Endocytosis against high turgor: intact guard cells of Vicia faba constitutively endocytose fluorescently labelled plasma membrane and GFP-tagged K+-channel KAT1


For correspondence (fax +49 6151 164630; e-mail tobias.meckel@web.de).


The relevance of endocytosis in plants against high turgor pressure has frequently been questioned on the basis of energetic considerations. Here, we examine the dynamics of the plasma membrane (PM) in turgid guard cells of Vicia faba by monitoring with confocal microscopy the fate of fluorescent styryl dyes (FM1-43, FM2-10 and FM4-64). As a second marker, we also observe the retrieval of a fluorescent chimaera of the K+-inward rectifying channel from Arabidopsis thaliana and the green fluorescent protein (KAT1::GFP). Analysis of cytoplasmic structures, which became labelled by the different styryl dyes, revealed that only FM4-64, the most hydrophobic dye, was a reliable marker of endocytosis, whereas the two other styryl dyes resulted also in an unspecific labelling of different cytoplasmic structures including mitochondria. Over some minutes of incubation in continuous presence of these dyes, endocytic vesicles in the cortical cytoplasm beneath the PM were fluorescently labelled. The identification is based on the observation that the size distribution of these structures is very similar to that of endocytic vesicles obtained from patch-clamp capacitance recordings. Also, these structures are frequently co-labelled with KAT1::GFP. Taken together, the data show that turgid guard cells undergo vigorous constitutive endocytosis and retrieve membrane including the K+-channel KAT1 from the PM via endocytic vesicles.


Opening and closing of stomata is an osmosis-driven process. Following the accumulation or discharge of K+ salts, water enters or leaves the cells, respectively. As a consequence of these water fluxes, guard cells undergo large changes in turgor pressure and volume, which finally result in opening or closing of the stomatal pore. The large changes in cell volume during stomatal movement are thought to be associated with substantial excursions in the surface area of the guard cell plasma membrane (PM). Previous investigations with guard cell protoplasts have shown that these cells accomplish an increase in surface area by incorporating exocytic vesicles into the PM. Vesicles of about the same size are retrieved in an endocytic process to achieve a reduction of the surface area when cells shrink (Homann, 1998).

It has been a matter of debate whether intact plant cells are able to perform endocytosis against the generally high turgor pressure in plant cells (Cram, 1980; Gradmann and Robinson, 1989; Robinson et al., 1992; Saxton and Breidenbach, 1988). The hurdle of turgor pressure for endocytosis is in guard cells even more pronounced than in other plant cells. It is therefore not possible to a priori extrapolate the data on endocytic activity in guard cell protoplasts to intact guard cells. Indeed, uptake of extracellular markers has so far only been shown in intact plant cells with a relatively low turgor pressure such as cells of Nicotiana tabaccum cv. bright yellow-2 (BY-2) (Emans et al., 2002), root apices of Zea mays (Baluska et al., 2002) and pollen tubes of Lilium longiflorum (Parton et al., 2001).

The available data question the relevance of endocytosis in turgid guard cells. For example, no uptake of the fluid phase marker lucifer yellow could be detected in intact guard cells (Diekmann et al., 1993). Also, electron micrographs of closed stomata showed no occurrence of vesicles, which would point to an endocytic activity (Diekmann et al., 1993). In a recent publication, Shope et al. (2003) were able to show a general uptake of the endocytic marker FM4-64 in osmotically shrinking guard cells. However, no direct evidence for the formation of endocytic vesicles was found.

In the present study, we have chosen for a number of reasons intact guard cells to investigate the possibility of endocytosis in turgid plant cells. First, while normal plant cells have a turgor pressure of about 1 MPa or less, cells of open stomata have pressure values of up to 4.5 MPa, which are among the highest known in plants. Even cells in closed stomata have pressures in the range of 1 MPa (Franks et al., 2001). Second, unlike in the majority of plant cells, shrinking is a normal physiological process for these cells. Closing of stomata and the concomitant reduction in PM surface area can be triggered by known physiological stimuli like abscisic acid, darkness, or high CO2 concentrations (Fricker and Willmer, 1996). Finally, electrophysiological investigations on guard cell protoplasts have revealed an estimate for the size of the expected endocytic vesicles (Homann and Thiel, 1999). These data offer a reference point for the interpretation of data from intact cells.

Following a preceding study with guard cell protoplasts (Kubitscheck et al., 2000), we examine in the present work the process of endocytosis in turgid cells using styryl dyes and confocal microscopy. Among the numerous styryl dyes available, FM1-43, FM2-10 and FM4-64 are the most frequently used to study endocytosis. With few exceptions (Emans et al., 2002; Shope et al., 2003), these dyes have, in the case of plant cells, only been applied to single, isolated cells. Therefore, we started off investigating which of these dyes, which differ mainly in their hydrophobicity, is best to study membrane dynamics in intact guard cells. The requirements for the dye are a sufficient hydrophilicity in order to pass the hydrophilic cell wall, while at the same time, it should be hydrophobic enough to intercalate into the PM with sufficient affinity. But, in order to report the endocytic pathway correctly, it is of capital importance that the dye must not penetrate a biomembrane, at least in the time window of observation.

The present data show that for investigations on intact guard cells of Vicia faba, only FM4-64 of the tested styryl dyes can be used as a reliable marker to follow endocytic processes. This is in contrast to the successful application of numerous derivatives of these dyes on animal cells. Using this dye, we are able to show that endocytosis occurs against high turgor and that PM proteins are endocytosed by vesicles.


Comparison of staining with different FM dyes

Styryl dyes are frequently used for staining the PM and monitoring endocytosis in living cells. In order to find the appropriate FM dye for investigating endocytosis in turgid guard cells, we first compared staining of these cells incubated in FM1-43, FM2-10 and FM4-64 with respect to their toxicity.

After 10 min of incubation, all three dyes were evenly distributed in the PM. Apparently, all dyes needed about the same time to stain the PM, despite the differences in their hydrophobicity. However, while it was possible to remove the least hydrophobic dye FM2-10 from the PM and the cuticle by a 1-h wash-out, this was impossible with the most hydrophobic dye FM4-64 even after washing for more than 6 h. The cuticle remained brightly stained and most likely acted as a dye buffer storage, which also kept the PM stained (data not shown). In continuous presence of dye, staining of cytoplasmic structures was observed after incubation for more than 2 h for all three styryl dyes (details see below).

While the three dyes exhibited an overall similar staining kinetic, they revealed marked differences in their toxicity. Treatment with FM1-43 for 2 h resulted in serious damage of more than 50% of the cells. In damaged cells, the cytoplasmic streaming came to a halt. In most of these cells, the PM lost its integrity, which was inevitably paralleled by a rapid and intense staining of all intracellular membranous compartments. A visible loss of turgor in many stained cells provided an additional hint for a leaky PM (Figure 1b, right cell). As FM1-43 was used in a concentration lower than its competitors, it clearly turned out to be the most toxic dye in our study. In contrast, no perceivable damage was visible after treatment with FM2-10 for 2 h (Figure 1e). Even after 12 h in FM2-10 and subsequent high exposure to the laser excitation (e.g. during 3D scans), more than 90% of the cells were still unaffected in terms of cytoplasmic streaming and cell shape. There was also no sign of unspecific intracellular staining. However, if the dye was used in a fivefold concentration (100 µm), damage was observed after 6 h only. Incubation in FM4-64 gave an intermediate result. The dye caused visible damage in about 20% of the cells after incubation for 2 h and in about 40% after 12 h.

Figure 1.

Size distributions of fluorescently labelled structures.

Size distribution histograms for intracellular structures labelled with FM1-43 (a), FM2-10 (d), FM4-64 (g), Mitotracker Red CMX Ros (m) and mitochondria measured in electron micrographs of guard cells (left column). The histograms were fitted with a multiple lognormal function (solid line). Parameters of these fits are listed in Table 1. The chemical formula of the respective dye and the number (n) of measured structures is given beneath each histogram. Right column: examples of single optical paradermal confocal sections through the aequatorial (left) or cortical (right) region of guard cells labelled with the respective dyes. In the electron micrograph (q), mitochondria and lipid bodies are visible. Scale bars for fluorescent images = 10 µm, and for the electron micrograph = 500 nm.

All in all, toxicity was found to be correlated to dye concentration, quantum yield of the dye and duration of incubation.

Cytoplasmic structures labelled by FM dyes in intact cells

The analysis of cells without any visible damage revealed that the different styryl dyes resulted in labelling of cytoplasmic structures of different size. For analysis, sizes of a large number of fluorescent structures were measured and plotted in a histogram. The data were fitted by multiple lognormal distributions in order to estimate the mean diameters and the relative distributions of the individual structures (Figure 1; Table 1).

Table 1.  Each size distribution histogram illustrated in Figure 1 was fitted with a multiple lognormal function, yielding these fit parameters
 FM1-43FM2-10FM4-64MitotrackerMitochondria (EM)
  1. The center (xc), width (w) and amplitude (A) for each peak are given.

xc1290.15 ± 2.01291.29 ± 6.13311.15 ± 1.91  
w10.14 ± 0.010.18 ± 0.020.23 ± 0.01  
A19.15 ± 0.4210.82 ± 0.6529.86 ± 1.78  
xc2 479.58 ± 7.35472.37 ± 2.48474.55 ± 4.56 
w2 0.10 ± 0.020.11 ± 0.010.08 ± 0.01 
A2 12.82 ± 1.0116.57 ± 1.437.77 ± 0.81 
xc3601.22 ± 3.98601.03 ± 4.14 606.01 ± 4.55597.71 ± 1.14
w30.09 ± 0.010.09 ± 0.01 0.09 ± 0.010.04 ± 0.01
A35.99 ± 0.2434.10 ± 2.36 19.48 ± 1.3617.53 ± 1.21
xc4729.56 ± 5.47731.81 ± 9.38 730.57 ± 7.04722.36 ± 2.65
w40.08 ± 0.010.14 ± 0.01 0.09 ± 0.010.07 ± 0.01
A43.72 ± 0.1927.97 ± 1.67 12.96 ± 0.7310.58 ± 0.39

To determine the lower end of possible size measurements with the confocal technique, fluorescent beads with a diameter of 100 nm were imaged through the cuticle to mimic the conditions under which all fluorescent signals in guard cells were recorded. The full width at half maximum (FWHM) peak height of the resulting diffraction-limited spot was measured in order to determine the dimensions of the point spread function (PSF) under this condition. In the recorded images, the beads appear to have a lateral diameter of 275 ± 6 nm (Figure 2) and an axial diameter of 1131 ± 28 nm (pinhole size: 1 airy). These values are close to those determined for the same measurements in Cammelia communis (White et al., 1996). Thus, any – even infinite small – structure, which is bright enough to give a signal, appears to have at least these lateral and axial dimensions in our confocal recordings.

Figure 2.

Size measurements of fluorescently labelled structures.

The distribution of signal intensity against the x and y dimensions of fluorescent images of a bead (a) and of a single vesicle (b) are shown. The bead had a diameter of 100 nm and was imaged through the cuticle to mimic the conditions under which all fluorescent signals in guard cells were recorded. Both distributions were fitted with a Gaussian function (red net) in order to obtain the FWHM of the signal intensity distribution, which corresponds to the diameter of the structures.


In those cases where cells incubated with FM1-43 remained intact, the dye was detected in numerous big cytoplasmic (Figure 1b, left cell) and small cortical (Figure 1c) structures. Because of the toxicity of FM1-43, the number of cells contributing to this size distribution histogram was lower than that for the other dyes. Nevertheless, analysis of the size distribution of the stained structures (Figure 1a) revealed three clear peaks. The first peak at approximately 290 nm corresponds to small cortical and the peaks at approximately 600 and 730 nm to big cytoplasmic structures. Close inspection of the distribution also revealed a very small peak at approximately 470 nm.


Like in FM1-43-stained cells, the most prominent cytoplasmic target of FM2-10 were big cytoplasmic structures (Figure 1e) with diameters of approximately 600 and 730 nm (Figure 1d). This dye showed the brightest staining of these structures. A close scrutiny of FM2-10-stained cells also shows very small cortical and intermediate structures, which accumulate this dye (Figure 1f, arrowheads and arrows, respectively). The respective histogram (Figure 1d) shows corresponding peaks at approximately 290 and 470 nm, respectively. Thus, according to their size, the small structures can clearly be separated into two groups.


Spherical cytoplasmic structures can also be detected after incubation with FM4-64. However, in this case, mainly intermediate and small structures are detectable (Figure 1h,k, respectively). The histogram reveals that the structures resemble those of small cortical and intermediate structures stained by FM1-43 and FM2-10 (Figure 1g). The very small cortical structures give a peak at approximately 310 nm, the intermediate structures at approximately 470 nm.

After incubation for 4 h in FM4-64, it was possible to detect labelling also of the tonoplast. This was not the case with the two other dyes. Figure 3(a) shows one optical slice of an FM4-64-stained cell with a labelled PM and tonoplast. In three successive optical slices of a different cell, the small FM-labelled structures can clearly be located in the cytosol, i.e. between the brightly stained PM and the lesser bright tonoplast (Figure 3b–d). In a brightest-point-projection, it can further be seen that many structures are distant from the PM in a cytoplasmic strand (Figure 3e), which were found to have an average size of 465 ± 2 nm (n = 5).

Figure 3.

Staining of the tonoplast by FM4-64.

(a–d) Paradermal confocal sections of one guard cell (a) and three consecutive sections of a different cell (b–d) labelled with FM4-64 for 4 h (a–d). The brightly stained PM (long arrow), vesicles (arrowheads) and the less bright labelled tonoplast (short arrows) are displayed. Because of the chlorophyll, fluorescence in the detection band of FM4-64 chloroplasts (asterisks) are also visible.

(e, f) A cytoplasmic strand (arrowhead) filled with intermediate sized structures can bee seen in the fluorescent (e) and transparent (f) channels of a brightest-point-projection of 15 consecutive paradermal sections. The structures have a mean diameter of 465 ± 2 nm (n = 5).

Scale bars = 10 µm.

In FM4-64-stained cells, only very few big structures were observed after prolonged incubation. In addition, some structures, which fall in a very broad size distribution ranging from 600 to 1100 nm (Figure 1g), were observed.

Identification of the stained cytoplasmic structures

Lipid bodies

Guard cells contain numerous lipid bodies, which could, in principle, be the target for the styryl dyes, because of their hydrophobic nature and predicted size. These structures can be distinguished from other organelles by their contrast in differential interference contrast (DIC) microscopy (Figure 4b, arrows). Inspection of FM2-10-stained guard cells showed that these lipid bodies were never fluorescent (Figure 4a,c). They can therefore be excluded as targets of the styryl dyes.

Figure 4.

Lipid bodies are not labelled by FM2-10.

Numerous structures, which are labelled by FM2-10 (a, arrowheads) after 2 h do not co-localise with lipid bodies (b, arrows), which can be identified in DIC images by their contrast. In the merged image (c), a red colour table was applied to the DIC image (b) for better contrast. Scale bar = 10 µm.


The chemical structure of the FM dyes is similar to that of dyes used to label mitochondria (e.g. 4-Di-ASP). To examine the possibility that the large FM-labelled structures in the guard cell cytoplasm are mitochondria, we exposed cells to FM2-10 or FM1-43 and Mitotracker Red CMX Ros. Overlays of both detection channels show that the Mitotracker clearly co-localises with FM2-10 and FM1-43 (Figure 5). In contrast to FM dyes, the Mitotracker labels neither the PM nor the small cortical structures of around 300 nm (Figures 1n,o and 5c,d). Note, however, that the Mitotracker accumulates in the lumen of the vacuolar compartment (Figure 5, asterisks), whereas the tonoplast was never found to be stained by FM2-10 or FM1-43.

Figure 5.

Co-localisation of the Mitotracker with FM2-10 and FM1-43 and the dependency of staining on membrane potential.

(a–f) Double labelling of guard cells with FM1-43 (a) or FM2-10 (b) and Mitotracker (c,d) reveal a clear co-localisation of each styryl dyes with the Mitotracker in the same cytoplasmic structures (e,f) but not in the PM, which is not stained by the Mitotracker (arrowheads). Note that after more than 3 h of incubation, the Mitotracker also accumulates in the lumen of vacuolar compartments (asterisks).

(g, h) In intact guard cells were pre-treated with 10 mm sodium azide for 10 min; neither the Mitotracker (g) nor FM2-10 (h) stains these structures. FM2-10 still brightly labels the PM (h, arrowhead).

Scale bars = 10 µm.

The size distribution histogram of structures stained with the Mitotracker is very similar to that of cells treated with FM1-43 and FM2-10, except that it lacks the peak at approximately 300 nm (Figure 1m). In addition to the most prominent peaks at approximately 600 and 730 nm, a minor peak at approximately 470 nm is also present. Only few structures have sizes between 800 and 1200 nm and more. However, the size distribution for mitochondria obtained from electron micrographs in turn lacks the peak at approximately 470 nm (Figure 1p). Hence, only the peaks at approximately 600 and 730 nm are caused by mitochondria. As these organelles have an elongated shape, our size measurement method reports two diameters.

The co-localisation of the Mitotracker and FM2-10 and striking similarities of the size distributions for both markers strongly suggest that the large cytoplasmic structures stained with the styryl dyes FM2-10 and FM1-43 are mitochondria. On the other hand, the intermediate structures with a diameter of approximately 470 nm are obviously not mitochondria; the ability of the Mitotracker to stain them may point to a (pre-)vacuolar nature of these compartments.

The ability of the cell-permeant Mitotracker Red CMX Ros to accumulate in active mitochondria after passively diffusing across the PM depends on the charging of the mitochondrial membranes (Haugland, 2002). To examine whether labelling of mitochondria by FM dyes occurs via the same pathway, we monitored the staining pattern in cells where the mitochondrial membrane potential had been depolarised by treatment with sodium azide. The results reported in Figure 5(g,h) show that neither the Mitotracker nor FM2-10 reveals under this condition any appreciable staining of cytoplasmic structures. This implies that staining of the mitochondria by FM dyes also depends on the mitochondrial membrane potential and is not associated with endocytic processes.

The small structures are endocytic vesicles

The examination of staining with different FM dyes implies that only the small structures of approximately 290/310 nm can be considered as a result of endocytosis. To further examine the possibility that these structures are indeed of endocytic origin, we expressed a GFP chimaera of the plant potassium inward rectifier (KAT1) in guard cells. Complementary electrophysiology studies on guard cell protoplasts demonstrate that this channel is inserted into the PM via exocytosis and retrieved by endocytosis during pressure-driven volume changes (Hurst et al., 2004).

Figure 6(a) shows a typical image of a KAT1::GFP-expressing guard cell. Abundant green fluorescence can be detected at the cell periphery, suggesting its PM localisation. This localisation is confirmed by a labelling of the PM of the same cell with FM4-64 dye (Figure 6b), which co-localises with the GFP fluorescence (Figure 6c). GFP labelling can also be detected in the form of small spherical structures inside the cytoplasm (Figure 6d, arrowhead). An overlay of the images reveals for many of these structures a clear co-localisation of the GFP and FM4-64 signals (Figure 6c,f, arrowheads). The obvious interpretation is that those structures, which carry both labels, are endocytic vesicles, which retrieve KAT1 channels from the PM.

Figure 6.

Co-localisation of FM4-64 and KAT1::GFP endocytic vesicles.

(a, b) A brightest-point-projection of seven consecutive sections of a guard cell expressing KAT1::GFP (a), additionally labelled with FM4-64 (b), reveals a co-localisation of both fluorescent signals in small cortical structures (arrowheads). Scale bar = 10 µm.

(d–f) The corresponding magnifications show that these structures are distinct from the PM. Not all endocytosed vesicles carry GFP tagged KAT1 molecules (arrows). Scale bar = 5 µm.

(g) The size distribution histogram of structures labelled with KAT1::GFP, FM4-64 or endocytotic steps recorded with the patch clamp technique (data from Figure 4a provided by Homann and Thiel, 1999) reveals a high similarity of the three different size measurements. The parameters of the lognormal fits are listed in Table 2.

Previous electrophysiological measurements revealed that exo- and endocytosis of small vesicles accommodate for osmotically driven changes in surface area in guard cell protoplasts (Homann and Thiel, 1999). We therefore examined whether the small fluorescent structures labelled with FM4-64 and/or with KAT1::GFP have any relation to the sizes of single vesicles determined by the electrophysiological assay. Figure 6(g) illustrates a histogram for the size distributions of vesicles obtained from the fluorescent images and from the high-resolution capacitance recordings. The data show that GFP and FM4-64-labelled structures as well as those from the electrophysiological assay have a very similar distribution. This is underlined by the values obtained from fitting the size distributions with the lognormal distribution (Table 2). The similarity of the size distributions is true not only for the peaks (xc) but also for parameters defining the shape of the fitting functions (A, w). The difference of the peaks between GFP- and FM4-64-labelled vesicles can be explained by the longer emission wavelength of FM4-64 (λem = 650 nm) with respect to GFP (λem = 510 nm) (see Discussion).

Table 2.  Parameters of the lognormal functions, fitted to the size distribution histograms illustrated in Figure 6
 FM4-64KAT1::GFPPatch clamp
  1. The center (xc), width (w) and amplitude (A) for each peak are given.

xc312.79 ± 2.97275.23 ± 4.01277.59 ± 3.26
w0.34 ± 0.020.41 ± 0.020.48 ± 0.02
A106.80 ± 4.32107.68 ± 3.79106.48 ± 3.09


The quantum yield of styryl dyes increases by more than two orders of magnitude upon partitioning into a lipid environment, and they are thought to be membrane impermeable (Cochilla et al., 1999; Henkel et al., 1996). Because of these properties, FM dyes are frequently used for monitoring endocytosis in living cells. The appearance of fluorescent label inside cells is generally taken as evidence for endocytic activity, and the respective fluorescent structures are interpreted in the context of compartments along the endocytic pathway. The present study now shows that different styryl dyes are able to label three distinct populations of cytoplasmic structures in intact guard cells. These structures can be distinguished on the basis of their diameter and their affinity to the three dyes tested. An analysis of the fluorescent images reveals that a population of small cortical structures, but not all labelled structures inside a cell, can be interpreted in the context of endocytosis.

Small FM-labelled structures are endocytic vesicles

The expected small size of endocytic vesicles (Holstein, 2002) raises the question whether these small structures can be investigated with a confocal microscope. The resolution of an optical microscope is defined as the shortest distance between two points on a specimen that can still be distinguished as separate entities. However, objects with subresolution dimensions like single vesicles can still cause a recordable signal, if only the signal-to-noise ratio is high enough. The lower end of possible size measurements with the confocal technique is marked by the dimension of the PSF, which, in our setup, had a lateral dimension of around 275 nm. Thus, all fluorescent structures measured through the cuticle, which are bright enough to be detected and which are smaller than the PSF, will appear to have a lateral diameter of 275 nm.

The green (FM1-43 and FM2-10, λem = 580 nm) and red (FM4-64, λem = 650 nm) emitting styryl dyes resulted in labelling of a population of small vesicles with a mean diameter of approximately 290 and 310 nm, respectively. This difference can be explained by the fact that resolution depends on the emission wavelength (Jonkman and Stelzer, 2002):


Calculating the theoretical difference with equation 1 and the emission maxima of the dyes yields to ΔFWHMxy = 19 nm, which is close to the observed difference of approximately 20 nm. As the green styryl dyes have emission spectra, which are more similar to that of the beads, their values for the lateral resolution limit are closer to the values determined for the PSF. From these observations that all the vesicles have diameters that exactly match the dimensions of the PSF, we conclude that all vesicles have, as expected, subresolution diameters.

The small fluorescent structures, which are found most prominently in the cortical regions of cells, can for the following reasons be identified as endocytic vesicles. First, they co-localise with structures carrying KAT1::GFP proteins. KAT1 is a PM K+-channel, and the activity of the KAT1::GFP chimaera can be measured with electrophysiological methods in the PM of transfected guard cell protoplasts. Furthermore, it has been shown that this chimaera can be retrieved from the PM via endocytosis (Hurst et al., 2004). Hence, it is reasonable to assume that structures, which are co-labelled with KAT1::GFP and FM4-64, are endocytic vesicles, which have retrieved the K+-channel protein from the PM via endocytosis. Second, previous capacitance recordings have resolved endocytosis of single vesicles in guard cell protoplasts (Homann and Thiel, 1999). The authors note that vesicles with a subresolution diameter contribute to the recorded signal but cannot be resolved as single endocytic steps and do not contribute to the observed size distribution. Hence, it is only reasonable to compare the size distributions of both methods in their well-resolving range (e.g. above around 300 nm). But even though, the high similarity implies that both methods measure the same endocytic vesicles.

Previous experiments with V. faba guard cell protoplasts have collected a body of evidence for a dynamic nature of the PM. It was found that an excursion in the size of the PM surface area, which occurs during swelling and shrinking, is accomplished by an exo- and endocytic incorporation and retrieval of membrane into and from the PM, respectively (Homann and Thiel, 1999). Together with the vesicular membrane, also K+-channel proteins are delivered or retrieved via exo- and endocytosis, respectively (Homann and Thiel, 2002). The present data now show that this system can be extrapolated to turgid guard cells. The fact that endocytic vesicles and even more so endocytic vesicles carrying the GFP-tagged K+-channel KAT1 were identified in turgid cells shows that endocytosis is possible against the high turgor pressure of intact guard cells. As all intact guard cells were investigated under constant osmotic conditions, controlled endovesiculation rather than vesiculation of the PM upon an osmotic shock led to the observed endocytic vesicles (Hawes et al., 1995). Notably these (double labelled) vesicles have only been observed in the cortical cytoplasm. This might point to a reserve or recycling pool of vesicles carrying K+-channels, which may be retrieved from and inserted into the PM during the closing and opening cycles of guard cells (Hurst et al., 2004).

The failure to show endocytosis against high turgor in intact guard cells in the past was most likely caused by a signal problem rather than by an energetically restriction to the process itself: (i) the (constitutive) endocytic rate in these cells is low; and (ii) the size of endocytic vesicles is expected to be very small, as this reduces the energy demand of endocytosis in presence of a high turgor (Saxton and Breidenbach, 1988). Either contributes to a very low signal.

We assume that endocytosis against high turgor is clathrin mediated with no special mechanism contributing to this process. In the mammalian and yeast fields there is growing evidence that actin plays an important role in endocytic events. Numerous proteins have been identified that functionally link actin with the endocytic machinery and at the same time regulate actin filament dynamics. These proteins could thus harness forces produced during actin polymerisation to facilitate steps in the endocytic process (Engqvist-Goldstein and Drubin, 2003).

It will now be of great interest whether plants make use of such forces to promote endocytosis against high turgor.

Intermediate structures are compartments of the endocytic pathway

The size distribution histograms of all fluorescent markers reveal a clear peak at approximately 470 nm. The fact that this peak occurs with all FM dyes suggests that it reflects a real structure, which is discernable from the small endocytic vesicles. Circumstantial evidence suggests that they are pre-vacuolar or endosomal-like compartments. We will further refer to them as endosomes.

An argument in support of this hypothesis is the finding that FM4-64 reveals the highest relative number of these intermediate-sized structures. This dye is also the only among the FM dyes, which after long incubation stains the tonoplast. In addition, structures around that size can be shown distant from the PM, e.g. in cytoplasmic strands. Together, this suggests that the intermediate-sized structures are compartments in the endocytic pathway from the PM to the tonoplast. This is in agreement with the finding that FM4-64 co-localises with the Arabidopsis-ras-related GTPases ARA6 and ARA7 in early endosomes in Arabidopsis thaliana protoplasts which are members of the Rab/Ypt family (Ueda et al., 2001).

Notably, structures of the very intermediate size are also detected in cells treated with the Mitotracker. As this dye also gives a well-detectable signal in vacuoles, it is reasonable to assume that this membrane-permeable dye can also label endosomes. This is in agreement with the finding that the size distribution histogram of guard cell mitochondria, measured in electron micrographs, lacks a peak at approximately 470 nm.

FM dyes also label mitochondria

It is a significant observation that, in particular, the FM1-43 and FM2-10 also label large cytoplasmic structures with a mean diameter of approximately 600/730 nm. For two reasons, we believe that these structures are mitochondria. First, the structures have the expected size of guard cell mitochondria, according to size distribution histograms from confocal laser scanning microscopy (CLSM) images of guard cells labelled with the Mitotracker Red CMX Ros and measurements from electron micrographs. In addition, both histograms confirm the double peak, which is caused by the elongated shape of mitochondria. Second, they are co-labelled with Mitotracker CMX Ros, a specific fluorescent label for mitochondria. FM1-43 and FM4-64 have also been reported to stain mitochondria of Neurospora crassa hyphae, even though the latter only after prolonged incubation (Fischer-Parton et al., 2000). Collectively, this indicates that the large FM-labelled structures in guard cells are mitochondria.

Styryl dyes where former developed as membrane potential sensors (Fluhler et al., 1985; Grinvald et al., 1988) and other styryl dyes with a very similar structure are used as both Mitotrackers and markers for endocytosis at the same time (i.e. 4-Di-1-ASP, Haugland, 2002). In this context, we have also shown that a pre-incubation of cells with sodium azide abolished the appearance of cytoplasmic label with FM and Mitotracker dyes in guard and epidermal cells. The absence of cytoplasmic labelling in the presence of metabolic inhibitors has generally been used as an indication for the energy requirement of endocytosis. However, as styryl dyes are voltage sensitive, they only label the mitochondria when their membrane is charged. Hence, the absence of label in azide-treated cells is most likely because of the inability of the dyes to label the depolarised mitochondria. Differential labelling in the presence versus absence of a metabolic inhibitor is, thus, not necessarily a proof for the existence of energy requiring endocytosis.

FM dyes can penetrate a phospholipid bilayer

The Mitotracker used in this study does not enter cells via an endocytic pathway but is cell permeant. It is thought to passively diffuse across the PM and accumulates in active mitochondria (Haugland, 2002). Accordingly, staining of endocytic vesicles was never observed with this dye. We can therefore conclude that this Mitotracker is directly entering the cytosol from where it stains three compartments: mitochondria, endosomes and vacuoles. As the mitochondrial staining can be effectively prevented by sodium azide, the membrane potential of these organelles must play a crucial role in order to be labelled by this dye.

The same behaviour was found for FM2-10. As the existence of a vesicular pathway between the PM and mitochondria is unlikely to exist, this dye must also find its way to mitochondria guided by their membrane potential and therefore must enter the cytosol. Thus, if these dyes can enter the cytosol, they must be able to cross a membrane. This may happen via any of the below-mentioned mechanisms.

It has been reported that mechanosensitive channels of Chinese hamster ovary (CHO) cells can transport FM1-43 across the PM, which in turn led to FM-labelled mitochondria (Gale et al., 2001; Nishikawa and Sasaki, 1996). In view of the fact that mechanosensitive channels have been described in guard cells of V. faba (Cosgrove and Hedrich, 1991) and that we found FM-labelled mitochondria, these dyes may enter the cytosol by this pathway in guard cells as well. Uptake via a mechanosensitive channel has not been observed for bigger dyes such as FM3-25 (Meyers et al., 2003). As FM2-10 in turn is smaller than FM1-43, it might enter the cytosol more easily via this pathway, which would explain the more intense and rapid staining of mitochondria.

Uptake of dyes may also result from local damage of labelled PM and endocytosed membrane. Illumination of fluorescent dyes inevitably produces hyperoxide radicals by an energy transfer from their triplet state to molecular oxygen (Betz et al., 1992), which may directly, or via lipid peroxidation, lead to membrane damage. This effect is used by a technique called photochemical internalisation to deliver membrane-impermeable macromolecules into the cytosol (Selboa et al., 2002). The unspecific uptake of FM dyes via photochemical internalisation is in agreement with the observation that guard cells more often became leaky at the PM, in correlation to illumination intensity (e.g. 3D stack recordings), duration of incubation, concentration and quantum yield of the dye. This is reflected in the observed toxicity of the dyes and may render some of these dyes almost useless for long-term investigations of membrane trafficking, at least in guard cells.

Differences in hydrophobicity may cause differential labelling

It is apparent that FM4-64 passes through the entire endocytic pathway up to the vacuole but does not stain mitochondria. This is consistent with previous investigations showing that FM4-64 is staining the tonoplast of various cell types from a broad origin such as pollen tubes of L. longiflorum (Parton et al., 2001), protoplasts of A. thaliana (Ueda et al., 2001), Saccharomyces cerevisiae (Vida and Emr, 1995), numerous fungal hyphae (Fischer-Parton et al., 2000) and Dictyostelium (Heuser et al., 1993).

However, according to the present data, it appears that all FM dyes also stain early structures in the endocytic pathway such as endocytic vesicles and presumably endosomes. Hence, the fact that FM2-10 does not label the tonoplast implies either that this dye escapes at early stages from the endocytic pathway into the cytosol (from where it can label mitochondria) or that it recycles back to the PM, while FM4-64 remains in the pathway to the vacuole. As the three dyes mainly differ in their hydrophobicity, it can be reasoned that this parameter is a major cause for their differential behaviour.

Investigations by Mukherjee et al. (1999) have shown, for CHO cells, that lipid analogues with long acyl chains are sorted to late endosomes (lytic pathway), whereas analogues with short acyl chains are delivered to the endocytic recycling compartment (ERC) and are targeted back to the PM (recycling pathway). The same may apply for the differential sorting of FM2-10 and FM4-64 in guard cells and would explain why FM4-64 but not FM2-10 is targeted to the latest possible endosome in plants – the vacuole. Such a sorting into different endocytic pathways could be realised by the partly coated reticulum (PCR). However, for the time being, this assumption is solely based on morphological similarity between the PCR and mammalian ERC or sorting endosomes (Fowke et al., 1991; Tanchak et al., 1988). Moreover, it is known from the animal system that the membranous components get sorted to late endosomes, whereas aqueous solutes are recycled via the ERC back to the PM (Mukherjee et al., 1999). The solubility of FM dyes in the lipid environment is favoured by longer tails and more double bonds (Betz et al., 1996). In accordance, Hao and Maxfield (2000) have shown that FM2-10 behaves more like an aqueous solute in recycling through the ERC. As FM4-64 has a longer and more unsaturated tail than FM2-10 has, it is reasonable to assume that its hydrophobicity is sufficiently high to be sorted away from FM2-10 into a different pathway leading to the vacuole. This pathway, in turn, is likely to lead via multi-vesicular bodies (MVB), which might be functional related to late endosomes (Fowke et al., 1991; Tanchak and Fowke, 1987). Thus, the chemical difference of the dyes causes a different affinity to membranes and, as a consequence, the dyes may be sorted into different endocytic pathways.

Lack of appropriate markers for endosomal compartments in plant cells (especially those for live cell investigations) and the similar sizes of the involved organelles in electron microscopic (EM) studies (Low and Chandra, 1994) leave us with the speculation that any of the described plant endosomal compartments (i.e. PCR and MVB) may have caused the signals corresponding to the 470-nm-sized structures. Considering that sorting of different FM dyes does occur, it may be speculated that those labelled by FM2-10 are mainly PCR whereas those labelled by FM4-64 are MVB.


The present data show that not all fluorescent structures labelled with FM dyes can be interpreted in the context of endocytic activity. Unspecific uptake of dye and/or early escape from endocytic compartments may deliver dye to the cytosol and lead to a staining of compartments such as the mitochondria, not involved in endocytosis. FM4-64, the dye with the highest hydrophobicity, turns out to be the most suitable tool for monitoring endocytosis in guard cells. This dye reveals endocytic vesicles in turgid guard cells, which are abundant in the cortical area in the vicinity of the PM. Some of these vesicles are also involved in the retrieval of the K+-channel KAT1 from the PM.

Experimental procedures

Plant material

Guard cells were investigated on abaxial epidermal strips from about 3-week-old leaves of V. faba L. cv. Bunyan. The plants were grown at 20°C, 60% RH, 300 µE light for 14 h and 17°C, 70% RH for 10 h dark. Epidermal strips were anchored in small dishes with the cuticle facing a coverslip and incubated and investigated in a buffer solution consisting of 10 mm 2-Morpholinoethanesulfonic acid (MES, pH 6.1/KOH), 20 mm KCl and 100 µm CaCl. All investigations were carried out in stomata with constant stomatal aperture. Cells were checked for cytoplasmic streaming as a viability test before, during and after each investigation.

Fluorescent dyes

FM1-43, FM2-10, FM4-64 and the Mitotracker Red CMX Ros were purchased from Molecular Probes, Leiden, the Netherlands. Concentration of the styryl dyes was adjusted to ensure a similar brightness after 1-min incubation: 5 µm FM1-43, 20 µm FM2-10 and 10 µm FM4-64. The Mitotracker Red CMX Ros was applied in a final concentrations of 500 nm.

Construction of the KAT1::GFP fusion protein and transfection of intact guard cells via particle delivery were carried out according to Hurst et al. (2004).

Confocal laser scanning microscopy

Confocal microscopic analysis was carried out using a Leica TCS SP (Leica Microsystems GmbH, Heidelberg, Germany), equipped with a 63× water immersion objective (plan apo, N.A. 1.2). Excitation levels were adjusted to cause minimal autofluorescent signal. All styryl dyes were exited using the 488-nm line of a 25 mW Ar/Kr ion laser. Emission for FM1-43 and FM2-10 was detected at 530–630 nm and for FM4-64 at 600–700 nm. Chlorophyll fluorescence was allowed to be detected in the FM4-64 channel in order to maximise detection efficiency while minimising excitation energy. Co-localisation of FM2-10 and Mitotracker Red CMX Ros was performed by a sequential scan: FM2-10 was excited at 488 nm, but emission was detected at 600–560 nm. The Mitotracker Red CMX Ros was excited at 568 nm, while emission was recorded between 600 and 650 nm to avoid chlorophyll fluorescence in the images. With this approach, both fluorophores can be detected separately without noticeable cross-talk. The signals have not been accumulated in order to get reliable co-localisation during sequential scans in presence of cytoplasmic streaming. Each guard cell pair was imaged as a stack of paradermal images at 500-nm spacing using the Leica Confocal Software 2.00 (LCS, Leica Microsystems GmbH). imagej (http://rsb.info.nih.gov/ij) was used for further processing of the images.

Size measurements

From each guard cell, sizes of about 10 structures were measured in order to give each cell the same statistical weight. The intensity values (z) of eight-bit-image matrices (x,y) of single structures were fitted with a Gaussian function of the form


in order to obtain the parameters wx and wy for each structure. These parameters are proportional to the FWHM peak height of the Gaussian curve in x and y directions, respectively, which in turn corresponds to the diameter of the imaged structure.


Separate parameters for the diameter of the structures in x and y directions (i.e. wx and wy) were chosen in order to identify if the structures are ideal spheres ore not.

Fit of size distributions

The size distributions were fitted with a multiple lognormal function to determine the mean value of the peaks, as it takes the asymmetrical shape in our size distributions into account.



We thank Dr Ulrich Kubitscheck for his continual help. This work was supported by the Deutsche Forschungsgemeinschaft.