Aged mother cells of Saccharomyces cerevisiae show markers of oxidative stress and apoptosis



Recently, we and others have shown that genetic and environmental changes that increase the load of yeast cells with reactive oxygen species (ROS) lead to a shortening of the life span of yeast mother cells. Deletions of yeast genes coding for the superoxide dismutases or the catalases, as well as changes in atmospheric oxygen concentration, considerably shortened the life span. The presence of the physiological antioxidant glutathione, on the other hand, increased the life span of yeast cells. Taken together, these results pointed to a role for oxygen in the yeast ageing process. Here, we show by staining with dihydrorhodamine that old yeast mother cells isolated by elutriation, but not young cells, contain ROS that are localized in the mitochondria. A relatively large proportion of the old mother cells shows phenotypic markers of yeast apoptosis, i.e. TUNEL (TdT-mediated dUTP nick end labelling) and annexin V staining. Although it has been shown previously that apoptosis in yeast can be induced by a cdc48 allele, by expressing pro-apoptotic human cDNAs or by stressing the cells with hydrogen peroxide, we are now showing a physiological role for apoptosis in unstressed but aged wild-type yeast mother cells.


Yeast mother cell-specific ageing has been intensively researched and reviewed (Jazwinski, 1999; Johnson et al., 1999; Guarente, 2000) in the last few years. This simple model system for cellular and, perhaps, organismic ageing shows similarity to ageing in higher cells in terms of well-characterized morphological and physiological changes. Old cells are much bigger than young cells (Mortimer and Johnston, 1959; Egilmez et al., 1990; Nestelbacher et al., 1999), the cell cycle as well as protein synthesis is slowed down (Mortimer and Johnston, 1959; Motizuki and Tsurugi, 1992), and the cell surface has a loose and wrinkled appearance (Pichova et al., 1997). The median life span of most laboratory Saccharomyces cerevisiae strains is about 25–35 generations or about 3 days (Jazwinski, 1993). Given the short life span of yeast cells and the ease of genetic analysis in yeast, it is feasible to elucidate general mechanisms of ageing by studying the yeast ageing process.

In recent years, Sinclair and Guarente (1997) proposed the hypothesis that DNA minicircles derived from chromosomal rDNA but lacking centromeres accumulate in mother cells and eventually prevent ordered replication and cell cycle progression by titrating out an essential DNA-binding factor, thus leading to senescence. These minicircles have a strong tendency to accumulate in the mother cell (Sinclair and Guarente, 1997; Heo et al., 1999) because they cannot be segregated regularly on the mitotic spindle. However, so far, rDNA minicircles have not been demonstrated in senescent cells of multicellular organisms. Many different hypotheses have been put forward to explain cellular and organismic ageing (Finch, 1990) including, for example, that ageing may be caused by the accumulation of somatic mutations in senescent cells. These theories are, of course, not mutually exclusive. Moreover, almost certainly, several independent causes can contribute to the ageing process. As we have argued previously, an accumulation of mutations in genomic DNA can be ruled out as a cause of yeast ageing because of its mother cell specificity (Nestelbacher et al., 2000). On the other hand, it is quite possible that damaged cellular material other than DNA could be decisive for yeast ageing provided that this damaged material is inherited asymmetrically.

Cellular materials can be damaged by oxygen radicals or oxidizing molecules originating from a ‘leaky’ respiratory chain when those radicals are no longer detoxified by the cell (Halliwell and Gutteridge, 1989). This ‘oxygen theory of ageing’, first formulated by Harman (1962), has gained increased acceptance based on recent experimental results (Ames et al., 1993; Orr and Sohal, 1994; Sohal and Weindruch, 1996; Gems, 1999). In particular, it has been found that longevity genes of the nematode Caenorhabditis elegans, whose biochemical functions were previously unknown, are now shown to play a role in the detoxification of oxygen radicals (Gems, 1999; Taub et al., 1999).

Recently, we and others have shown that genetic and environmental changes that increase the burden of reactive oxygen species (ROS) on cells of the yeast S. cerevisiae lead to a shortening of the life span of mother cells. Deletions of yeast genes coding for the superoxide dismutases (Barker et al., 1999; Wawryn et al., 1999) or the catalases, as well as changes in atmospheric oxygen partial pressure and the presence of the physiological antioxidant glutathione, all had the expected distinct effects on life span and indicated a role for oxygen in the yeast ageing process (Nestelbacher et al., 2000). We therefore sought to show in a more direct way that old yeast mother cells accumulate oxidizing molecules and build up internal oxidative stress in the absence of any external stress condition.

Here, we present biochemical evidence that old mother cells, but not virgin cells, contain strongly oxidizing molecules originating from the mitochondria. On standard media, old cells suffer oxidative stress in the absence of any external source of oxygen radicals. These cells show markers of apoptosis, including positive terminal deoxynucleotidyl transferase-mediated dUTP end labelling (TUNEL) and annexin V tests and chromatin fragmentation seen after 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI) staining. Apoptosis has been shown to occur in certain yeast mutants and in wild-type cells when they are oxidatively stressed (Madeo et al., 1997; 1999; Frohlich and Madeo, 2000). The present work describes yeast apoptosis in ageing wild-type cells. This is the first report of a physiological role for apoptosis in yeast. Thus, we present another strong parallel between yeast ageing and the ageing processes of cultured human cells that have recently been shown to undergo apoptosis induced by oxidative stress (P. Jansen-Dürr, personal communication).


Isolation of large yeast cells by elutriation allows enrichment of very old cells

The haploid wild-type yeast strain JC482 (Pichova et al., 1997) was grown and treated as described in Experimental procedures for the isolation of large and small cells by elutriation centrifugation. The large cells obtained after 1 day (about 10 generations) of growth in 100 ml of YPD were cultivated again for 1 day in 100 ml of YPD, and large cells were again isolated by elutriation. Fraction V contained large undamaged cells, whereas fraction II contained very uniform small cells. The total yield of large cells was about 1.6 × 108 cells. These two fractions were characterized in detail.

Fraction V cells were oval with a major axis diameter of 10–15 µm, and fraction II cells were rounder with a diameter of about 5–7 µm, as determined by phase-contrast microscopy. The median life span of fraction V cells was 4 ± 3.5 generations, whereas the median life span of fraction II cells was 18 ± 3.8 generations, similar to the standard life span for this strain of 23 ± 6 generations (Fig. 1). The small difference between the life span of fraction II and the standard life span of the strain is statistically significant. Note that about 70% of the fraction II cells were virgins, and the rest showed one or two bud scars, which would account for the difference in life span. About 35% of the large cells (fraction V) entered but did not complete a single cell division, although they were not lysed or damaged as shown by fluorescence microscopy (see below, Fig. 6) and thin-section electron microscopy (data not shown). Some of the cells in fraction V were actually young, as shown by the ‘tail’ of long-lived cells in the life span analysis of fraction V cells (Fig. 1). These cells are the few daughters that did not separate in the last cell cycle but were still adhering to their mothers.

Figure 1.

Life spans of elutriated wild-type yeast cell fractions of strain JC482 were determined as described in Experimental procedures. The error bars indicate standard deviations from the median. Forty randomly chosen cells from fraction V and fraction II were analysed. In fraction V, only one of 40 cells never budded. In fraction II, six out of 40 cells never budded and were excluded from the calculation. The standard life span of strain JC482 was determined as described previously (Pichova et al., 1997).

Figure 6.

A. Fraction V cells were stained for exposed phosphatidylserine with FITC-conjugated annexin V after digestion of cell walls with glusulase/lyticase and viewed under a fluorescence microscope using the fluorescein filter set (Madeo et al., 1997).

B. The same sample washed and stained with propidium iodide and viewed using the fluorescein filter set. This control shows that the annexin-positive cells in (A) are not lysed or damaged.

C. Fraction II cells stained for phosphatidylserine with FITC-conjugated annexin V and viewed under a fluorescence microscope using the fluorescein filter set. A sample with a very infrequently observed annexin-positive cell (arrow) is shown.

D. The same sample stained with propidium iodide showing that the marked cell is lysed (arrow).

E. The same sample as in (C) and (D) shown in phase contrast. Also here, it is obvious that the marked cell is lysed (arrow).

Figure 2 shows representative views of the fraction V and fraction II cells after Calcofluor White staining. Some of the fraction V cells showed a large number of bud scars (Fig. 2A). Note that only half of each cell's surface can be seen in the micrographs, so the real number of bud scars is probably about twice as great as the number of bud scars visible in Fig. 2A. However, it is also clear from Fig. 2A that some of the large cells have only about 5–10 bud scars. We hypothesize that these are already senescent, as the standard life span of the strain shows that a certain fraction of mothers is senescent after every cell generation. It is also obvious from Fig. 2A that some of the old cells carry unseparated daughters, which themselves have several bud scars. This is an indication of cell cycle irregularities during the last cell divisions of a mother cell, as can be seen by comparison with the micrographs shown in our previous papers (Pichova et al., 1997; Nestelbacher et al., 2000). The cells in fraction II were mostly daughter cells showing a dark birth scar but no bud scars, although fraction II also contained some cells with one or two bud scars (Fig. 2B).

Figure 2.

A. Fraction V cells after staining with Calcofluor White M2R and viewing under a fluorescence microscope. The budding pattern on mother cells is typical of a haploid cell. Irregular patterns are seen on some contaminating daughters (see text).

B. Fraction II cells treated exactly as in (A). The majority of these cells are virgins with just one (dark) birth scar. The cells are smaller than in (A).

Old yeast mother cells contain ROS in the absence of external oxidative stress

Figure 3 shows that strongly oxidizing molecules (ROS) were detected by dihydrorhodamine 123 (DHR) staining in the mitochondria of fraction V cells (Fig. 3A) but not in fraction II cells (Fig. 3C). For comparison, the staining with 2-[4-(dimethylamino)styryl]-1-methylpyridinium iodide (DASPMI) indicates the location of mitochondria in these cells (Fig. 3E and G). Double staining with DHR and DASPMI was not possible for technical reasons. However, the morphology shown in Fig. 3 is clearly indicative of a mitochondrial location for the strongly oxidizing molecules. Mitochondria of young cells do not contain measurable levels for strongly oxidizing molecules (Fig. 3C and G). As these cells have been grown in YPD under standard conditions and have never been exposed to external oxidative stress apart from that found in a normal aerobically grown cell, we must assume that the strongly oxidizing molecules that are detected in Fig. 3A arise internally, either through accumulation by a ‘leaky’ respiratory chain or triggered by the physiological process of ageing.

Figure 3.

A. Fraction V cells were stained with DHR (5 µg ml−1; stock solution 2.5 mg ml−1 in ethanol) and viewed and photographed under a confocal laser-scanning fluorescence microscope after 10 min using the rhodamine filter set. The stained cells show typical mitochondrial morphology.

B. The same sample in phase contrast.

C. Fraction II cells treated as above show only very weak staining.

D. The same sample as in (C) viewed in phase contrast.

E. Fraction V cells stained with the mitochondrial-specific dye, DASPMI (25 µg ml−1), and immediately viewed under a confocal laser-scanning fluorescence microscope using the fluorescein filter set.

F. The same sample shown in phase contrast.

G. Fraction II cells stained with DASPMI and viewed under a fluorescence microscope.

H. The same cells as in (G) shown in phase contrast.

Taken together, Fig. 3 shows that old cells (fraction V), but not young cells, exhibit strongly oxidizing molecules and that these molecules are contained in the mitochondria.

Old yeast mother cells display markers of apoptosis

Figure 4 shows the nuclear morphology of fraction V cells (Fig. 4A) and fraction II cells (Fig. 4B). The young cells display compact single nuclei of normal appearance, whereas the old cells showed diffuse staining indicating chromatin fragmentation similar to that seen when apoptosis is induced by hydrogen peroxide in yeast cells (see Fig. 1 in Madeo et al., 1999). Some of these old cells showed more than one nucleus as a result of endomitosis caused by irregular cell division cycles as shown earlier (Pichova et al., 1997).

Figure 4.

A. Fraction V cells after staining of nuclei with DAPI. Note diffuse chromatin and, occasionally, multiple nuclei.

B. Fraction II cells treated exactly as above. Note compact well-defined nuclei.

Figures 5 and 6 illustrate our analysis of apoptotic landmarks in fraction V cells with young cells (fraction II) included as important controls to exclude methodological artifacts. Positively stained cells were detected using DHR, TUNEL and annexin V tests in about 20% of the cells for each method. It is impossible for technical reasons to perform double staining by these methods; however, the complete absence of positively staining cells in fraction II strongly indicates that it is the same subfraction of old cells that stain positively with all three fluorescent dyes.

Figure 5.

A. Fraction V cells were fixed, cell walls were digested and strand breaks in DNA were detected according to the TUNEL protocol (Madeo et al., 1997). Nuclei containing large amounts of DNA strand breaks were stained black by the diaminobenzidine–H2O2 reaction after incorporation of fluorescein isothiocyanate (FITC)-labelled dUTP and treatment with anti-FITC antibody Fab fragment from sheep coupled with horseradish peroxidase. Viewing 500 cells, positive staining was observed in about 20% of them. In some cases (arrows), mother and daughter cells from a pair were both TUNEL positive, indicating that the last daughters of old apoptotic mother cells are sometimes also apoptotic.

B. Fraction II cells were treated and stained as in (A). Practically no TUNEL-positive staining was observed.

Figure 5A and B shows that old cells and not young cells are stained by the TUNEL method that detects DNA strand breaks. Many of the TUNEL-positive cells carry one or more unseparated daughters that are also TUNEL positive (arrow), as if those cells could not complete the last or the last few cell cycles, thus corroborating the finding that about 35% of the fraction V cells start but do not complete one cell cycle (Fig. 1).

Figure 6 illustrates the results of fluorescent staining with the annexin V method. The old cells of fraction V were frequently stained at the cell periphery after digestion of the cell wall (Fig. 6A), indicating plasma membrane inversion (externalized phosphatidylserine, which is an established marker for early apoptotic events). The same cells showed no staining with propidium iodide (the usual control), indicating that these cells are not unspecifically damaged (Fig. 6B). On the contrary, fraction II cells were not stained by this method (Fig. 6C). Very infrequently, young cells were observed that were stained, but they were also stained by propidium iodide, indicating that they had lysed as a result of unspecific damage. The damaged cell is indicated by an arrow in Fig. 6C–E.


Here, we present a morphological and physiological characterization of aged yeast mother cells that were isolated by a new method. These cells are very close to the terminal senescent state, as shown by determination of their remaining median life span. Our data presented here, together with earlier work (Nestelbacher et al., 2000) showing that the physiological antioxidant, reduced glutathione, can substantially increase the life span of yeast cells under conditions of oxidative stress, strongly support the notion that ROS originating in the mitochondria are a causative factor in senescence. The present data also show that senescent yeast mother cells undergo apoptosis.

The method developed by us for the isolation of old yeast mother cells depends on only one parameter (cell size) and is limited by the fact that it is impossible completely to avoid contamination by some young cells (virgins) that adhere to their mother cells (Fig. 1). In spite of this limitation, the life span determinations performed with the size-fractionated old and young cells showed that the old cells have a remaining median life span of only 4 ± 3.5 generations in a strain whose standard median life span is 23 ± 6 generations. The young cells isolated in the same experiment have a median life span that closely resembles the standard life span of the strain, thereby excluding the possibility that the elutriation method itself has any influence on the life span of the cells or is grossly altering their physiology. About 35% of the old cells (fraction V) do not complete the first cell cycle that they enter. We assume that these are cells undergoing senescence. These cells are not lysed or damaged by the elutriation method, as shown by light microscopy, fluorescence microscopy with propidium iodide and electron microscopy of ultrathin sections (data not shown). A similar fraction of the old cell preparation (about 20%) shows the presence of strongly oxidizing molecules in the mitochondria and markers of yeast apoptosis. Fraction V contains many cells with more than 20 bud scars (Fig. 2A). This fraction also contains a small proportion of about 10% of younger cells, as seen by the ‘tail’ of longer lived cells in the life span determination (Fig. 1). These are daughters that were still attached to their mothers during the elutriation as well as an unavoidable subfraction of young cells with a larger volume. Taken together, these experimental results strongly support the conclusion that we have indeed isolated a population of cells that is very close to senescence and that the population is sufficiently enriched in senescent cells to enable a biochemical study of the phenomenon of senescence, which is crucial for understanding the ageing process. In order to confirm that these markers of oxidative stress and apoptosis are not simply caused by the intensity of respiration, we also stained cells of strain JC482 growing exponentially on glycerol as well as on glucose as the only carbon source. Neither mother nor daughter cells stained positively with DHR under exactly the same conditions that were used for staining fraction V (not shown in detail). This is in good agreement with earlier observations that non-fermentable carbon sources (glycerol) certainly do not shorten the life span (Egilmez et al., 1990; Barker et al., 1999).

In order to show that the phenotype of old mother cells described here does not depend on the genetic background of the strain, it was important to repeat the elutriation experiment with a second unrelated haploid MATα strain (BY4742). It was also seen for this strain that old mother cells, but not young cells, exhibited ROS in their mitochondria, diffuse or fragmented nuclei and many of those cells showed > 20 bud scars (data not shown in detail).

Other methods for isolating old yeast mother cells have been published (Egilmez et al., 1990; Woldringh et al., 1995; Smeal et al., 1996). Using the method presented here, the residual life span of the mother cells is very short (only about 10% of the total life span of the strain), and a substantial fraction of the old cells are post-mitotic and senescent. Old mother cells and young cells are isolated from the same initial population of batch-grown cells (see Experimental procedures) that are only minimally influenced physiologically by the separation method. These cells are clearly not killed by an artifact of the method.

We have used these cells to search for phenotypic markers of oxidative stress and apoptosis and, in addition, to characterize them with respect to cell surface and nuclear morphology. The results have provided strong support for the theory that oxidative stress is implicated in the cell ageing process. How do these findings fit into the general hypotheses that are presently in vogue to explain ageing in both yeast and higher organisms? Could the oxygen theory of ageing for which we have provided evidence here and the theory that is based on the formation of rDNA minicircles as a primary cause of ageing be part of the same general picture?

In principle, one possible, but unlikely, explanation is that the perturbation of replication caused by the accumulation of rDNA circles leading to a titration out of essential protein factors (Johnson et al., 1999) could indirectly lead to a perturbation of gene expression and therefore to an increase in the load of cells with ROS. This is improbable, because the correlation of ROS with ageing is a very general one, as it has also been demonstrated in cells of higher eukaryotes, whereas rDNA minicircles have been found only in ageing yeast cells, not in ageing cells of humans, flies or nematodes. Another, more likely explanation is that the autogenous oxidative stress observed by us indirectly leads to the accumulation of rDNA minicircles. This could be the case, for instance via an effect of ROS on recombination and repair. Intrachromosomal recombination, the process that leads to rDNA minicircles, has been shown to be induced by oxidative mutagens (Brennan et al., 1994).

The two processes could even be interdependent: oxidative stress could cause a moderate increase in rDNA minicircles that could lead in a vicious circle to more oxidative stress and more minicircles in the mother cell. Finally, the two processes could be independent of each other and both contribute to yeast ageing.

The morphology of the structures stained with DHR clearly indicates a mitochondrial location for the ROS. This fits excellently with a large number of investigations of oxidative ageing in higher cells and points to a parallel between the ageing processes of yeast and higher cells. It has been shown that, in higher cells, ROS (oxidative stress) are key regulators of apoptosis (Hockenbery et al., 1993; Kane et al., 1993; Ghibelli et al., 1995; Greenlund et al., 1995) and that primary human cells in culture undergo oxidative stress and apoptosis when they enter senescence (Jansen-Dürr, personal communication). Our observation of about 20% TUNEL-positive and annexin V-positive cells in the preparation of old, but not young, yeast cells (Figs 5 and 6) therefore further strengthens the oxygen theory of ageing and the generality and similarity of yeast and higher cell ageing finally leading to apoptosis. It is important that apoptosis has now been found in wild-type cells as a physiological process in the absence of external oxidative stress. The existence of apoptosis in yeast is now well established (for a review, see Ink et al., 1997; Frohlich and Madeo, 2000), but it has only been described previously in special mutants (cdc48-S565G) or in wild-type cells incubated in the presence of H2O2.

A final point concerns the origin of mother cell specificity of ageing in the context of the findings presented here. Mother cell specificity of yeast ageing implies that some ‘death factor’ (Jazwinski, 1990) must be inherited asymmetrically by the mother cell. In the framework of the hypothesis presented here, this death factor could be damaged cellular material. One very plausible candidate for damaged material would be damaged mitochondria, which produce an increasing amount of ROS. Mitochondria producing ROS are found only in senescent cells, not in virgin cells (Fig. 3). Consequently, we speculate that old (pre-existing and damaged) mitochondria should be inherited preferentially by the mother cell in any cell division, whereas newly synthesized mitochondria should be segregated to the daughter cell. The ordered inheritance of mitochondria in yeast during cell division has been well studied (Warren and Wickner, 1996; Yaffe, 1999; Catlett and Weisman, 2000), although the question whether the mother cell inherits old mitochondrial components is presently unanswered.

Experimental procedures

Yeast strains and media

A wild-type yeast strain (JC 482; Pichova et al., 1997) was used for all experiments. Some of the experiments were repeated using a second strain, BY4742 (Brachmann et al., 1998). Liquid YPD (2% glucose, 1% yeast extract, 2% peptone in H2O) was used for growing yeast cells. Life spans were determined on SC agar (2% agar, 2% glucose, 0.17% yeast nitrogen base, 0.5% ammonium sulphate). Media components were from Gibco.


Cells were separated according to their diameter using the Beckman elutriation system and rotor JE-6B with a standard elutriation chamber. Before the separation, the cells were grown in 100 ml of YPD medium at 28°C on a rotary shaker for 24 h. Then, the cells were harvested at 3000 r.p.m. and resuspended in 1× PBS buffer (8 g of NaCl, 0.2 g of KCl, 1.44 g of Na2HPO4, 0.24 g of KH2PO4, pH 7.4, in a total volume of 1 l) at 4°C. The elutriation chamber was loaded with 4.2 ml of cell suspension corresponding to about 109 cells. To separate cell fractions with different diameters, the chamber was loaded at a flow rate of 10 ml min−1 and a rotor speed of 3200 r.p.m. Cells with a diameter < 5 µm were elutriated (fraction I). To collect fraction II (diameter 5–7 µm), the flow rate was set to 15 ml min−1 and rotor speed to 2700 r.p.m. Fraction III (diameter 7–8.5 µm) was elutriated at 2400 r.p.m., fraction IV (diameter 8.5–10 µm) at 2000 r.p.m. and, finally, fraction V (diameter 10–15 µm) at 1350 r.p.m. The quality of separation of particular fractions was verified microscopically.

Calcofluor staining

Calcofluor staining of cell walls was carried out as described previously (Streiblova et al., 1984), using Calcofluor White M2R (Sigma F-3397).

DAPI staining

The basic protocol for DAPI staining (Streiblova, 1988) of nuclei was used. Cells were collected, resuspended in 70% (v/v) ethanol for brief fixation and permeabilization, stained with DAPI solution and observed under an epifluorescence microscope (Zeiss Axioskop).

DHR and DASPMI staining

Free intracellular radicals or strongly oxidizing molecules (ROS) were detected with DHR (Sigma D1054) as described previously (Madeo et al., 1999). We noticed that it was necessary to take pictures after no longer than 10 s laser irradiation to avoid photodamage. Excitation at 488 nm was used. Cells were viewed through a rhodamine optical filter. Mitochondria were visualized with DASPMI (2-[4-(dimethylamino)styryl]-1-methylpyridinium iodide; Sigma-Aldrich 280135) using the fluorescein filter set described previously (McConnell et al., 1990).

The samples were viewed using a Leica TCS 4D confocal laser-scanning microscope, and pictures were taken using a Hamamatsu video system.

Annexin V and TUNEL

Exposed phosphatidylserine was detected by annexin V labelling (Madeo et al., 1997). DNA strand breaks were demonstrated by TUNEL (Madeo et al., 1997) using the Boehringer Mannheim in situ cell death detection kit (POD) and observed with a Zeiss Axioskop microscope.

Life span procedure

Life span analysis was performed as described previously (Pichova et al., 1997).

To determine the remaining life span of fraction II and fraction V cells, cohorts of 40 randomly chosen cells per fraction were taken directly after elutriation and, for each cell, the number of remaining cell cycles was determined by micromanipulation. Cells that never budded were excluded from analysis.

Statistical analysis of life span data

The standard deviations of the median life spans at a confidence level of 95% were calculated by applying Kaplan–Meier statistics (Kaplan and Meier, 1958). To decide whether two given survival distributions were significantly different at a 95% confidence level, Breslow, Tarone–Ware and log-rank statistics were used. All statistical calculations were performed using the software package spss 9.0.


This work has been supported by grant no. P14574-MOB (FWF Austria; to M.B.), an ARC IREX grant to I.W.D. and M.B., and grant GA CR(CZ) GA204/97/0541 to A.P. We are grateful to the Faculty of Sciences of Salzburg University for making available the Axioskop fluorescence microscope.


  1. These authors contributed equally to this work.