• Open Access

Plasticity of death rates in stationary phase in Saccharomyces cerevisiae


  • Nadège Minois,

    1. Survival and Longevity Laboratory, Max Planck Institute for Demographic Research, Rostock, Germany
    Search for more papers by this author
    • *

      Present address: Research Institute of Molecular Pathology, Institute of Molecular Biotechnology, Vienna, Austria.

  • Francesco Lagona,

    1. Statistics Laboratory, Max Planck Institute for Demographic Research, Rostock, Germany
    Search for more papers by this author
    • Present address: Dipartimento Instituzioni Pubbliche, Economia e Società, University Roma Tre, Rome, Italy

  • Magdalena Frajnt,

    1. Survival and Longevity Laboratory, Max Planck Institute for Demographic Research, Rostock, Germany
    Search for more papers by this author
  • James W. Vaupel

    1. Survival and Longevity Laboratory, Max Planck Institute for Demographic Research, Rostock, Germany
    Search for more papers by this author

Nadège Minois, Vienna Drosophila RNAi Center, Dr Bohr-Gasse 3, A-1030 Vienna, Austria. Tel.: +43 1790444547; fax: +43 17987153; e-mail: minois@imp.univie.ac.at


For the species that have been most carefully studied, mortality rises with age and then plateaus or declines at advanced ages, except for yeast. Remarkably, mortality for yeast can rise, fall and rise again. In the present study we investigated (i) if this complicated shape could be modulated by environmental conditions by measuring mortality with different food media and temperature; (ii) if it is triggered by biological heterogeneity by measuring mortality in stationary phase in populations fractionated into subpopulations of young, virgin cells, and replicatively older, non-virgin cells. We also discussed the results of a staining method to measure viability instead of measuring the number of cells able to exit stationary phase and form a colony. We showed that different shapes of age-specific death rates were observed and that their appearance depended on the environmental conditions. Furthermore, biological heterogeneity explained the shapes of mortality with homogeneous populations of young, virgin cells exhibiting a simple shape of mortality in conditions under which more heterogeneous populations of older cells or unfractionated populations displayed complicated death rates. Finally, the staining method suggested that cells lost the capacity to exit stationary phase and to divide long before they died in stationary phase. These results explain a phenomenon that was puzzling because it appeared to reflect a radical departure from mortality patterns observed for other species.


The budding yeast Saccharomyces cerevisiae is a model widely used in aging research since it has been shown that individual cells have a limited potential of division (Mortimer & Johnston, 1959). This model organism shed light over the years on important genetic mechanisms of lifespan regulation, the most famous example probably being the sirtuins, with the discovery that the SIR genes modulate lifespan (Kaeberlein et al., 1999).

Two types of lifespan are studied in yeast. The replicative lifespan measures how many times individual cells divide (e.g. Kennedy et al., 1994). The vast majority of aging research in yeast is done using this method. So far, about 40 genes and several environmental factors have been shown to increase replicative lifespan (Jazwinski, 2001, 2002 and references therein; Kaeberlein et al., 2005). The chronological lifespan, in contrast, measures how long non-dividing cells live. This non-dividing phase of population growth is reached when cells have exhausted the available nutrients. In this measurement, a small subsample of known volume of the original populations is spread onto solid medium, and cells able to divide and to form a colony are counted. This approach is called the colony-forming unit (CFU) method.

Conditions in which cells are kept during the experiments vary between laboratories and lead to very different chronological lifespans (Fabrizio & Longo, 2003). If grown in a medium containing a limited amount of nutrients (SC medium for instance), cells will enter into the post-diauxic phase, keep a high metabolic activity and die within a few days if kept in their spent medium (Fabrizio et al., 2001, 2004). If grown in a nutrient-rich medium (YPD medium for instance), cells enter a hypometabolic, stress-resistant state called stationary phase (Werner-Washburne et al., 1996; Gray et al., 2004). When transferred into distilled water after reaching post-diauxic phase, the chronological lifespan extends to several months (Gendron et al., 2003). However different the conditions can be between these experiments and between the lifespans they lead to, it seems that analogous pathways and mechanisms regulate both survival in post-diauxic phase and survival in stationary phase in water. So far, mutants isolated for long-life in post-diauxic phase also lived longer in stationary phase in water (Fabrizio & Longo, 2003). A recent screen of a collection of open reading frames deletions of non-essential genes further uncovered 90 genes whose knockdown increased chronological lifespan (Powers et al., 2006).

Yeast is not used only in studies of genetic mechanisms of longevity. Its characteristics make it also a very suitable model organism for biodemographic studies of mortality: large population size in stationary phase and relatively short chronological lifespan. Previous studies have shown that age-specific death rates calculated from the CFU method in stationary phase exhibited a complicated pattern: they rose, fell, and rose again late in life (Vaupel et al., 1998; Gendron et al., 2003). This shape is at variance with the one observed for mortality in other carefully studied species, which shows a rise then a plateau or decline late in life (Vaupel et al., 1998). In this paper, any shape of mortality differing from a rise with age followed or not by a decline or plateau in late life is considered complicated.

A hypothesis put forward by Gendron et al. (2003) was that such a complicated shape was triggered by heterogeneity. However, they showed that the mortality shape was not due to genetic heterogeneity caused by mutations during population expansion to stationary phase. In this article, we investigate whether other factors can modulate mortality shape obtained from the CFU method. It is known that medium containing fewer nutrients than the standard one can increase replicative lifespan (Lin et al., 2000). We tested whether mortality differs with the medium given to the cells during the CFU assay. Although genetic heterogeneity has been ruled out, other types of heterogeneity could explain the complicated shape of mortality in yeast. It is known that populations of yeast cells in stationary phase are composed of individuals in different physiological states, depending upon their reproductive history before entering into stationary phase: small, young, having never budded cells, and bigger, older, having budded cells. To test whether such heterogeneity could trigger the shape of age-specific death rates, we fractionated the cells in the populations according to their size – small, young cells and big, replicatively older cells – and follow the mortality of each subpopulation separately. Finally, we investigated if environmental conditions can modulate mortality shape.

Another hypothesis to explain that mortality shape in yeast is different from the one observed in other organisms is that a different definition of death is used for yeast. Both replicative and chronological lifespans assume that a cell unable to divide is dead. However, if this assumption is not verified, then the measurement of mortality by the CFU method is not comparable with mortality measurement in other organisms. It seems to be the case. For instance, Cho & Kim (1999) showed that survival in Salmonella typhi assessed by CFU declined much faster than survival assessed by a green fluorescent protein (GFP)-based direct viable count of the cells in the population. A study of the thermal resistance in S. cerevisiae showed that after a prolonged exposure to 52 °C, cells were not able to divide and form colonies. In contrast, many of them were still able to exclude the dye propidium iodide, suggesting that their cellular integrity was still intact (Attfield et al., 2001). Maskell et al. (2003) studied chronological lifespan in lager strains of S. cerevisiae and showed that the decline of cells able to divide in the populations (CFU) was much faster than the changes in viability observed with the age of the cultures as assessed with two different dyes (methylene violet and oxonol). The evidence gathered from the literature and the observation that a lot of cells were still present in the populations in stationary phase when none could form a colony when provided with food prompted us to measure chronological lifespan with a method not requiring cells to divide and to investigate whether cells in stationary phase failing to form colonies when provided with food were dead while in stationary phase or died as a result of exposure to nutrients.


In this paper, we aimed at understanding the factors triggering the complicated shape of age-specific death rates in the budding yeast in stationary phase. To analyze the shape of mortality, age-specific death rates were computed by estimating the mortality hazard function from the data. Relevant changes of direction in mortality were studied by detecting significant zero-crossings of the hazard derivative, where mortality risk reaches either a relative maximum or a relative minimum level. The occurrence of consecutive zero-crossings of the hazard derivative, in particular, indicates complicated mortality patterns, as those previously reported by Vaupel et al. (1998) and Gendron et al. (2003). Appropriate estimation of the hazard derivative allows us to test whether these zero-crossings are significant or due to sampling error (see Appendix S1 and Fig. S1 in the Supporting Information).

Effect of medium on CFU counting

In a first experiment, we grew four populations of S288C cells to stationary phase in YPD. In contrast with previously published results, age-specific death rates in these populations increased with age (Fig. 1). The CFU results were not dependent on the medium used for the assay. We conducted the CFU assay on five media with decreasing concentrations of yeast, peptone and glucose as well as on SC medium and found no difference in the number of cells estimated in the populations from the CFU counts for the 78 days of the experiment (P-values for the medium effect ranging from 0.62 to 0.97 for the four populations, see Fig. 2 for an example). During this time the cell number in the populations dropped by around 98% and if an effect of the medium were to be seen, it would have been observed (age effect: P < 0.0001 for all the populations).

Figure 1.

Age-specific death rates in standard conditions. Age-specific mortality of four replicate populations of genotype S288C grown on YPD and kept in sterile distilled water at 30 °C after having reached stationary phase.

Figure 2.

Effect of medium on estimated number of cells in the populations. This figure shows results for population of genotype S288C replicate 1.

The only noticeable difference was that after 3 days of incubation at 30 °C, the colonies formed on standard YPD were bigger than on any other medium (data not shown). Given the difference in growth, we kept the plates longer to see if any new colonies would appear on the less rich media after a longer incubation time. We did not observe so.

To compare further the effect of medium, we grew populations of S288C cells either on YPD or SC medium to stationary phase, the latter condition replicating the conditions used in previously published studies of mortality (Vaupel et al., 1998; Gendron et al., 2003). Again, S228C cells grown on YPD showed an increase in mortality. In contrast, cells grown on SC medium showed first a decrease in age-specific death rates, then an increase (Fig. 3). This is at variance with published experiments done in the same conditions (Vaupel et al., 1998; Gendron et al., 2003). The initial mortality observed in the present experiments was much higher than usually reported. For instance, in Fig. 3, the initial values range from 0.05 to 0.32 for the populations grown on SC medium but are not different from 0 for the populations grown on YPD. The initial high mortality rates of cells grown on SC medium mean that a lot of cells lost the ability to divide very early in the experiment. We did not see the initial increase in mortality observed in previous studies (Vaupel et al., 1998; Gendron et al., 2003) because the cells grown on SC medium that would die over a week or so in previous experiments died on the first day in the present experiment.

Figure 3.

Age-specific death rates on different culture media. Age-specific mortality of three populations of genotype S288C grown on YPD and three populations of the same genotype grown on SC medium. Cells were kept in sterile distilled water at 30 °C after having reached stationary phase.

Effect of temperature and genotype on CFU counting

To complete our investigation of the effect of environmental conditions on mortality shape, we decided to keep cells in stationary phase at 34 °C instead of 30 °C. We also included different genotypes to test for strain-specific effect. As seen on Fig. 4, S288C cells kept under these conditions exhibited a complicated shape of mortality: the age-specific death rates increased, decreased then increased again, as originally observed (Vaupel et al., 1998; Gendron et al., 2003). The strain carrying a deletion of the SIR2 gene (Y13738) showed the same pattern. In contrast, its control strain (BY4742) had increasing death rates.

Figure 4.

Age-specific death rates at high temperature. Age-specific mortality of the three studied genotypes (S288C, BY4742 and a SIR2Δ in the BY4742 background). The populations were grown on YPD and kept in sterile distilled water at 34 °C after having reached stationary phase.

Effect of fractionation on CFU counting

Complicated shapes of mortality are triggered by heterogeneity. Gendron et al. (2003) had previously shown that mortality shape was not due to genetic heterogeneity caused by mutations arising during population expansion to stationary phase. Our results show that mortality shape is not dictated by strain-specific heterogeneity, since the same strain can exhibit different shapes of age-specific death rates: S288C cells have an increasing mortality when grown on YPD and kept at 30 °C (Figs 1 and 3) but show a complicated mortality pattern when kept at 34 °C (Fig. 4).

If the heterogeneity observed on mortality in yeast is not genetic, then it is probably biological. Allen et al. (2006) reported that yeast populations in stationary phase consisted of two different subpopulations – quiescent, unbudded, and nonquiescent, replicatively older cells – and that nonquiescent cells lost their viability earlier than quiescent cells.

To test the contribution of this type of heterogeneity to mortality shape, we fractionated populations grown to stationary phase on SC medium into the small cells that never budded before entering stationary phase (‘young’ populations) and the bigger cells that divided before entering into stationary phase (‘old’ populations) and followed the mortality of each subpopulation. Figure 5 shows the results of one of the experiments. The mixed, control population exhibited a shape of age-specific death rates similar to the one observed in our previous experiment for cells grown on SC medium: a high initial mortality followed by a decrease then an increase in mortality. Cells from the ‘old’ population showed a similar pattern. In contrast, the population consisting of young, unbudded cells exhibited a low, flat mortality then an increase from around 30 days of age.

Figure 5.

Age-specific death rates of fractionated populations. Age-specific mortality of populations of S288C cells grown on SC medium and kept in sterile distilled water at 30 °C, fractionated after entry into stationary phase according to the size of the cells in the populations. Control: whole population. Young: small, virgin cells. Old: big cells having divided before entering into stationary phase.

These results suggest that the two subpopulations behave differently mortality-wise and that the population of young, unbudded cells appear less heterogeneous than the populations of replicatively older cells and of control cells. A second replicate gave similar, but not as clear results (Supporting Fig. S2). It thus seems that the complicated shape of mortality in yeast can at least be partly explained by the biological heterogeneity of the cells in the populations and that heterogeneity is observed depending on the environmental conditions experienced by the populations.

Viability measurement by staining

Although the level of population heterogeneity and the effect of the environment seem to explain how complicated mortality patterns can occur in yeast populations in stationary phase, another hypothesis for the difference observed is that the CFU method is not comparable with lifespan measurements in other organisms. The CFU method will give an incorrect estimation of mortality if cells in stationary phase are still alive but unable to divide.

To test this hypothesis, we measure viability using the dye Phloxine B (see Supporting Information), which stains the metabolically inactive cells (Minois et al., 2005). As previously reported (Maskell et al., 2003; Allen et al., 2006), we found that the number of living cells estimated by the two methods was similar in young cultures in stationary phase and that this number decreased faster with the age of the cultures when counted by the CFU method (data not shown). When the number of viable cells as a function of age was estimated using another dye, rhodamine 123, similar results as with Phloxine B were observed, ruling out a dye-specific effect (data not shown). Finally, the death rates calculated from the Phloxine B staining exhibited complicated patterns (data not shown), showing that it is an inherent feature of yeast populations and not an artifact of the CFU method.

Since the discrepancy between the CFU and the staining methods seems to come from an inability of the older cells to form colonies, the reaction of cells to nutrient exposure might explain the difference. Cells in stationary phase sense the nutrients given to them and decide to implement the processes leading to exit from stationary phase and division. However, cells are less and less able with time to successfully exit stationary phase and die trying to do so. To test this hypothesis, we plated on YPD plates in which Phloxine B was added individual cells from young (3 days) and old (70 days) cultures in stationary phase to assess the percentage of cells able to give a colony. All the cells plated were alive (unstained) at the beginning of the experiment. In young cultures, 74% of the cells divided and gave a colony. In contrast, in old cultures, although a few cells (16%) began growing their first daughter, they died in the process and did not give a colony. The percentages of cells giving a colony are in accordance with the percentages of living cells observed at the same ages with the CFU method, albeit lower (74% vs. 98% for young cultures and 0% vs. 1% in old cultures) probably because the plating of individual cells required more handling of the cells. The cells that did not divide began dying a few hours after plating (4% in young cultures and 10% in old ones were dead after 3 h). All the cells that did not divide were dead after 28 h in young cultures and 31 h in old cultures. In contrast, if the cells are kept in water (during the staining method), they exhibit 98% and 56% of survival for 3-day- and 70-day-old populations, respectively. This experiment shows that in old cultures, cells survive better if kept in water than if plated on YPD. It thus seems that cells die as they are not able any longer to progress through the cell cycle when nutritional cues for division are present. From the number of living cells in stationary phase, fewer and fewer cells are able to progress through the first division on rich medium with the age of the cultures. However, further tests would be required to determine the significance of these results.


We have confirmed in this paper that yeast cells in stationary phase can really exhibit complicated mortality shapes. This is at complete variance with the traditional theories of the evolution of senescence, which predicts an exponential increase of mortality with age. It is not the first time that a departure from a logarithmic increase with age of death rates is reported. For instance, Abrams (1991) showed that mortality in historical populations followed a linear increase. He also cited examples showing the same pattern in wild animals. However, these studies were based on limited sample sizes.

In large populations, changes in mortality have also been described. Mortality levels off or even declines at late ages in several species, including humans (Vaupel et al., 1998). Environmental conditions also modulate death rates in Drosophila melanogaster. Temperature changes the slope of the increase in mortality and dietary restriction does not alter the slope, but delays aging-related mortality (Mair et al., 2003). However, complicated patterns of mortality have so far been reported only in S. cerevisiae, maybe because mortality trajectories have been studied particularly intensively in this organism.

Importantly, and for the first time, we showed here that the shape of death rates in yeast depended on the environment. When grown to stationary phase on YPD, S288C cells had a simple mortality when kept at 30 °C but exhibited a complicated shape of death rates at 34 °C. Mortality shape was also influenced by the medium on which the cells had been grown to stationary phase. It was already reported that medium affects chronological lifespan in yeast. For instance, MacLean et al. (2001) showed that cells grown on glycerol were able to divide and give colonies longer than cells grown on YPD. However, they compared a fermentable (glucose) carbon source with a nonfermentable (glycerol) one and they looked only at the percentage of cells able to divide. They did not calculate the age-specific death rates. Growth on SC medium, a medium less nutritious than YPD led to a shorter chronological lifespan than growth on YPD (Weinberger et al., 2007). Recently, Boer et al. (2008) reported that starvation of yeast populations by auxotrophic requirements triggered shorter chronological lifespans than starvation for phosphate or sulfate.

In contrast with the effect on mortality of the medium used for growth to stationary phase, we showed here for the first time that the medium used for exit from stationary phase (during the CFU assay) did not modulate mortality.

Previous results had shown that the age-specific death rates in yeast populations in stationary phase rose, fell and rose again (Vaupel et al., 1998; Gendron et al., 2003). A strain-specific mechanism for this observation was discarded because yeast strains of different genotypes exhibited the same shape of mortality.

An alternative hypothesis to explain such a pattern is the heterogeneity of the individuals in the populations (Vaupel & Yashin, 1985; Vaupel et al., 1998). Heterogeneity is an inherent feature of populations, even when they consist of genetically identical individuals kept in the same conditions. Rea et al. (2005) observed that isogenic Caenorhabditis elegans worms differed in their ability to activate a heat shock protein (hsp16.2) promoter (as measured by the expression of GFP fused to the hsp16.2 promoter) during a sublethal heat shock and that this ability predicted their remaining life expectancy. The nematodes with a high hsp16.2 promoter activity lived much longer than the worms with a low hsp16.2 promoter activity. The longevity difference between nematodes with low and high hsp16.2 promoter activity was more pronounced when a longer, more stressful heat shock was used to induce the activity of the promoter, showing that heterogeneity was greater under sub-optimal conditions.

We observed here the same phenomenon in death rates in yeast cells in stationary phase. When cells were grown and kept in optimal conditions (YPD and 30 °C) we saw a regular increase of mortality: heterogeneity of the cells in the populations did not translate into the death rates shape. In contrast, when cells were kept in less optimal conditions (grown on SC medium or kept at 34 °C), the mortality differential between the subpopulations increased and the mortality shape then showed waves of high or increasing mortality separated by a decrease in mortality: mortality of the subpopulations with different heterogeneity can be separated. Finally, when conditions become harsh for all the cells, they all die quickly and a simple shape of death rates can be observed again. However, what constitutes a benign or harsh environment will differ from strain to strain depending on their intrinsic strength. This may explain why at 34 °C, the BY4742 strain exhibited a simple increase in mortality and not a complicated pattern as observed in the other genotypes in the same conditions. The BY4742 strain was short-lived compared with the others, meaning that this strain is likely weaker and the temperature used was comparatively harder to withstand for this strain than the others. Sir2p is a histone deacetylase. It reduces gene expression, genomic instability and silences the mating-type genes (Hekimi & Guarente, 2003) and sir2Δ mutants have a lower replicative lifespan than the controls (Kaeberlein et al., 1999). In contrast, we found here that a deletion in the SIR2 gene increased survival compared with the control strain. Similar results have been published by Fabrizio et al. (2005). We also showed here that not only the lifespan of the sir2Δ mutants was increased but that the mutation conferred an overall decrease in frailty of the cells in stationary phase with cells exhibiting mortality patterns similar to the longer-lived S288C strain and not similar to their control BY4742.

We decided to test for the effect of heterogeneity on mortality shape in yeast. The biological causes for heterogeneity in yeast cells have only been recently investigated. Gendron et al. (2003) showed that the complicated mortality shape was not due to genetic heterogeneity caused by mutations during population expansion to stationary phase. In contrast, nongenetic, biological heterogeneity has recently been observed in yeast populations in stationary phase. Allen et al. (2006) showed that S. cerevisiae populations in stationary phase were composed of two subpopulations: quiescent and nonquiescent cells. Compared with quiescent cells, nonquiescent ones lost more rapidly their ability to divide and form colonies in the presence of food, were less resistant to heat and showed signs of oxidative damage and apoptosis. Weinberger et al. (2007) showed that yeast cells more efficiently able to arrest in G1 phase when entering into stationary phase exhibited longer chronological lifespans than cells arrested in S phase. The study of gene expression in quiescent and nonquiescent cells (Aragon et al., 2008) showed that nonquiescent cells seemed unable to arrest growth. In contrast, quiescent cells exhibited gene expression showing they had the capacity to survive prolonged starvation and to respond quickly if conditions became favorable to growth. This suggests that biological heterogeneity is still somehow triggered by genetic heterogeneity. This is also supported by the fact that Aragon et al. (2008) identified mutations affecting reproductive capacity and/or viability.

Another source of heterogeneity in cells in stationary phase is their reproductive history. A population at the beginning of stationary phase consists for a half of cells that never budded before entering stationary phase, the other half containing cells having budded one and more times. Apart from a possible effect of budding per se on death rates, older cells (that have budded) have accumulated damage that they retain and do not pass to their progeny (Aguilaniu et al., 2003) and the level of damage within the cell contributes to heterogeneity between the cells. We tested the effect of reproductive heterogeneity on mortality by fractionating populations into virgin (young) and nonvirgin (old) cells and following mortality of each subpopulation. We showed here that this type of heterogeneity explains at least partly the death rates observed in yeast: populations of young, virgin cells displayed increased mortality rates with age, whereas the control population and the population consisting of older cells, more heterogeneous, exhibited a more complicated shape of death rates.

However, heterogeneity is inherent to every population, so why should it trigger such a pattern only in the budding yeast? Compared with all the other species for which death rates have been thoroughly studied, the budding yeast is the simplest and the only unicellular one. As discussed by Vaupel (2003), this may mean that populations of yeast are much more heterogeneous than populations of more complex species. This, however, is a tentative hypothesis that requires further research.

Another hypothesis to explain the difference of death rates shapes in yeast and other organisms is that the CFU method may not accurately measure lifespan in stationary phase. This hypothesis is driven by the observation that when the CFU assay is stopped because none of the cells are able to exit stationary phase and divide, there remain many cells in the populations. We decided to evaluate mortality in yeast cells in stationary phase with a method not requiring the cell to be able to divide. We chose to use a dye, phloxine B, that stains dead, metabolically inactive cells. Using this staining method, we found that the estimated number of living cells in the populations decreased with time, albeit more slowly than with the CFU method.

By using simultaneously the CFU and the staining methods, our results suggest that cells in stationary phase, while still alive, lose their ability to exit stationary phase and divide again with time spent in stationary phase. It is unclear why after some time in stationary phase cells become unable to divide. Entering into and exiting from stationary phase are complex processes (Gray et al., 2004). Cells in stationary phase undergo specific changes that allow them to withstand this life without nutrients, the most striking of them being the absence of division. Among other changes, we can list a higher stress resistance (Piper, 1993), thickened cell wall, reduced transcription and translation rates (Fuge et al., 1994), inhibited mRNA degradation (Jona et al., 2000) and induced autophagy (Noda & Ohsumi, 1998). Differences observed in gene expression are probably responsible for these changes (Gray et al., 2004). Yeast cells possess a set of plasma membrane sensors of extracellular nutrients that allows them to evaluate the amount of available nutrients (Forsberg & Ljungdahl, 2001). Entry into stationary phase is triggered by nutrient limitation. Once in stationary phase, cells will stay in this phase until they can detect key nutrients (especially carbon sources) that will trigger the implementation of at least the first steps towards exit from stationary phase (Gray et al., 2004). However, the success of the exit from stationary phase will depend on the nutrients available (Granot & Snyder, 1991, 1993). The reaction of cells to nutrients might thus explain the difference observed between the CFU method and the staining method. Cells in stationary phase sense the nutrients given to them and decide to implement the processes leading to division, but as our results suggest, old cells die while trying to implement these processes.

Two other studies have reported that cells lost their ability to divide before they died while in stationary phase (Maskell et al., 2003; Allen et al., 2006). In the first paper, the authors reported long-term survival in stationary phase measured with the CFU method and with staining with methylene violet (reduced by active enzymes) and oxonol (membrane integrity). They followed survival in stationary phase of polyploid lager production strains with the three methods every 5 days for 50 days. They showed that with time, cells in stationary phase lose first their reproductive ability (fast decline observed with the CFU method), then their membrane integrity (slower decline with oxonol) and finally their intracellular reducing capacity of methylene violet (almost no decline over the 50 days of the experiment). They observed this phenomenon in the four different strains studied and in both virgin and non-virgin cells. Allen et al. (2006) measured survival with the CFU method and with the FUN1 dye in quiescent and nonquiescent cells in stationary phase at 7, 14, 21 and 28 days. They showed that the percentage of living cells decreased fast when measured by the CFU method with nonquiescent cells exhibiting already only 36% survival at 7 days. In contrast, when measured with the FUN1 dye, the authors showed that the number of living quiescent cells did not significantly decrease and that the survival of nonquiescent cells was much higher than when measured with the CFU method and decreased later (21 and 28 days). For the quiescent cells the results reported by Allen et al. (2006) are similar to the ones reported here with both methods giving similar viability in young cultures and differing in older cultures. In contrast, the authors reported a difference already in young cultures of nonquiescent cells.

To conclude, our study brought two results that cannot be explained by our current understanding of theories of evolution of senescence.

First, we showed that yeast cells in stationary phase can exhibit depending on the environment various shapes of mortality, including very complicated ones. This experimental demonstration of the plasticity of age-specific mortality is one of the most important results about mortality trajectories since the discovery of the leveling-off of death rates at old ages observed in many species (Vaupel et al., 1998). As a matter of fact, the shapes we described in this paper, as well as the leveling-off of death rates at old ages, do not support the traditional theories of the evolution of senescence, that predict an increase of mortality with age.

Second, we observed that cells lose the capacity to exit stationary phase and to divide long before they die in stationary phase. Thus, it seems that cells in stationary phase exhibit a post-reproductive life. It is hard to see what evolutionary mechanism could lead to this. In order to explain this phenomenon we need to know why cells in stationary phase do not die soon after losing the ability to reproduce. One hypothesis might be that the cells do not assess their ability to exit stationary phase before trying to re-enter the cell cycle, that is before they find in their environment the nutritional clues triggering the exit from stationary phase.

Experimental procedures


Yeast haploid strains used in the experiments were the following: S288C (MATα mal gal2), BY4742 (MATα; his3Δ1; leu2Δ0; lys2Δ0; ura3Δ0) and Y13738 (BY4742; YDL042 c::kanMX4) that carries a deletion of the SIR2 gene in the BY4742 background.

Preparation and maintenance of the populations

Experimental cohorts were obtained from single cells as follows. A sample of frozen culture was streaked onto standard YPD (2% glucose, 2% peptone, 1% yeast extract, 2% agar) or SC medium (Longo et al., 1997) plates. After 3 days of incubation at 30 °C, a single colony was sampled and cells were transferred to 250 mL glass flasks with 40 mL liquid (without agar) YPD or SC medium, where they were grown (30 °C and 200 r.p.m.) until stationary phase was reached. At this point, each cohort was transferred to 50 mL Falcon tubes, and cells were collected by centrifugation. Cells were then washed using autoclaved Millipore water following the procedure: (i) 8 min centrifugation at 5200 g; (ii) removal of the supernatant; and (iii) resuspension with 45 mL of fresh water. This procedure was repeated three times. After the last wash, the cohort was resuspended in 45 mL fresh water and placed at the desired temperature and 200 r.p.m. in 250 mL glass flasks. The cells were resuspended in the maximum amount of water (45 mL in 50 mL tubes) to ensure a thorough washing. This point marked the first day of the survival assay. To preclude reproliferation of cells in the cultures during the experiments, the cells were washed following the same procedure as above every other day to ensure fresh water, which was void of nutrients. Under these experimental conditions, the cells remain in stationary phase and undergo no budding (Fabrizio & Longo, 2003).

Lifespan measurement: CFU method

Daily measures of CFUs in each experimental cohort were estimated as follows. One hundred microliter from each cohort was removed from the flask containing the experimental population and serially diluted (from 105 to 10 depending on the number of cells in the population). Four hundred and fifty microliter from each cohort was removed from the flask containing the experimental population when the surviving population was so small that no dilution was used. The dilution was changed to a lower one when the number of colonies counted was too small and could increase sampling error. Five days a week, 40 µL of the appropriate dilution was plated onto a half YPD plate for a total of eight (dilutions 105 to 10) or five (no dilution) plates. After incubation at 30 °C for 3 days, the number of colonies on each half-plate was counted and used to estimate the total number of cells able to divide in the flask (see Statistical analyses in the Supporting Information).

Effect of medium on CFU counting

Four populations of the strain S288C were grown to stationary phase on YPD, transferred into distilled water and kept at 30 °C. Survival was estimated by the CFU method. Between days 21 and 99 after entry into stationary phase, cells were plated on different media and the number of colonies compared. The different media used were standard YPD, YPD with only 0.5% glucose, YPD with only 0.5% glucose and 0.5% peptone, YPD with only 0.5% glucose, 0.5% peptone and 0.25% yeast extract and SC medium. The number of cells able to divide again when given food in each population was estimated from the number of colonies counted on the different types of medium and compared with Kruskal–Wallis tests. The analysis was performed separately for each population.

We also tested whether the medium used to grow the cells to stationary phase influenced the shape of mortality. Three populations of the strain S288C were grown on YPD and three populations on SC medium to stationary phase, transferred into sterile distilled water and kept at 30 °C. This experiment allows a direct comparison between the previously published experiments (Vaupel et al., 1998; Gendron et al., 2003), in which cells were grown to stationary phase on SC medium, and the experiment reported in the paragraph above.

Effect of temperature and genotype on CFU counting

In this experiment, four populations (one of the S288C strain, one of the BY4742 strain and two of the Y13738 strain) were grown to stationary phase in standard YPD at 30 °C, transferred into sterile distilled water and kept thereafter at 34 °C.

Effect of fractionation on CFU counting

In this experiment, populations of S288C cells were grown to stationary phase on SC medium. For each of two independent experiments, three populations of S228C cells in stationary phase were fractionated in order to separate virgin cells from cells having dividing before entering stationary phase. This could be done because yeast cells get bigger with the number of divisions (Egilmez et al., 1990).

Sucrose gradients were prepared by slowly pouring into 50 mL falconer tubes 15 mL of a 50% sucrose solution and then 15 mL of a 10% sucrose solution. Tubes were gently put horizontally in a fridge for 11 one 1 h. After that time, tubes were put back vertically and on the top of each gradient, 600 µL of cells was slowly layered. The tubes were centrifuged for 2 min at 2100 g. After centrifugation, two layers of cells could be observed, an upper layer consisting of the cells that divided before entering stationary phase (old cells) and a lower layer with the small virgin cells (young cells). Each category of cells was retrieved from the gradient with a gradient fractionator (Auto Densi-Flow, Ref. 4517200, Labconco, Kansas City, MO, USA), transfer into a 15-mL plastic tube, washed three times with distilled water and then kept in water. Eight gradients were prepared at a time and the procedure was continued until all the cells in the three populations were separated. For the control population, cells from one of the populations were mixed again together after fractionation. All the young cells separated from the other two populations were pooled and kept in 45 mL of sterile distilled water. The same was done for the old cells. The efficiency of the fractionation was checked regularly during the fractionation procedure by inspection under the microscope of the different populations.


We thank A. Storek-Langbein, N. Sievers, B. Pfeiffer, A. Neumann and B. Klissing for their technical assistance and M. Breitenbach (University of Salzburg) for the strains BY4742 and Y13738.