Introduction

Plasmodesmata (PD), the small pores that interconnect most types of plant cells, are about 50 nm in diameter and are therefore difficult to image using conventional optical-based methods. Despite their very small size, PD can be seen but not resolved with a conventional light microscope. The minute dimensions of PD did not deter many of the light microscopists of the nineteenth century from making painstaking observations of living plant cells that have seldom been surpassed in their attention to detail. Tangl (1879) first described ‘open communications’ between endosperm cells, and Strasburger (1901) later called these channels ‘plasmodesmata’, challenging the widely held cell theory of Schleiden and Schwann (1804) that each cell functioned as an autonomous unit (as cited by Roberts 2005).

The advent of electron microscopy in the 1950s brought a new level of attention to PD ultrastructure and remains the method by which current models of PD structure are derived. However, new advances in fluorescence imaging have broken the ‘diffraction barrier’, long held to limit the resolution of optical microscopy (Huang et al. 2009; Huang 2010; Patterson et al. 2010), and are beginning to blur the boundary between light and electron microscopy, bringing PD into a new imaging ‘window’ that was previously thought impossible. This review will consider a range of imaging approaches for studying PD, encompassing more conventional methods alongside emerging technologies that are likely to find their way into mainstream imaging in the near future.

Transmission electron microscopy

There is no doubting the impact of transmission electron microscopy (TEM) on the field of PD biology. Virtually all we know of PD structure is derived from this approach (Roberts 2005; Fig. 1). TEM was instrumental in demonstrating that PD are plasma membrane (PM)-lined pores containing a central desmotubule connected to the cortical ER of adjacent cells (Roberts 2005; Fig. 1).

Fig. 1
figure 1

Diagrammatic representation of the PD pore as derived from TEM studies. The central desmotubule (pale blue) is continuous with the cortical ER. Proteinaceous particles (dark blue), sometimes helically arranged, are associated with the outer leaflet of the desmotubule. Additional particles (red) are associated with the inner leaflet of the plasma membrane (dark green). Spoke-like extensions of unknown origin connect the desmotubule with the plasma membrane. In some species, a prominent raised wall collar is present around the neck region of the pore. Diagram courtesy of Alison Roberts

Full size image

In the 1980s and 1990s, there was a significant interest in the fine structure of PD from a wide range of evolutionary groups. Notable contributions were magnificent descriptions of PD by Esau (e.g. Esau and Thorsch 1985), Kollmann (see Ehlers and Kollmann 2001) and Evert (e.g. Robinson-Beers and Evert 1991). In our view, the quality of these TEM images of PD has not been surpassed, despite significant improvements in resolution. Example images of PD are shown in Fig. 2 and reveal significant substructural details of the PD pore, including the desmotubule and central rod, together with ‘spokes’ that extend from the desmotubule to the PM (Fig. 2a, b). Small particles, often showing ninefold symmetry, and possibly arranged in a spiral fashion, (Ding et al. 1992a, Blackman and Overall 2001), have been observed to be associated with the outer leaflet of the desmotubule and the inner leaflet of the PM (Figs. 1 and 2f). These particles are about 4–5 nm in diameter, slightly smaller than the width of the cytoplasmic sleeve (5–6 nm), while the radial gaps between the spiralling particles may be as small as 3 nm (reviewed in Robards and Lucas 1990). PD may have a simple or complex architecture (see Burch-Smith et al. this issue) and may become highly specialised at specific interfaces. In the case of the PD that connect the sieve element (SE) with its companion cell (CC), a single callose-lined pore leads into the SE while branched pores occur in the CC wall (Fig. 1c). In Strasburger cells within the phloem of pine needles, large complex PD form, each containing several pores derived from the adjoining cells. Such complex PD often show conspicuous central cavities shared by each of the contributing pores (Fig. 2d). Many PD, when sectioned across the neck region of the pore, often display a densely staining cytoplasmic sleeve (Fig. 2f) while those sectioned below this region often show the sleeve to be electron lucent (e.g. Fig. 2b). There has been much speculation as to the nature of the structural components of PD, including suggestions that they may be cytoskeletal elements (Aaziz et al. 2001; Blackman and Overall 2001; Oparka 2004; Roberts 2005). However, for reasons of molecular constraint, cytoskeletal elements would be difficult to incorporate into the cytoplasmic sleeve of PD (Tilsner et al. 2010).

Fig. 2
figure 2

Images of PD derived from TEM. Simple PD connecting mesophyll cells in a sugarcane leaf are shown in longitudinal (a) and transverse sections (b). Note the prominent neck constrictions in a. In b, the PD are sectioned below the neck region, revealing an electron-lucent cytoplasmic sleeve. Spoke-like extensions are apparent in some of the pores. a, b are from Robinson-Beers and Evert (1991) and reproduced with permission from the authors and the publisher. Scale in a = 200 nm; scale in b = 100 nm. c The specialised pore-PD connecting a sieve element (SE) with its companion cell (CC). Prominent callose collars (C) line the pores leading into the SE. The PD are branched within the CC wall. From Evert and Mierzwa (1989) with permission of the authors and the publisher. Scale = 150 nm. d A longitudinal section through a complex PD structure connecting Strasburger cells in the phloem of a pine needle. Multiple PD pores connect into a shared central cavity (Ca). Arrows indicate neck constrictions and darts show open desmotubule structures. From Glockmann and Kollmann (1996) with permission from the authors and the publisher. Scale = 1 μm. e PD connecting the trichome with the epidermis in a tobacco leaf. The PD are sectioned across the neck region and show conspicuous wall collars (WC). From Faulkner et al. (2008) with permission from the publisher. Scale = 100 nm. f Single PD pore from the trichome/epidermis interface of a tobacco leaf. Individual components can be discerned, including the plasma membrane (PM), desmotubule (Dt) and central rod (CR). The wall collar (WC) is also visible. Note that the cytoplasmic sleeve (CS) in the neck region contains electron-dense material. Scale = 100 nm. Oparka unpublished data. g–i PD imaged after tannic acid fixation (from Olesen 1979). Wall collars are visible in g. h A single PD after rotational enhancement. Note that the particles within the wall collar show a ninefold symmetry. i A model derived from the TEM images, showing the wall collar and the dimensions of the different regions in the neck of the pore. Reproduced with permission from the author and the publisher

Full size image

Chemical fixatives may cause structural artefacts due to poor sample penetration and the relatively slow rate at which the cellular matrix is crosslinked (Gilkey and Staehelin 1986). However, rapid low-temperature fixation, followed by cryo-EM, allows imaging at a near-native state (McIntosh et al. 2005). In this procedure, heat is quickly removed from the sample before ice crystals can form, preserving the cellular architecture. High-pressure freezing, followed by freeze substitution, has revealed the basic structure of PD, confirming that many of the structures seen in conventional TEM images are not artefacts associated with aldehyde fixation (Ding et al. 1992a). However, it is worth noting that despite the fixation/freezing protocol used, conventional resin-embedding procedures may lead to tissue shrinkage, potentially underestimating the dimensions of structures within PD. In most studies of PD, postfixation in osmium, followed by conventional staining with uranyl acetate and lead citrate has been the norm. However, a number of staining techniques have revealed additional details of PD structure and these will be dealt with briefly.

Tannic acid fixation

Many PD display a conspicuous wall collar around the neck of the pore (Fig. 2e–i). Olesen (1979) used high-contrast ultrathin sections following tannic acid fixation (2% tannic acid added to the primary fixative) to demonstrate the presence of large particles (20–30 nm diameter) immediately below the plasma membrane, in the wall collar around the neck constriction (Fig. 2g–i). Rotational enhancement of the images revealed a ninefold symmetry of the external collar particles (Fig. 2h, i). The overall dimensions of the external ring structure described by Olesen and others (about 100–130 nm in diameter, Robards and Lucas 1990) correspond well to the raised wall collar seen in field emission scanning electron microscopy (FESEM) images (Fig. 4e, f; see below). However, the nature of the particles within the raised collar remains to be determined. Olesen (1979) suggested that these external particles might be components of a hypothetical ‘sphincter’, involved in the regulation of solute transport through PD. This remains a possibility as a number of proteins, including a class I β 1,3-glucanase (Levy et al. 2007), a viral movement protein-interacting host factor that affects callose deposition (Fridborg et al. 2003), and a novel PD-localised callose-binding protein (PDCB1; Simpson et al. 2009) have all been shown to be associated with the neck region of PD. However, it remains to be determined whether the entire machinery necessary for callose deposition at PD resides exclusively within the raised wall collar.

Zinc iodide osmium impregnation

Zinc iodide osmium (ZIO) impregnation of tissues, sometimes known as the ‘Maillet method’ (Maillet 1959; Pellegrino de Iraldi 1977), was used effectively for the demonstration of autonomic nerves at the light microscope level and subsequently at the TEM level. This stain accumulates as an electron-dense product between double-membraned structures (for recipe see Hawes 1994). In plants, it has been used to selectively stain the cortical ER, Golgi and nuclear membranes, as well as thylakoids within chloroplasts (Barlow et al. 1984; Hawes 1994). A single maize root cell stained with ZIO is shown in Fig. 3a and reveals a significant contrast within the endomembrane system. A variation of the ZIO method utilised osmium ferricyanide (White et al. 1979), which like ZIO accumulates between double membranes. This method was used by Hepler (1982) to demonstrate the entrapment of ER within the phragmoplast to form the future PD of the cell plate. Images of ZIO-fixed maize roots showing the entrapment of ER strands in a developing cell plate are shown in Fig. 3b, c. Interestingly, in the Hepler (1982) study, as the PD pores continued to develop in the new cell wall, the electron-dense stain was gradually excluded from the desmotubules, suggesting that the closely appressed ER membranes of this structure had insufficient lumenal space to allow accumulation of the ZIO. There has been a general consensus that the ER membranes forming the desmotubule are always in close contact. However, this is not always the case (e.g. Glockmann and Kollmann 1996; Fig. 2d; see also Tilsner et al. this issue). In our lab, we have found that ZIO stain accumulates within the desmotubule in PD from a range of tissues, for example in the PD that connect the protophloem sieve elements with surrounding pericycle cells in the Arabidopsis root tip (Fig. 3d). As it now appears that the desmotubule may not be a fixed structure and may dilate in response to factors such as virus infection (Guenoune-Gelbart et al. 2008; Epel 2009), the use of ZIO staining may find uses in confirming that under certain conditions the membranes that form the desmotubule may indeed become separated.

Fig. 3
figure 3

Zinc iodide osmium (ZIO) impregnation of plant tissues. a Maize root tip cells after ZIO postfixation. The endomembrane system is clearly delineated, including ER, Golgi bodies and the nuclear envelope. Scale = 10 μm. b, c Differentiating cell plates viewed after ZIO postfixation. Cortical ER strands have become trapped in the developing cell wall (W). Microtubules (MT) are also visible in c. Scale in b = 5 μm. Scale in c = 1 μm. Images a–c, courtesy of C. Hawes (unpublished data). d PD between a protophloem sieve element (SE) and a neighbouring pericycle cell (Pe) in the Arabidopsis root show accumulation of ZIO in the desmotubule (arrow). Scale = 1 μm. Oparka, unpublished data

Full size image

Field emission scanning electron microscopy

Scanning electron microscopy, which is normally restricted to exposed surfaces, would appear superficially, to be unsuitable for imaging PD. However, FESEM is potentially a powerful tool in imaging PD (Vesk et al. 2000), provided that they can be suitably exposed on the cell surface. In addition, cryo-FESEM, in which the tissue is rapidly frozen and fractured within the specimen chamber, is likely to reveal details of PD close to their functioning state. We have been exploiting this method to obtain new images of the neck region of PD and on the ways in which secondary PD develop within a single wall interface (Faulkner et al. 2008). Leaf trichomes are a suitable model for FESEM as they comprise a linear chain of cells that contain PD only at their end walls (see also Burch-Smith et al. this issue). Using tobacco, small pieces of leaf were frozen rapidly in sub-cooled liquid nitrogen and rapidly transferred to the freezing stage of the microscope. While still frozen, the trichomes were removed with a scalpel, exposing the cell–wall interface between the basal trichome cell and the epidermis. Using this approach, all the PD at a single wall interface can be exposed (Fig. 4a, b; see also Faulkner et al. 2008). A second approach involves placing the leaf surface into colloidal graphite and prising the leaf surface away from the frozen graphite, revealing predominantly the PM surface between the basal cell and the epidermis (see diagram in Fig. 4). We have also used fracturing, followed by FESEM, to examine the interfaces between vascular parenchyma cells, revealing extensive PD pit fields in areas of shared wall interface (Fig. 4d). Such large groupings of PD are seen only rarely in TEM images when the wall has been sectioned tangentially (c.f. Fig. 4c). These freeze-fracturing approaches have revealed new details of the neck region of PD. As chemical fixation and embedding are absent from this procedure, and the PD are imaged while still frozen, the images represent PD in as close to their functioning state as possible (Fig. 4e, f).

Fig. 4
figure 4

FESEM images of PD. a The trichome/epidermis interface in a tobacco leaf after freeze fracturing and FESEM. Note that the fracture plane has occurred across the cell wall (W) and also along the plasma membrane (PM) of the underlying cell. Individual PD are visible in both the wall and membrane faces. b A diagram of the fracture, with the positions of all PD at this interface marked in green. The inset diagram shows the different wall/membrane faces exposed during freeze fracturing. Scale = 5 μm. From Faulkner et al. (2008) with permission from the publisher. c A rare glancing TEM section across the neck region of several PD in a pigment strand cell of a developing rice grain. Note that the wall collars have an electron-lucent appearance. Scale = 10 μm. Oparka, unpublished data. d Freeze fracture of a tobacco vascular parenchyma element viewed by FESEM. Numerous PD with raised collars are visible and are aggregated into distinct pit fields. Scale = 5 μm. Oparka, unpublished data. e High-magnification FESEM view of PD pores with raised wall collars (wall fracture, see a). Note that many of the pores are twinned. Scale = 1 μm. From Faulkner et al. (2008) with permission from the publisher. f Equivalent PM view to that shown in e reveals PD in the surface of the membrane. Here, the wall collars appear as dark imprints. The central desmotubule can be seen in some of the pores. Note also the imprints of cellulose microfibrils on the surface of the frozen membrane. Scale = 1 μm. From Faulkner et al. (2008) with permission from the publisher. Inset shows a single PD pore in the plasma membrane. A desmotubule is apparent, as well as radially arranged particles within the wall collar (WC; c.f. Fig. 2). Scale = 100 nm. Oparka, unpublished data

Full size image

Images of fractured wall faces reveal prominent wall collars and significant ‘twinning’ of PD pores (see Faulkner et al. 2008; Fig. 4e). Given the rapidity of freezing, it is significant that the wall collars are clearly defined in FESEM images, suggesting that they are an integral component of PD in the functioning state, and not an artefact of chemical fixation. Measurements of tobacco and Arabidopsis PD using these methods have revealed significant differences in pore diameter between these species (Bell and Oparka, unpublished data) and are likely to reveal interesting differences in the dimensions of PD connecting different cell types. Fractures in which the PM is exposed reveal a unique view of the neck of PD, namely from the outer leaflet of the PM looking into the cell (Fig. 4e; Faulkner et al. 2008). In these images, the imprint of the wall collar on the PM can be seen (Fig. 4e), and in suitable views, the desmotubule entering the neck of the pore can be discerned (Fig. 4f, inset). Similar to TEM, high-magnification FESEM views of trichome PD have revealed radially arranged particles on the PM in contact with the wall collar (Fig. 4f inset, c.f. Fig. 2h, i).

FESEM was also used recently by Mullendore et al. (2010) to obtain remarkably clear images of sieve plates from a range of species. Sieve-plate pores are derived from PD (Esau and Thorsch 1985) and as such represent an extreme form of modification of the basic PD structure in which the PD pore becomes enlarged and lined by callose, while the central desmotubule is lost during development (Esau and Thorsch 1985). Mullendore et al. (2010) developed an intricate enzymatic clearing method devised to remove the occluding cytoplasm from sieve tubes following freezing, revealing details of the sieve-plate pores and the callose collars lining them. Significantly, in the absence of rapid freezing, the callose collars occupied a large proportion of the pore diameter (Fig. 5a–d). The scanned surface of the frozen and cleared tissue was then used to empirically derive the width of sieve plates and the diameter of their pores, and these data were subsequently used to derive a new model of phloem flux.

Fig. 5
figure 5

FESEM images of sieve plates after removal of their cell contents. a Proteus vulgaris, scale = 5 μm. b Cucurbita maxima, scale = 20 μm. c Solanum lycopersicum, scale = 5 μm. d Ricinus communis, scale = 20 μm. From Mullendore et al. (2010). Reproduced with permission from the authors and the publisher

Full size image

Fluorescent tags for PD

Viral movement proteins (MPs), fused to green fluorescent protein (GFP), provided the first in vivo tags for imaging PD (Epel et al. 1996). However, it is not yet clear if MPs from different viral groups label exactly the same components of the PD pore, due to fluorescence resolution constraints. For Tobacco mosaic virus (TMV) MP, it seems likely that this MP accumulates, both during infection (Tomenius et al. 1987) and in transgenic plants expressing MP (Moore et al. 1992), in the central cavities of complex PD (Ding et al. 1992b; Roberts et al. 2001). As pointed out by Burch-Smith et al. (this issue), a consensus appears to have arisen that TMV MP labels only secondary PD. However, the labelling pattern of this MP may not be as specific as this, and it is likely that TMV MP is more accurately a marker for those PD, either primary or secondary, within which central cavities have developed. The MP of Potato leafroll virus (MP17) appears to label PD in a different manner to that of TMV, being absent almost entirely from the PD in young, developing tissues whilst predominating in the PD found in older mature tissues (Vogel et al. 2007; Fig. 6a; Burch-Smith et al. this issue). Regardless of the precise localisation of viral MPs within PD, there is no doubting their usefulness in a range of studies of PD (see Lucas 2006).

Fig. 6
figure 6

a MP-GFP associated with PD in a surface view of the tobacco epidermis/trichome interface. Scale = 10 μm. Oparka, unpublished data. b A schematic showing the photoswitching of PD labelled with photoswitchable cyan fluorescent protein. On exposure to UV, the PD are switched from cyan to green. New PD then develop during wall extension (labelled green at 488 nm), resulting in two distinct labelled populations. Later in development, all the PD are again switched to green using UV light. Using this approach, the appearance of new PD can be monitored over relatively long time periods (days). c The leaf epidermis of an Arabidopsis transgenic expressing both PDLP1-mRFP and MP17-GFP. Two distinct populations of PD are present, simple (red) and complex (green). Scale = 20 μm. Fitzgibbon, Faulkner and Oparka, unpublished data. d Point spread function (PSF) from an oil immersion lens. The dimensions of a single PD pore are shown to scale. Note that the entire PD pore falls within the PSF. Modified from Huang et al. 2009 with permission from the author and the publisher. e Interaction between PDLP1 and the MP of GFLV assessed using FLIM. The tubule structure induced by the MP extends from PD and interaction with PDLP1 is strongest within the PD pore. Fluorescence lifetime measurements are illustrated using a false colour code, ranging from 1 (blue) to 3 ns (orange). Ritzenthaler, unpublished data

Full size image

Stains for PD

Aniline blue

Decolourised aniline blue has been used extensively for labelling the callose associated with PD (Currier 1957). However, aniline blue tends to fade rapidly and is cytotoxic to plant cells (Oparka and Read 1994). Aniline blue has been used both qualitatively and quantitatively to image PD (Guenoune-Gelbart et al. 2008) and to determine the amount of callose associated with PD in response to viral pathogen attack (e.g. Iglesias and Meins 2000).

DiOC6

This stain has been used extensively in plant tissues to stain both mitochondria and ER (Oparka and Read 1994). It also stains PD under suitable conditions. For example, in plasmolysed cells, DiOC6 clearly labels the ER that is associated with the PD pore (Oparka et al. 1994). DiOC6 has also been used to stain PD in isolated cell walls derived from Arabidopsis cell suspension cultures (Ritzenthaler et al. 2000).

PD tags arising from proteomics studies

For several years, callose labelling and/or viral MPs remained the only bona fide tags for labelling PD. However, a number of proteomics approaches have begun to yield valuable, previously unknown PD proteins whose functions are the subject of much recent study. For these proteins, fusion to GFP has revealed their unequivocal localisation to PD (Thomas et al. 2008; Simpson et al. 2009) and provided a new suite of PD markers for developmental and physiological studies of PD (Maule 2008). Plasmodesma-localised protein 1 (PDLP1) was isolated from the walls derived from Arabidopsis cultured cells and is a member of a large family of PDLP proteins (Thomas et al. 2008; Fig. 6c). Subsequently, a family of PD callose-binding (PDCB) proteins was isolated using this approach (Simpson et al. 2009; Fig. 7h–j). Interestingly, both PDLP1 and PDCB1 appear to label all PD, including the primary PD present in newly formed cell walls. The presence of different classes of proteins within PD offers the intriguing possibility that these proteins might interact functionally in situ within the PD pore. Such interactions could potentially be monitored using fluorescence resonance energy transfer or fluorescence lifetime imaging (FLIM). It has been shown recently, using FLIM, that PDLP1 interacts at the PD pore with the MP of a tubule-forming virus, Grapevine fan leaf virus (GFLV), perhaps acting as a PD receptor for the viral MP (Fig. 6d; see Amari et al. 2010). Interestingly, the PDLP1 did not interact with the MP of TMV within PD although it did interact with the MP of Cauliflower mosaic virus (CaMV), a second tubule-forming virus. These data suggest that PDLP1 may be a receptor-like protein usurped specifically by viruses whose passage through PD is guided by tubules.

Fig. 7
figure 7

Super-resolution imaging of PD using 3D-SIM. a A conventional confocal image of an Arabidopsis leaf expressing a MP-GFP fusion (MP17; see Vogel et al. 2007). The epidermal walls have been stained red with propidium iodide. PD connecting individual epidermal cells can be seen (arrow) as well as PD grouped into pit fields between the epidermis and the mesophyll (dart). Scale = 15 μm. Fitzgibbon and Oparka, unpublished data. b Glancing TEM section of pit fields showing central cavities separated by small wall partitions. Scale = 1 μm. Oparka, unpublished data. c–f 3D-SIM images of individual PD. MP-GFP (green) is visible in the central cavities and callose (red) has been stained with a callose antibody. Paired callose platelets are clearly visible. Scale = 5 μm. g 3D-SIM image of a single pit field reveals MP-GFP within central cavities similar to those observed using TEM (c.f. b). Scale = 5 μm. h Confocal image of a callose-binding PD protein (PDCB1-mCherry; see Simpson et al. 2009). Epidermal PD can be seen but not resolved. Scale = 20 μm. Oparka and Faulkner, unpublished data. i Immunogold localisation of PDCB1 to the neck region of PD. Scale = 500 nm. From Simpson et al. 2009. Reproduced with the permission from the authors and the publisher. j A 3D-SIM image of PD labelled with PDCB1-mCherry. The association of the protein with individual callose collars can now be resolved. Scale = 5 μm. Oparka and Faulkner, unpublished data. k 3D-SIM image of the parietal layer of a sieve element. The callose collars (red) associated with individual PD pores are resolved, along with MP-decorated strands (MP, green) that interconnect the PD. 3D-SIM images are from Fitzgibbon et al. (2010) and reproduced with permission from the publisher. Scale = 1 μm

Full size image

Imaging different populations of PD

The growing repertoire of PD proteins holds great promise in imaging PD development. Recently, we crossed Arabidopsis plants expressing PDLP1-mRFP with a second transgenic line expressing the MP17 protein of PLRV (Vogel et al. 2007). In the leaves of these plants, all of the simple PD in immature cells were labelled red. As the epidermal cells differentiated, new branched PD, expressing both red and green signals, were detected in a discrete developmental pattern (Fig. 6c; Fitzgibbon and Oparka, unpublished data). It seems likely that as increasing numbers of PD proteins are isolated, some of these will prove to be valuable tags for both primary and secondary PD and some for different PD morphotypes (e.g. simple and complex PD). Another intriguing possibility is that some PD proteins may prove to be unique to specific cellular interfaces within the plant, for example for the specialised PD that occur between the sieve element and companion cell. The identification of such proteins awaits the isolation of PD from specific wall interfaces.

Pulse-chasing PD using photoswitchable probes

In the last 5 years, a large number of photoswitchable and photoconvertible fluorescent reporters have been introduced that hold great promise in the study of PD dynamics and development (reviewed in Lippincott-Schwartz and Patterson 2009; Tilsner and Oparka 2010). This approach will require the selection of specific wall interfaces (e.g. the trichome/epidermis boundary) where all the PD can be imaged in a single plane (Fig. 6a). For example, photoswitchable cyan fluorescent protein (PSCFP) is cyan when imaged at 488 nm laser light and is converted to a green form by a short burst of illumination at 405 nm (see also Chapman et al. 2005). Subsequent imaging at 488 nm reveals two populations of protein, cyan from the unswitched region of the cell and green from within the switched region. The use of transgenic lines expressing PSCFP fused to known PD proteins offers the opportunity of selectively photoswitching populations of PD and tracking their fate over time during wall extension. By photoconverting all of the PD in the same cell wall (cyan to green), the areas of wall within which new PD develop (cyan only) can, in principle, be determined over time (Fig. 6b).

Super-resolution imaging of plasmodesmata

Regardless of the number of new PD proteins that are isolated, the resolution imposed by the diffraction barrier of the emitted light (Abbe 1873) represents a serious limitation to bridging the imaging gap between fluorescence and electron microscopy. Since 2005, however, there has been an explosion in imaging approaches that break the diffraction barrier (reviewed in Huang et al. 2009 and Huang 2010), collectively referred to as ‘super-resolution imaging’ or ‘nanoscopy’.

The diffraction limit

Due to diffraction, light from a point in the object does not converge to an exact point in the image and so is blurred. Blur represents information lost from the final image. Ernst Abbe (1873) formalised the limits of resolution attainable in a diffraction-limited system in the nineteenth century. The mathematical expression for this is d = λ/2n sin α where (d) is the minimum distance between objects required to resolve them as single entities and is a product of the wavelength (λ) of light divided by the numerical aperture of the lens (n sin α) multiplied by 2. A minimum value for d is reached as it is not possible to arbitrarily increase the numerical aperture because light will no longer be able to enter the lens. Nor is it feasible to use wavelengths of light shorter than 400 nm as this can damage biological samples. Abbe’s theorem predicted that objects closer than approximately 200 nm laterally (x–y) and 500 nm axially (z) cannot be resolved.

Whilst the lateral resolution limit is a direct consequence of light diffraction, the much higher axial limit is due to microscope design. Anisotropy results as light is collected from only one side of the specimen, causing an elongation of the point spread function (PSF) in the axial plane (point spread function is the 3D intensity distribution of a point object). Optical sectioning using a confocal or multiphoton microscope allows greater sample penetration and increased resolution by creating an effectively smaller PSF, but the axial resolution remains comparatively poor. Whilst frustrating for cell biologists generally, this is of particular disadvantage to plant scientists who, unless using protoplasts, have to contend with the cell wall. Figure 6e shows the PSF of a light microscope fitted with a high numerical aperture (NA) oil lens. For comparison, the dimensions of a typical PD pore are shown. Note that the entire PD structure falls within the PSF and so cannot be resolved. However, the locations of PD in the wall can be seen due to the fluorescence they emit (e.g. Fig. 7a). Unfortunately, it is not possible to distinguish the number of PD present, nor their architecture, within a single point of fluorescence.

I5M and 4Pi microscopy boost axial resolution by using opposing objective lenses with high NAs. In this scheme, a much greater portion of the full spherical angle is covered and a common focal area is imaged coherently. I5M is a wide field system in which an arc lamp illuminates the sample (Gustafsson et al. 1999). In 4Pi microscopy, a laser is used and the samples are optically sectioned (Hell and Stelzer 1994) producing a 3D resolution of 200 nm. Leica launched a commercial 4Pi confocal microscope in 2005. Due to the optical sectioning capacity, 4Pi is likely to be of most use to plant biologists who are interested in imaging cells within tissues but, as yet, no 4Pi publications using plants have been published. Whilst the increased NA of the coherent objectives boosts axial resolution significantly, it does so within the limits of diffraction set by Abbe’s theorem. To break that limit requires more than a change in microscope design.

3D-structured illumination microscopy (3D-SIM) achieves a resolution of 100 nm in the x–y dimension and 200 nm in z dimension. In optical microscopy, the image produced is a product of the incident light that interacts with the sample, as represented by the distribution of the dye/fluorescent protein. Gustafsson (2000) proposed that by introducing spatial structure to the incident light, it is possible to reveal a greater level of detail in the unknown sample structure. The interaction between the structured incident light and the structure of the sample produces emitted light that contains interference fringes (Gustafsson 2000, 2005). The interference fringes can be much courser than either of the original structural patterns and as such are more readily observable even if one, or both, of the original patterns are too fine to resolve. Through processing a series of images, it is possible to extract additional information about the sample, thereby generating an image with improved resolution. Originally demonstrated in 2D, 3D imaging capacity soon evolved. To achieve this, laser light is first diffused and polarised before passing it through a diffraction grating where it is split into two parallel beams. These beams then enter the objective where they are focused to interfere at the sample plane. This interference generates a sinusoidal illumination pattern. Multiple images are taken by scanning and rotating the excitation pattern, and these are processed linearly to reconstruct an image with up to twice the normal resolution of a conventional light microscope. In addition to providing extended resolution, 3D-SIM is also compatible with standard fluorochromes (e.g. GFP, mCherry, DAPI) and with the sample preparation methods commonly used for light microscopy. In 2008, the application of 3D-SIM was first reported in a multi-colour 3D study of the nuclear periphery (Schermelleh et al. 2008). Concomitantly, a commercial 3D-SIM platform was released by Deltavision, called OMX (Optical Microscope eXperimental).

3D-SIM of plasmodesmata

We recently used 3D-SIM in a study of Nicotiana tabacum PD constitutively expressing the TMV MP (Fitzgibbon et al. 2010). By double labelling the PD with a red callose antibody, we were able to show that these two signals (callose and MP) could be resolved easily (Fig. 7c–f). The width of the callose collar was estimated to be about 150 nm, slightly larger than the raised wall collar seen in FESEM images (c.f. Fig. 4e). We were also able to resolve an unlabelled region of the pore, about 135 nm in length, between the callose collar and the MP within the central cavity (Fig. 7c–f). 3D-SIM also offered unique views of central cavities when PD pit fields were optically sectioned (Fig. 7g). Here, central cavities closely resembled those images seen only rarely in the TEM in glancing sections of the cell wall (c.f. Fig. 7b). More recently, we have used 3D-SIM to image the localisation of the putative callose-binding protein, PDCB1. Immunogold labelling of this protein showed it to be localised to the neck region of PD (Fig. 7i; Simpson et al. 2009), a feature confirmed using 3D-SIM (Fig. 7j). Within the phloem, we were able to resolve sieve-plate pores and the specialised pore-PD that interconnect the companion cell with the sieve element. Significantly, we were able to image individual callose-lined pores leading into the sieve element parietal layer. In this region, we noticed that the TMV MP was no longer associated only with the central cavities of PD but became distributed on a tubular material that interconnected all the pore-PD in the sieve element parietal layer (Fig. 7k). These unique views of the sieve element–companion cell interface cannot be obtained using confocal microscopy and are seen only rarely in fortuitous glancing sections of this interface.

Although lacking the capacity of 3D-SIM to resolve fine detail, confocal microscopy is used routinely to ascribe subcellular locations of proteins fused to GFP (Escobar et al. 2003). Liu et al. (2008) recently highlighted an important limitation when using confocal microscopy to identify PD-associated proteins. A dramatic distortion of the PSF occurs when imaging close to the cell wall due to the different refractive indices of the cell wall and the cytosol. A consequence of this is that a single point of fluorescence may reflect across the cell wall resulting in the appearance of paired punctae. Liu et al. (2008) proposed a simple polarisation method to recognise reflected signals that would allow fluorescent artefacts to be identified. The increased axial resolution and optical sectioning capacity of 3D-SIM ensures that it is more robust in identifying plasmodesmal associated proteins than confocal microscopy. Thus, 3D-SIM offers great potential in bridging the gap between confocal and electron microscopy.

Other super-resolution imaging approaches also offer potential but have yet to be tested on plant tissues. Saturating structured illumination microscopy (SSIM; Gustafsson 2005) retains compatibility with conventional fluorophores and achieves true super resolution. SSIM uses structured light but of very high intensity to saturate the emission of stimulated fluorochromes. As the saturation level is approached, the fluorescence emission is no longer proportional to the excitation light intensity. The effective resolution is thus limited only by the fluorescent lifetime of the probe and not by the exciting light. SSIM is not yet available commercially.

Stimulated emission depletion microscopy

Stimulated emission depletion (STED) was the first far-field super-resolution imaging technique to be applied to cell biology (Klar and Hell 1999). Like SSIM, it saturates fluorochrome emission but does so using a second, STED laser. Unlike SSIM, which is a wide field technique, STED uses point scanning. Through the introduction of a polymeric phase plate, the STED laser can be modified to contain a focal doughnut (Hein et al. 2010). This features an intensity zero at its focal centre and strong intensities at the periphery. The excitation laser is then precisely co-aligned to occupy the STED zero point. The STED beam is shifted towards the red end of the fluorochrome emission spectrum and so can de-excite potentially excited molecules through stimulated emission. Stimulated emission occurs when an excited state fluorochrome encounters a photon that contains the energy difference between the excited and ground state. Consequently, the fluorochrome is brought back to the ground state before emission occurs. Although reliant upon the pattern generated by two diffraction-limited laser light sources, STED is truly super resolution as the size of the STED pattern is limited only by the power of the STED laser. Furthermore, it is theoretically possible to use any fluorochrome in a STED study (Willig et al. 2006). Whilst organic dyes typically offer the greatest photostability under STED conditions, imaging parameters have been optimised for fluorescent protein imaging (Willig et al. 2006). By coupling STED with 4Pi microscopy, super resolution has been achieved in all three dimensions (Schmidt et al. 2008). Using STED, synaptic vesicle movement has been dissected at video rate (Westphal et al. 2008), and more recently, connexon-43 clusters were imaged crossing the cell membrane (Hein et al. 2010). Significantly, Leica Microsystems has recently launched a STED module that can upgrade its existing SP5 confocal platform. A super-resolution ‘bolt-on’ to an existing imaging system that is integrated with a familiar user interface is likely to prove popular amongst researchers.

The STED microscope has evolved significantly since it was first described (Hell and Stelzer 1994). Table 1 highlights its capacity to image in 3D, to achieve exceptionally high lateral resolution and image at a video rate. Whilst no published STED platform has integrated all these specifications, any one has the potential to contribute significantly to plasmodesmal research. For example, the capacity to image in 3D at high resolution could allow structural dissection of the PD pore, in a similar vein to that described by Fitzgibbon et al. (2010) whereas super-resolution imaging at a biologically relevant temporal resolution could be a powerful tool in unravelling plant virus dynamics.

Table 1 A comparison of the spatial and temporal resolutions of biological imaging techniques
Full size table

Single molecule imaging

STED and SSIM are super-resolution techniques that image an ensemble of molecules. It is worth noting that super resolution can also be attained by imaging single molecules (reviewed in Patterson et al. 2010). The principle underlying single molecule detection is that the position of a single emitter can be localised with a high degree of accuracy if enough photons are collected. This allows the centre of the fluorescent emission to be determined by fitting the photon output to the ideal PSF. Photoactivation localisation microscopy (PALM), stochastic optical reconstruction microscopy (STORM) and fluorescence photoactivation localisation microscopy are all examples of techniques that use this same principle. All independently conceived, publications introducing these technologies first appeared in 2006, and although not yet used on plant cells, it is worth considering these techniques briefly.

To detect a single emitter from a sample containing many fluorescent molecules requires a probe with requisite photophysical properties (Lippincott-Schwartz and Patterson 2009) and a sensitive detection system, usually a CCD camera. Dronpa is a monomeric fluorescent protein used in single molecule detection experiments (Andro et al. 2004; Habuchi et al. 2005). It photoconverts from a green fluorescence to a dark state in response to prolonged 488 nm laser irradiation, with a sharp burst of 405 nm laser light restoring the initial green fluorescence. In single molecule studies, activation is stochastic, and the light–dark cycling allows imaging and localisation without overlap. Through iteration of activation and imaging, a super-resolution image is constructed from the coordinates of many molecules. Resolution is directly dependent upon the total number of photons collected, and to that end, it is essential to maximise the signal to noise ratio. In addition to using probes with a high dynamic range and contrast ratio, the imaging platform is frequently integrated with a total internal reflection (TIRF) microscope. In TIRF (Axelrod et al. 1984), the angle of incident light is such that it cannot cross the sample refractive index and so is deflected back into the objective. When the incident light hits the coverslip, an evanescent wave is generated that decays exponentially, and so TIRF penetrates to only around 100 nm. Imaging in a limited focal volume limits background fluorescence from the bulk tissue increasing the signal to noise ratio by reducing noise. A lateral (x–y) resolution of 20 nm has been reported for PALM (Bates et al. 2007).

The developers of STORM were the first to generate a 3D image using single molecule detection (Huang et al. 2008). They did so by using a cylindrical lens to generate an elliptical image of a single fluorophore. From this, it was possible to measure the axial (z) position from the ellipticity and the lateral (x–y) position from the centroid. A resolution of 25 nm laterally (x–y) and 50 nm axially (z) was achieved (Huang et al. 2008).

Although capable of imaging live cells (Biteen et al. 2008) and in 3D (Huang et al. 2008), single molecule detection techniques are currently of limited practical use. The requirement for fluorescent probes with specific photophysical capacities (Lippincott-Schwartz and Patterson 2009) would require most labs to reengineer their suite of reporter proteins. Also, the necessity to have samples that produce low or no background provides another challenge to plant scientists (Moreno et al. 2006). Whilst TIRF limits bulk tissue auto-fluorescence, through reducing the focal volume, it does require areas of interest to be within approximately 100 nm of the coverslip.

A comparison of the spatial and temporal resolutions achievable with different types of optical-based microscopy is shown in Table 1. Note that whilst some have greatly improved x–y resolution, others show great improvements in z resolution. The final choice of system is likely to be dictated by the experimental system under study. Of the plethora of super-resolution approaches currently being developed, it will be interesting to see which of these will enter mainstream plant cell biology. Undoubtedly, many of these emerging techniques are applicable to the study of PD. However, it will require inventive tissue preparation and imaging conditions to achieve this goal.

Image processing

Although conventional imaging systems (CSLM and wide field) produce images with inherently less detail than their super-resolution counterparts, it is possible to extend the resolution significantly using imaging processing techniques. Through removing optical distortions, present in all images (regardless if they are collected by a super-resolution or a ‘conventional’ platform), it is possible to gain greater information about the sample. Many different image processing methods are available, and these differ largely in the algorithm used to process the raw data (Conchello and Lichtman 2005). One example of particular relevance is near neighbour subtraction (NNS). NNS is beneficial when the fluorescent signal is concentrated in thin filaments or discrete patches, as in PD (Conchello and Lichtman 2005). Using a 2D PSF, calculated for a miss-focus distance equal to the distance between the optical sections, NNS blurs the adjacent sections and then subtracts them from the section of interest to remove the out-of-focus light (Conchello and Lichtman 2005). This is then repeated for each of the remaining sections in turn before projecting the final image, minus much of the extraneous light. As computing powers increase, it will be possible to derive ever more accurate methods to process images using less simplifying assumptions. A limit is reached, however, as it is not possible to estimate frequency components of the specimen that are not passed by the objective (Conchello and Lichtman 2005).

Correlative light and electron microscopy

Because the boundaries between light and electron microscopy are blurring, there is a growing need to confirm that structures seen in the light microscope are indeed identical to those observed by TEM. Accordingly, a new area of biological imaging, correlative light and electron microscopy (CLEM), is emerging in which the same samples are viewed sequentially by light and electron microscopy (Koster and Klumperman 2003, reviewed in Tilsner and Oparka 2010). This approach was used to image mitochondria expressing a fluorescent reporter using both PALM and TEM (Betzig et al. 2006). To retain fluorescence, the cells were frozen and cryosectioned before imaging. The PALM images were found to superimpose exactly on the TEM images, confirming the resolution of the PALM technique and accurately ascribing the correct protein localisation.

Photo-oxidation is a method that translates a fluorescent signal to an electron-dense one. Upon illumination, virtually all fluorochromes produce singlet oxygen. Photo-oxidation harnesses singlet oxygen to convert the chromogen DAB (3,3′-diaminobenzidine tetrachloride) to a highly localised osmiophilic granular precipitate (Maranto 1982). The conversion takes place in fixed tissues that have been incubated in an oxygenated DAB solution (Meisslitzer-Ruppitsch et al. 2009). Upon illumination with high intensity excitation light, reaction products develop, visualised as a gradual browning under bright field. Following treatment with osmium tetroxide, the samples can then be processed for EM as normal (Maranto 1982). The success of the photo-oxidation reaction depends largely upon the capacity of the fluorochrome to generate sufficient singlet oxygen. Lucifer Yellow was the first fluorochrome used (Maranto 1982), but subsequent studies have used other fluorochromes due principally to their greater singlet oxygen yield (Geipmans et al. 2006; Meisslitzer-Ruppitsch et al. 2009). The omnipresent GFP has been used successfully to catalyse the photo-oxidation reaction but only in a few studies (Grabenbauer et al. 2005; Meisslitzer-Ruppitsch et al. 2009). Aside from conventional TEM imaging, photoconverted samples have also been used in electron tomography studies (Cortese et al. 2009).

A method that relies on conventional fluorescent labels and a light catalysed reaction to generate an electron-dense product is appealing. It would therefore seem worthwhile to invest time in testing the suite of markers already available for their photo-oxidising potential. As the capacity for generating singlet oxygen varies and DAB precipitate is best preserved when concentrated within an organelle or a limited area, this approach might be particularly suited to PD.

Many findings based on fluorescent protein imaging need support from EM (Lippincott-Schwartz and Patterson 2009). By using a super-resolution method in a correlative approach, a direct comparative analysis of imaging capabilities is possible. Perinetti et al. (2009) correlated 4Pi images of Golgi stacks with TEM serial sections and 3D reconstruction. They first mapped the positions of separated Golgi stacks with a wide field microscope before imaging them in 3D using confocal and then 4Pi microscopy. The correlation was achieved by detecting a fluorescent Golgi protein using a nanogold conjugated antibody, followed by careful TEM serial sectioning and recording. Whilst this approach required a second step to make the fluorescent signal visible in the EM, a bifunctional probe allows the sample to be directly detectable at both the light and EM level.

FluoroNanogold is an example of a dual labelling reagent that contains fluorescent and electron-dense particles. It is a Fab probe that is covalently bound to a fluorochrome and gold particle. As it is an immunoglobulin fragment (the fragment of antigen binding), it is smaller, and as such can yield enhanced labelling efficiency. Greater penetration is also afforded by the comparatively small gold conjugate (between 0.8 and 1.4 nm; Robinson and Takizawa 2009). A consequence of this is that silver enhancements may be needed to increase the size of the gold particles to allow ready light microscope visualisation. As immunological microscopy is already a routine lab technique, with an ever-expanding toolbox of antibodies, using bifunctional probes in place of standard secondary antibodies is a potentially high-return imaging investment.

As studies of PD advance, there will be a growing need to ascribe the location of newly discovered PD proteins with accuracy to their native location within the PD pore. EM immunogold labelling is one approach to achieving this, but will require the production of antibodies for each new protein detected. Given the small size of PD, immunogold labelling is subject to its own suite of problems, including the need for stringent controls and the inherent problems associated with different gold particle sizes. A simpler solution would be to conduct correlative imaging in which fluorescent PD protein fusions are imaged first using a super-resolution instrument, followed by correlative TEM in which the protein is rendered electron dense. We have shown that using 3D-SIM, proteins can be ascribed to different locations within PD. The challenge for the future is to accurately ascribe PD proteins to their correct location and to conduct meaningful functional studies on the role of these proteins within PD.