been the focus of many studies in angiosperms, due to the key role of these cells in transporting metabolites, macromolecules and other signaling molecules [1]. During resolution than could be obtained with untreated leaves. As sugar export capacity may decrease rapidly in Arabidopsis leaves following their excision from the plant [21], we investigated the possible impairment of phloem transport after the cutting of the petiole and peeling off of the leaf surface with a razor blade. We used the phloem symplasmic tracer 5,6 carboxyfluoresceindiacetate (CFDA) to investigate both phloem transport and sieve element integrity [22]. CFDA is a membrane-permeant dye that is cleaved by cellular esterase to release carboxyfluorescein (CF), a non membrane-permeant fluorescent form of the dye. Fluorescence rapidly progressed from the treated area into the veins (Fig. 1 A-B, S1 Movie), with CF reaching the main vein at an apparent velocity of 6-10 mm min -1 , moving in a proximal direction toward the petiole of the detached leaf. This value was in the same range as the velocity determined in intact Arabidopsis plants (100 μm/s) [14], indicating that the treatment did not prevent phloem transport from the treated area to the petiole (i.e. sink-ward, as expected in intact leaves), and that leaf excision did not trigger the immediate sealing of the sieve tubes connected to the treated area. We also assessed the transport activity of the sieve tubes located immediately beneath the treated area. Two areas located 5 mm apart, close to the midrib, were peeled off, with one used for tracer application and the other, in a more proximal position, used for observations of tracer (c-f) Observation of the transport of the fluorescent label after the application of CFDA to the treated area. (c) Observation of leaf autofluorescence immediately after application of the tracer. Treated areas are less fluorescent, enhancing observation of the vascular network (white arrows). For this experiment, the superficial layers were peeled away from a single vein in two areas: the loading area (la) and the transport area (ta). CFDA was loaded in the loading area. (d) 10 seconds after the application of CFDA. (e) 40 seconds after the application of CFDA, the fluorescence is transported and is visible in the transport area. (f) 100 seconds after the application of CFDA, the fluorescence has moved beyond the transport area. La: loading area; ta: transport area; mv: minor vein, sv: secondary vein; v: main vein. Scale bar = 5 mm. translocation ( Fig. 1 C-F). After loading in the acropetal area, the tracer moved basipetally across the second area on its way to the main vein, at a flow velocity of up to 10 mm min -1 ( Fig. 1 D-F). The fluorescence continued to progress along the veins in this observation area, with minimal fluorescence detectable outside the veins, indicating that the sieve elements were continuing to translocate material beneath the treated area. Thus, the treatment of abaxial tissues from a detached Arabidopsis leaf does not impede phloem transport function, at least during the first few minutes after cutting. Phloem imaging is, therefore, possible.
Use of the PP2-A1:GFP fusion protein as a marker of companion cells Various transgenic Arabidopsis Col-O lines producing green fluorescent protein (GFP) or other fluorescent proteins (CFP or YFP) were generated for live imaging of the phloem (Table 1). Genes encoding free, soluble GFP or GFP fused to the phloem lectin PP2-A1 [19] were expressed under the control of a companion cell-specific promoter (Fig. 2 A, S1 Table). The SUC2 promoter [23] was used in all constructs in preference to the PP2-A1 promoter [19], which frequently drives cosuppression of the endogenous PP2-A1 gene (unpublished results). The veins were observed in vivo by confocal laser scanning microscopy (CLSM), immediately after leaf peeling. Fluorescence was bright in veins of the various orders (i.e. main, secondary and minor; Fig. 2 B). Longitudinal observations of the phloem cells were carried out. In pSUC2: PP2-A1:GFP plants, fluorescence was observed in the cytosol and nucleus of companion cells, with no fluorescence in the sieve elements or phloem parenchyma cells (PPCs) (Fig. 2 D,E). This subcellular distribution was identical to that in undetached leaves, as shown by in vivo observations of leaves in planta using stereo fluorescence microscope (S1 Fig.). Similar findings were recorded for all veins with the pSUC2:GFP:PP2-A1 and pSUC2:PP2-A1:CFP constructs (Fig. 2 F-G). In an individual confocal section, up to 20 companion cells could be observed in the main veins, whereas three to five companion cells were observed in the smallest minor veins. Similar images were obtained with the YFP-tagged PP2-A2, a phloem lectin closely related to PP2-A1 [19], in pSUC2:PP2-A2:YFP plants (S2 Fig., A). By contrast, free GFP fluorescence was observed in both phloem parenchyma cells and companion cells in pSUC2:GFP plants (Fig. 2 C). As the promoter sequence used was the same as that used for PP2 fusions, for which no signal was detected in phloem parenchyma cells, these findings demonstrate the high specificity of this promoter for companion cells, and they indicate that the soluble GFP was transported from the companion cell into the phloem parenchyma cells via the plasmodesmata. Phloem parenchyma cells were larger (up to 17 μm in diameter) and were found in lateral  positions within the veins. By contrast, PP2 fusions, which have an apparent molecular weight (MW app ) of 55 and 45 kDa for the PP2-A1 and PP2-A2 fusions, respectively, did not diffuse from the companion cells to the phloem parenchyma cells, unlike free GFP (MW app of 27 kDa), consistent with the plasmodesmata having a size exclusion limit for nonspecific trafficking of between 30 and 40 kDa in Arabidopsis leaves [24]. Moreover, no fluorescence was observed in the sieve elements with GFP and PP2 fusions, because of rapid translocation towards sink organs, dilution in the cytosol of the sieve element, or, for PP2 fusions, an inability of the fluorescent proteins to pass through pore-plasmodesmata units (PPUs).

Cytological criteria for phloem cell identification
In pSUC2:PP2-A1:GFP and pSUC2:PP2-A1:CFP plants, in which the fluorescent signal was strictly limited to companion cells, we analyzed in more detail the subcellular organization of companion cells. The elongated shape and dimensions of the companion cells (mean diameter of 4 μm and length of up to 70-100 μm) labeled with the fluorescent protein were easily discerned, except at the vein junctions, where curved companion cells were characteristically enlarged at the bend ( Fig. 2 D). Assessments of chlorophyll autofluorescence showed a remarkable alignment of chloroplasts in single file, in successive companion cells (Fig. 2 G). Up to 10 chloroplasts could be observed in individual confocal sections. The chloroplasts occupied most of the diameter of the cells. This feature was sufficiently characteristic to constitute a hallmark of companion cells, because no other cell type displayed such an alignment of chloroplasts. In most companion cells, one chloroplast was found adjacent to one side of the cell nucleus. The mean diameter of the chloroplasts in companion cells was 3.21 ± SD 0.66 μm (n = 662), smaller than that in phloem parenchyma cells (4.08 ± SD 0.99 μm; n = 278) and mesophyll cells (5.58 ± SD 1.01 μm; n = 250) (S3 Fig.). Other cell types also displayed markedly different chloroplast organizations. For example, in phloem parenchyma cells, the chloroplasts were arranged at the cell periphery, mostly on the side opposite the companion cell (

Subcellular organization of companion cells
Companion cells are highly compartmentalized (Fig. 2 G). Their subcellular organization was analyzed in crosses between the pSUC2:PP2-A1:CFP line and transgenic lines producing GFP or YFP fluorescent marker molecules (S2 Table) (Fig. 3, S4 Fig.). These cells could be identified on the basis of chloroplast alignment or CFP fluorescence from the PP2-A1:CFP fusion (Fig. 3 A). An example of this approach is provided by the analysis of a cross with the p35S:GFP:LTI6b line [25], which carries a plasma membrane marker. Fluorescence was readily observed at the plasma membrane (PM) in companion cells (Fig. 3 B). In p35S:H2B:RFP plants expressing the fluorescent histone 2B (H2B:RFP) nuclear marker, the nuclei of companion cells appeared dense and compacted (Fig. 3 C). Similarly, large numbers of mitochondria (

Subcellular organization of the sieve elements
The subcellular organization of sieve elements was also examined. The GFP:LTI6b marker labeled the plasma membrane in the sieve elements (1-3 μm diameter in minor veins and 3-4 μm in larger veins), with brighter fluorescence was observed in the vicinity of the sieve plates ( Fig. 3 B). Plants expressing the ER marker (ER:YFP) displayed a succession of fine fluorescent stacks at lateral positions in sieve elements (Fig. 4 A). For the mitochondrial marker (COX4: YFP), we observed faint fluorescence colocalized with the vital dye MitoTracker Red (Fig. 4 B,C), indicating this marker is active in sieve elements. Other fluorescent GFP markers were more difficult to visualize. For the nuclear marker (H2B:RFP) or the tonoplast marker (γTIP:YFP), the breakdown of the nucleus and tonoplast during differentiation of the sieve elements accounted for the absence of fluorescence. For the actin marker (fABD2:GFP), the microtubule marker (GFP:MBD) and the chloroplast marker (RbcS:YFP), no fluorescence signal was detected in the sieve elements. Sieve element plastids are not photosynthetic, and RuBisCO is probably degraded during sieve element differentiation. The actin cytoskeleton has been identified in sieve elements [14,26], and the lack of fluorescence may be due to the signal being below the detection threshold, or the GFP and YFP fusions being unable to pass through the PPUs between the companion cells and sieve elements.
We also examined the subcellular organization of sieve tubes in transgenic Arabidopsis lines producing GFP-tagged sieve element proteins ( Table 1, S2 Table). As expected, filaments of up to 200 μm long and resembling those described in roots in planta [14,15] were observed in the sieve elements (Fig. 4 D-F) of the leaves of pSEOR2:SEOR2:GFP plants producing a GFP-labeled structural phloem filament protein, SEOR2 [15]. Similar findings were obtained for undetached leaves observed in planta using stereo fluorescence microscope (S1 Fig., C). These filaments were not colocalized with mitochondria ( Fig. 4 E,F). In some cells, the protein appeared to accumulate as spherical protein agglomerates similar to those reported in immature sieve elements in roots [14,15]. Similar images were obtained for plants expressing the SEOR1: GFP fusion. We then investigated the sieve tube proteins RTM1 and RTM2.
We analyzed the lines expressing the pRTM1:GFP:RTM1 and pRTM2:RTM2-GFP constructs in the TEV-susceptible C24 (rtm1/rtm1) and Col-O rtm2-1 mutant (rtm2/rtm2) background (Table 1), in which they provided functional complementation for TEV restriction with RTM1 [20,27] and for LMV restriction with RMT2 (Revers, personal communication). In pRTM1:GFP:RTM1 plants, an alignment of spherical protein bodies with an apparent diameter of 1.16 ± 0.01 μm (n = 612) was found at the periphery of individual sieve elements ( Fig. 5 A), with up to 30-40 bodies in the sieve elements found in the larger veins. These bodies were not colocalized with mitochondria ( Fig. 5 B). Their dimensions were of a similar range to that of sieve element plastids [14,28,29]. By contrast in pRTM2:RTM2:GFP plants, a reticulate organization lining the plasma membrane and surrounding bodies smaller than mitochondria was observed in the sieve elements ( Fig. 5 C,D). The nature of these bodies remains unknown, but structures of a similar dimension were observed in transmission electron micrographs of sieve elements in Arabidopsis leaves (S5 Fig.). In the lines expressing these sieve element fluorescent markers, as lines expressing companion fluorescent markers, similar findings were obtained for all orders of veins, other than for RTM2, which was difficult to image in the minor veins (Fig. 6).
Localization of PP2-A1 in the sieve elements PP2 phloem lectins were once considered to P-protein components [30], but their localization in the sieve elements remains a matter of debate [31]. Immunolocalization studies have shown that PP2-A1 accumulates in small bodies in the sieve elements [31]. In the pSEOR2:PP2-A1: GFP lines, in which expression of the GFP fusion is driven by the sieve element-specific SEOR2 promoter [15,32], fluorescence was detected around bodies at the lateral cell borders of the sieve elements (Fig. 5 E-I). These bodies had a bimodal size distribution, with a main peak for their longest dimension at 0.52 μm and a second, less prominent peak at 0.96 μm, accounting for about one fifth of the bodies (n = 328) (S6 Fig., G). They were motionless, indicating they were anchored to membranes. The GFP fluorescence signal could be mostly observed as small rings surrounding an unlabeled area (Fig. 5 F-G, S6 Fig., A-F), showing that the GFP-tagged proteins accumulated principally at the periphery of some organelles. In pSEOR2:PP2-A1:GFP  leaves stained with MitoTracker Red, the fluorescence was surrounded by GFP fluorescence (Fig. 5 H,I), demonstrating that the class of PP2-A1-GFP-labeled bodies with a small diameter (0.52 μm) corresponded to sieve element mitochondria. Transmission electron micrographs of sieve elements confirmed that the mitochondria were aligned along the plasma membranes, with an apparent diameter of 0.458 ± 0.026 μm (n = 12) (S6 Fig., H-J). The larger bodies labeled by PP2-A1:GFP (mean diameter of 0.96 μm) did not colocalize with MitoTracker Red.

Distribution of PP2-A1:GFP in differentiating sieve elements
The SEOR2 promoter is active in immature sieve elements [32], making it possible to localize PP2-A1:GFP in differentiating sieve elements. In pSEOR2:PP2-A1:GFP plants, differentiating sieve elements were readily recognizable on the basis of their bright fluorescence signal. Differentiation could be monitored by staining nuclei with propidium iodide and recording plastid autofluorescence. At early stages of differentiation in cells that still contained a nucleus, PP2-A1:GFP fluorescence was found in the cytoplasm and in the nucleus (Fig. 7 A), as in companion cells. In differentiating cells displaying plastid autofluorescence but with no observable nucleus, PP2-A1:GFP continued to accumulate in the cytosol (Fig. 7 B). At later stages, in cells retaining large organelles, presumably lytic vacuoles, but no longer displaying plastid autofluorescence and nuclear staining, PP2-A1:GFP was observed both in the cytosol and as small dots of fluorescence, indicating the progressive anchoring of PP2-A1 to mitochondria or other organelles (Fig. 7 C-E). In mature sieve elements, the cytosolic fluorescence had completely disappeared, and PP2-A1:GFP appeared as discrete spots (Fig. 7 F).

Live imaging of phloem cells in Arabidopsis
Fluorescent markers were used for the in vivo imaging of subcellular components of phloem cells in detached Arabidopsis leaves (Fig. 8). Observations were made for a few minutes, immediately after sample preparation. SE observations on detached leaves may be problematic due to hydraulic disruption, as the pressure is likely to drop, making it difficult to refill the sieve tubes, in leaves disconnected from the plant xylem [33]. We carried out an experiment with CFDA to check that the transport activity of the phloem was preserved for several minutes. We also confirmed that leaf peeling did not substantially modify the subcellular organization of sieve elements, with respect to other methods, in experiments using fluorescently tagged proteins or vital markers [14,34]. Virtually identical methods have been used on intact leaves from Vicia faba and tobacco, for the visualization of subcellular structures in phloem cells [17,34]. However, transport in the phloem is an integrative phenomenon operating at whole-plant level that is dependent on sources, sinks, a transport tube and a source of water; live imaging results for phloem cells over more extended periods should therefore be interpreted with caution. Companion Cell and Sieve Element Imaging

Landmarks for the identification of companion cells and sieve elements
On longitudinal sections, the companion cells in leaf phloem present a distinctive feature, in the form of a remarkable single-file alignment of the chloroplasts (Fig. 8), which occupy a large proportion of the cell volume. As the chloroplasts and vacuoles of these cells take up most of their volume, the nucleus, mitochondria and ER are compacted, as observed by transmission electron microscopy [4,35]. This situation is very different from that in phloem parenchyma cells, in which the distribution of chloroplasts at the cell periphery, opposite the companion cells, also constitutes a useful landmark for identification. Some of the fluorescent markers  Companion Cell and Sieve Element Imaging were not detected in mature sieve elements. Most organelles, such as nuclei and vacuoles, disintegrate during sieve element differentiation, and we would therefore not expect to observe them. Nevertheless, the membrane markers GFP:LTI6b (PM marker) and ER:YFP (ER marker), and the mitochondria COX4-GFP marker were observed. Either these markers diffused from the companion cell to the sieve element, or the fusion proteins were produced in differentiating sieve elements and were resistant to degradation in mature sieve tubes, unlike RuBisCO and histone 2B that are probably subjected to proteolysis during sieve element differentiation. By contrast, sieve element-resident proteins fused to GFP (GFP:RTM1, RTM2:GFP, SEOR1: GFP, SEOR2:GFP) produced under the control of their own promoters are evidently resistant to proteolysis and constitute reliable markers of sieve elements (Fig. 8). None of these fusion proteins diffused to adjacent cells, due to low solubility, anchorage to organelles or dimensions preventing passage through plasmodesmata due to the size exclusion limit.

The cytoskeleton in phloem cells
The organization of the cortical microtubules and actin filaments was observed in the companion cells. The cortical microtubules displayed transverse, oblique or longitudinal orientations, with some heterogeneity between samples. F-actin was more difficult to visualize in the companion cells. Only longitudinal actin bundles and a few thin actin filaments were observed. In the sieve elements, no fluorescent signal was detected for F-actin or microtubule markers. This lack of signal may result from the inherent difficulty of observing thin cytoskeleton elements. As for some other GFP fusions, the fusion proteins may be too large or have an unsuitable chemical composition for passage through the pore-plasmodesmata units or another structure. Nonetheless, a recent study established the presence of a peripheral actin meshwork in the sieve elements of fava bean leaves, by immunocytochemistry and the use of the phalloidin fluorescent marker [26]. As far as cortical microtubules are concerned, small amounts of α-tubulin have been detected in the phloem sap exudate of pumpkin [36,37], but β-tubulin is not detectable in the sap of most species [38] and there is currently no direct evidence for the assembly of microtubules in the sieve elements [2].

Sieve elements: not so empty after all
The nature of many sieve element components remains unknown [39]. The use of fluorescent markers and fluorochromes facilitated the identification of some of the organelles in the sieve elements. For example, the filamentous structural P-proteins were found in a reticulate pattern, similar to that previously reported in roots [14,15] for the sieve element reticulum, as reported for the sieve elements of tobacco [34]. There was also a large number of mitochondria, as reported in a previous study [4]. PP2-A1-GFP fluorescence was partially colocalized with mitochondria and with another class of organelles not stained with MitoTracker Red. A number of GFP-RTM1-labeled bodies, about 1 μm in diameter, were found evenly aligned along the sieve tubes. GFP-RTM1 did not colocalize with mitochondria, suggesting that at least some of these bodies may be plastids, the largest sieve element bodies other than mitochondria, with diameters of about 1 μm in Arabidopsis, as estimated by transmission electron microscopy [14,28,29]. RTM2:GFP fluorescence was found in a layer lining the plasma membrane that might be part of the parietal ground matrix. At several locations, it surrounded small round bodies at the periphery of the sieve elements. These organelles have yet to be identified, but small subcellular structures of similar diameters were also observed on transmission electron micrographs of the sieve elements (SE). Similar bodies have also been observed in the SE parietal ground matrix in Arabidopsis [14] and in potato SEs, in which it was suggested that they might be endocytotic vesicles [40].
The numbers and dimensions of bodies detected and stacked at the cell periphery or located in the lumen of sieve elements (up to 4 μm diameter in the leaf) also indicated that these cells might contain a large volume of organelles and protein bodies. These data are consistent with those recently reported for Arabidopsis and tobacco, in which an estimated 35% of the lumen is occupied by sieve tube constituents [14].

PP2-A1:GFP as a marker for the live imaging of sieve elements
The events occurring during the maturation of sieve elements, a process taking only 16 to 21 hours, are difficult to study in vivo, because of the small proportion of differentiating sieve elements. Indeed, this task has been compared to searching for the proverbial needle in a haystack [1]. GFP-tagged PP2-A1 produced under the control of the SEOR2 promoter produced a bright fluorescent signal in immature sieve elements, making it possible to image differentiating sieve elements easily. In sieve elements in the early stages of differentiation, PP2-A1:GFP was found in the cytosol and nucleus. This distribution is similar to that observed in companion cells. At later stages, during the disappearance of the nucleus and the reorganization of the organelles, PP2-A1:GFP was partially soluble in the cytosol and began to aggregate around bodies at the cell periphery. In mature sieve elements, PP2-A1:GFP was no longer soluble and was associated exclusively with the surface of mitochondria and a second class of larger organelles. These patterns define three stages, paralleling those described in the phloem of the fixed root tissues from Arabidopsis [29], which could be used as a blueprint for in vivo studies on the differentiation of the sieve elements.
Localization of PP2-A1:GFP in the sieve elements Differentiating sieve elements undergo major intracellular reorganizations, to create an unobstructed lumen through the anchoring of the remaining organelles to the plasma membrane at the lateral borders [41]. This enables the sieve tubes to support mass flow and to compensate for the high pressure generated by sugar loading. Improvements in fixation methods for phloem cell observation by transmission electron microscopy (TEM) have led to the identification of minute clamp structures coupling the membranes of the mitochondria, sieve element reticulum and plasma membrane, in tomato, Vicia fava and, more recently, Arabidopsis [14,41]. In Arabidopsis, TEM images have shown that the mitochondria and ER are covered with a halo of protein spikes forming the clamp structures and that these structures, together with vesicles, are embedded in an amorphous ground matrix forming a parietal layer [14]. The components of the parietal ground matrix and the clamps remain unknown. Immunolocalization studies have shown that PP2-A1 are located in the sieve elements at small, unidentified bodies, different from the P-proteins [31]. In mature sieve elements, PP2-A1:GFP was located at the periphery of mitochondria and of a second class of organelles with a diameter of about 1 μm. Immunogold labeling showed that PP2-A1 was associated with the sieve element plastids [31]. We therefore suggest that at least some of these bodies were plastids. In the parietal ground matrix, PP2-A1:GFP may thus be present in the clamp structures around mitochondria and plastids. However, the resolution obtained with CLSM was not sufficient to confirm whether it formed small protein spikes around the organelles, such as those observed by TEM and constituting the clamps [14].

Functions of PP2-A1:GFP in phloem cells
PP2-A1 is synthesized in the companion cells (CC), before being translocated to sieve elements [19]. In CC, PP2-A1-GFP was observed in both the cytosol and the nucleus. A nucleocytoplasmic distribution has also been reported for a PP2-like protein in the mesophyll cells of Nicotiana tabacum [42]. This suggests that this chaperone [43], which has been shown to bind GlcNAc residues [44], may be involved in the nucleocytoplasmic shuttling of proteins modified by O-GlcNAcylation in CC cells, before their trafficking through the companion cell-sieve element plasmodesmata. Such a role has already been proposed for the tobacco phloem protein NCAPP1, which also binds to O-GlcNAc proteins [45]. The differential localization of PP2-A1, soluble in the cytoplasm and in the nucleus of companion cells and immature sieve elements, but aggregated in mature sieve elements, suggests that this protein is multifunctional. Multifunctionality is a basic feature of many protein-binding proteins involved in structure and signaling, such as Armadillo/β-catenin, which acts as a platform for many binding partners and is involved in both cell-to-cell junctions in animals and transcriptional regulatory processes [46]. PP2-A1 binds proteins in phloem exudate [44], and other PP2s have been shown to bind mRNAs [47,48]. The PP2-associated structures found in the sieve elements may therefore retain proteins and RNAs within sieve elements, preventing their translocation by mass flow in functional sieve elements.

Plant materials and growth conditions
Arabidopsis thaliana (accession Columbia) plants were used for plant transformation and in vivo observations. Plants were grown in the greenhouse under long-day conditions (16 h light/ 8 h dark) in soil (Tref substrates or sand) and watered with Plant-Prod nutrient solution (Fertil-France, Boulogne Billancourt, France). The transgenic lines expressing subcellular markers are described in S2 Table. Crosses were carried out between homozygous plants.

Expression vectors
The binary vectors encoding translational fluorescent fusion proteins (S1 Table) were constructed with the GATEWAY system (Invitrogen, Life Technologies, Saint-Aubin, France). The promoter regions of AtSUC2 (1400 bp) and AtSEOR2 (1000 bp) were amplified from genomic DNA extracted from Arabidopsis Col-O and inserted into pCR II TOPO (Invitrogen). The coding sequence of AtPP2-A1 was amplified with the PP2A1Not and PP2A1XbaI primers (S3 Table) and inserted between the NotI and XbaI sites of pCR II TOPO. ATTB sites were created by PCR with the primers pSUC2attb1, pSEOR2attb1 and PP2A1attb2, and intact attB recombination sites were reconstituted with the primers attB1 and attB2. These PCR fragments were introduced into pDONR 207 (Invitrogen). The vectors for LR recombination are described in S1 Table. The constructs were verified by sequencing.

Plant transformation
Agrobacterium tumefaciens C58pMP90 [49] was transformed with the binary vectors, by electroporation. For vectors derived from pGKG, we added the helper vector pSoup [50] to the transfection mixture. Arabidopsis plants were transformed by floral dipping [51]. Transformants were selected on kanamycin (50 mg l -1 ), hygromycin (15 mg l -1 ) or glufosinate (15 mg l -1 ). For each construct, at least eight primary transformants were screened and one line was chosen for further analysis on the basis of its fluorescence quality and lack of phenotypic alterations.
Most observations were carried out on homozygous F3 plants.

Cytology
Observations were performed on the veins of excised rosette or stem leaves of Arabidopsis. Clear images of the leaf vein system were obtained by using a razor blade to remove the overlying cell layers, corresponding to the epidermis and spongy mesophyll, from the abaxial side of the leaf, essentially as previously described [17]. Immediately after the cutting of the leaf petiole, the leaf was immersed in water, a small area (1 cm 2 ) of the abaxial cell layers of the leaf was removed, and the leaf was rinsed and directly mounted in water or dye for imaging. Observations were carried out within a few minutes of preparation, by confocal laser scanning microscopy, and did not extend beyond 10 minutes after peeling. The clearest observations were obtained with leaves that were more than half-expanded, which should have gone through the sink-to-source transition [52]. For observations of mitochondria, samples were incubated for 5 min and mounted directly in 2.5 μg ml -1 MitoTracker Red CMXRos (Invitrogen) for observation. For live observations of nuclei in phloem cells, samples were mounted in 50 μg ml -1 propidium iodide (Molecular Probes). Observations on non-excised leaves from intact plants were performed on Arabidopsis seedlings grown in vitro for 15 days on vertical plates with a stereo fluorescence microscope. A rosette leaf was placed on a drop of water on a coverslip and the abaxial epidermis was removed with a razor blade. A second coverslip was added before fluorescence imaging with an Axiozoom V16 (Zeiss) microscope equipped with a Plan-Neofluar Z 2.3x/0.57 RWD 10.6 mm.

Phloem loading assay
Phloem loading in leaves was monitored by gently dry-peeling away a small area (*25 mm 2 ) of the abaxial surface, at the leaf margin, with a razor blade, and immediately applying 5 μl of 5,6 carboxyfluorescein diacetate (CFDA)-mixed isomers (60 μg ml -1 ) (Invitrogen) to the exposed area. The treated leaves were observed immediately, under a Nikon SMZ-1500 binocular microscope equipped with a high-pressure mercury lamp (C-SHG1, Nikon) and a GFP filter, and the loading of the fluorescent dye was monitored by video recordings every 1.5 s, for 3 minutes. The dye progressed basipetally in the minor veins and then in higher order veins, over at least 3 minutes.

Confocal laser scanning microscopy
At least 10 independent leaf samples were observed for each marker, and several sectors with phloem cells were observed per sample. For confocal laser scanning microscopy, leaves were mounted in water. Fluorescence images were obtained with either a Leica TCS-SP2-AOBS spectral confocal laser scanning microscope or a Leica SP5II-AOBS-Tandem HyD spectral confocal laser scanning microscope, both of which were equipped with a Leica HCX PL APO x 63 water immersion objective. CFP was excited with a 405 nm diode laser; GFP and chloroplasts were excited with the 488 nm line of the argon laser; YFP was excited with the 514 line of the argon laser, and RFP was excited with the 543 nm line of the He-Ne laser. Fluorescence emission was detected in the 450 to 520 nm range for the CFP construct, at 500 to 560 nm for GFP constructs, at 521 to 560 nm and 510 to 565 nm for SP2 and SP5, respectively, for YFP constructs, at 600 to 650 nm for MitoTracker Red, 610 to 650 nm for propidium iodide, 565 to 645 nm for RFP, and 684-784 nm and 645-738 nm, respectively, for SP2 and SP5, for chloroplast autofluorescence. Images were recorded and processed with LCS software version 2.5 (Leica Microsytems, Nanterre, France). CFP, YFP and GFP fluorescence data were acquired by sequential scanning, with the 405 nm laser diode used to excite CFP, the 488 nm laser line used to excite GFP and the 514 nm laser line used to excite YFP. All images were single confocal 512 x 512-pixel scans, except for images of the cytoskeleton (Z-stack of 6 images with a step of 0.32 μm) and mitochondria (Z-stack of 20 images with a step of 0.25 μm).

Ultrastructure on transmission electron microscopy
Small main vein or petiole sections were fixed by incubation, under vacuum, with 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.2) supplemented with 0.25 M sucrose for 3 h at 4°C. They were post-fixed by incubation with 1% osmium tetroxide in the same buffer for 2 h at 4°C. They were then dehydrated in a graded series of ethanol concentrations (30-100%), subjected to propylene oxide substitution, infiltrated and embedded in Epon resin. Ultrathin sections (50 nm) were cut on a Leica Ultracut-E microtome, stained with 7% uranyl acetate for 15 minutes and then with lead citrate for 5 minutes. Sections were examined at 80 kV in a FEI CM-10 transmission electron microscope equipped with an X-60 AMT digital camera (Elexience, Verrières-le-Buisson, France).

Image analysis
The size of the protein bodies observed by CLSM was analyzed with the NIH image analysis program ImageJ v1.43u (NIH, http://rsb.info.nih.gov/ij/). Plastid size was analyzed by using ImageJ for measurements on images of plastid autofluorescence. Similarly, TEM images were analyzed to determine the size of plastids and mitochondria in the sieve elements.