1. Biochemistry and Chemical Biology
  2. Structural Biology and Molecular Biophysics
Download icon

A physicochemical perspective of aging from single-cell analysis of pH, macromolecular and organellar crowding in yeast

  1. Sara N Mouton
  2. David J Thaller
  3. Matthew M Crane
  4. Irina L Rempel
  5. Owen T Terpstra
  6. Anton Steen
  7. Matt Kaeberlein
  8. C Patrick Lusk
  9. Arnold J Boersma  Is a corresponding author
  10. Liesbeth M Veenhoff  Is a corresponding author
  1. European Research Institute for the Biology of Ageing, University of Groningen, University Medical Center Groningen, Netherlands
  2. Department of Cell Biology, Yale School of Medicine, United States
  3. Department of Pathology, School of Medicine, University of Washington, United States
  4. DWI-Leibniz Institute for Interactive Materials, Germany
Tools and Resources
  • Cited 0
  • Views 1,015
  • Annotations
Cite this article as: eLife 2020;9:e54707 doi: 10.7554/eLife.54707

Abstract

Cellular aging is a multifactorial process that is characterized by a decline in homeostatic capacity, best described at the molecular level. Physicochemical properties such as pH and macromolecular crowding are essential to all molecular processes in cells and require maintenance. Whether a drift in physicochemical properties contributes to the overall decline of homeostasis in aging is not known. Here, we show that the cytosol of yeast cells acidifies modestly in early aging and sharply after senescence. Using a macromolecular crowding sensor optimized for long-term FRET measurements, we show that crowding is rather stable and that the stability of crowding is a stronger predictor for lifespan than the absolute crowding levels. Additionally, in aged cells, we observe drastic changes in organellar volume, leading to crowding on the micrometer scale, which we term organellar crowding. Our measurements provide an initial framework of physicochemical parameters of replicatively aged yeast cells.

Introduction

Cellular aging is a process of progressive decline in homeostatic capacity (Gems and Partridge, 2013; Kirkwood, 2005). Key molecules have been identified to govern the aging process and can extend health and lifespan by maintaining prolonged homeostasis (Kenyon, 2010). The function of these molecules ultimately depends on physicochemical properties, such as pH, macromolecular crowding, and ionic strength. As all these physicochemical properties require maintenance, the question how stable these properties are in aging is pertinent. Identifying age-related changes in the cytosol, where all proteins are synthesized and most metabolic processes take place is crucial to understanding how aged cells function differently from their younger counterparts. Saccharomyces cerevisiae is an excellent model system to quantify physicochemical changes during aging, as single cells can be directly monitored by microscopy as they age (Crane et al., 2014; Jo et al., 2015). Importantly, many of the molecular mechanisms that contribute to yeast aging are conserved in humans (Janssens and Veenhoff, 2016a).

pH homeostasis is an important parameter in human aging, as human senescent cells show increased lysosomal pH (Kurz et al., 2000), and in age-related pathologies such as Alzheimer’s and Parkinson’s disease, lysosomes are dysfunctional (Carmona-Gutierrez et al., 2016). The main proton pumps in the lysosomal membrane (termed vacuole in yeast), the V-ATPases, are highly conserved from yeast to human, and Pma1 - the yeast plasma membrane ATPase, shares structural and functional similarities with the Na+K+ ATPases in mammalian cells (Forgac, 2007; Morth et al., 2011; Nelson et al., 2000). Pma1 localizes in the plasma membrane and transports cytosolic protons out of the cell (Ferreira et al., 2001; Orij et al., 2011; Serrano et al., 1986), while the V-ATPase pumps protons from the cytosol into the lumen of various organelles and regulates their pH (Forgac, 2007; Kane, 2006). Both enzymes change in aging: Pma1 levels increase as this protein is asymmetrically retained in the mother cell (Henderson et al., 2014) and the components of the V-ATPase become substoichiometric (Janssens et al., 2015), potentially reducing the number of functional complexes. Concomitantly, changes in vacuolar and cytosolic pH have been reported in aging, namely, an alkalinization of the cortex (region close to the plasma membrane) (Henderson et al., 2014), and alkalinization of the vacuole (Chen et al., 2020; Hughes and Gottschling, 2012), both measured in single cells and occurring early in the lifespan. In addition, in a population-based study, an acidification of the cytosol at the end of the replicative lifespan was reported (Knieß and Mayer, 2016). So, while there is evidence for changes in pH in cellular aging, what is currently missing is a single-cell perspective on cytosolic pH in yeast replicative ageing.

Human senescent cells and aged yeast cells increase in size, which might result in dilution of the cytoplasm and changes in macromolecular crowding (Neurohr et al., 2019). Cells are highly crowded, with macromolecular concentrations estimated to be between 80 and 400 mg/mL (Cayley et al., 1991; Zimmerman and Trach, 1991). Macromolecular crowding retards diffusion, influences protein volume and association equilibria (Dix and Verkman, 2008; Ellis, 2001; Zhou et al., 2008), including, for example condensate formation in vitro and in vivo (Delarue et al., 2018; Woodruff et al., 2017). These effects are caused by steric exclusion, next to weak chemical interactions (Gnutt and Ebbinghaus, 2016; Rivas and Minton, 2016; Sarkar et al., 2013), and depend on the concentration, size, and shape of the molecules involved and are larger when crowders are smaller sized than the reacting molecule (Marenduzzo et al., 2006; Rivas and Minton, 2016). For example, an increased number of ribosomes slows down diffusion of 20 nm and 40 nm particles, but not average-sized proteins (Delarue et al., 2018). The propensity to undergo a phase transition to a ‘gel-like’ state is amplified by macromolecular crowding (Joyner et al., 2016) and also influenced by the pH: Starvation of yeast cells leads to acidification of the cytoplasm, which leads to a phase transition that hampers diffusion of µm-sized particles (Munder et al., 2016). Therefore, quantification of macromolecular crowding on the single-cell level could provide evidence on whether crowding could be a driver of aberrant biochemical organization in aging.

The volume of a cell needs to be coupled to biopolymer synthesis in order to maintain macromolecular crowding (Burg, 2000; Minton et al., 1992; van den Berg et al., 2017; Zimmerman and Minton, 1993). For example, it has been suggested that oversized yeast cells have reduced molecular density (Neurohr et al., 2019). One of the most dramatic features of aged yeast cells is a marked increase in cell size (Fehrmann et al., 2013; Janssens et al., 2015; Lee et al., 2012). Concomitantly, yeast organelles like vacuoles (Lee et al., 2012), the nucleus (including nucleoli Crane et al., 2019; Kennedy et al., 1997), and mitochondria (Scheckhuber et al., 2007) can exhibit changes to their respective morphology. These changes in compartment size and shape directly impact the cytosolic volume and additionally present physical barriers to molecular movement and surfaces on which molecules can be adsorbed. Furthermore, changes in compartmental volume alters energy consumption. For example, a small compartment, such as the endosome, needs to import fewer protons to maintain pH compared to larger compartments (Luby-Phelps, 1999). Additionally, specific cell types have different organelle sizes related to their function: for example, secretory cells have expanded ER (Federovitch et al., 2005). However, despite its importance, it is not clear how much compartmental volumes change during aging.

Here, we present a framework describing how the critical and interconnected physicochemical parameters of pH and crowding change during yeast replicative aging. We find that the cytosol shows modest acidification in early aging and drops significantly after the cells stop dividing. We optimize our macromolecular crowding sensors for long-term FRET measurements and show that longer-lived cells tend to maintain macromolecular crowding better than shorter-lived cells. While macromolecular crowding changes only modestly in aging, we observe drastic changes in organellar volumes, leading to crowding on the µm scale, which we term organellar crowding. In light of our evidence, pH and crowding must be taken into account when investigating and interpreting the hallmarks of aging.

Results

Yeast replicative aging leads to acidification of the cytosol, especially after entry into senescence

To follow cytosolic pH levels in aging, we utilized the fluorescence-based, genetically encoded pH sensor, ratiometric pHluorin (Miesenböck et al., 1998). pHluorin is a GFP variant that responds to changing pH: With increasing acidity, the excitation at 390 nm decreases, while the excitation at 475 nm increases. We expressed pHluorin in the BY4741 background strain and recorded the F390/F475 ratios from aging cells using the ALCATRAS microfluidic device (Crane et al., 2014; Figure 1A). We followed single-cell life histories from 80 cells over 80 hr, taking widefield fluorescent images every 10 hr and bright-field images every 20 min. Additionally, we performed an in vivo calibration of pHluorin with cells loaded in the microfluidic device (Figure 1B).

Figure 1 with 3 supplements see all
The cytosol of old cells acidifies, and cells display more substantial variability in cytosolic pH with aging, especially after entry into senescence.

(A) Representative images of the same cell expressing ratiometric pHluorin imaged at the start of the experiment and after 80 hr; the replicative age is indicated. Young cells are trapped in the microfluidic device, and bright-field images are taken every 20 min to determine the age of the cells. Fluorescent images are taken once every 10 hr (panels, C, D, F) or every hour (panel E) with excitation at 390 and 475 nm and emission at 525 nm. The images show DIC, DAPI, and FITC channels and merged fluorescence. The scale bar represents 5 μm. The trapped aging cell is indicated with white arrowheads. (B) Calibration curve showing the relationship between intracellular pH and pHluorin ratios. Cells expressing pHluorin from an exponential culture at OD600 of 0.5 were resuspended in buffers, titrated to pH 5, 5.5, 6, 6.5, 7, 7.5, 8, containing final concentrations of 75 μM monensin, 10 μM nigericin, 10 mM 2-deoxyglucose, and 10 mM NaN3. Monensin and nigericin are ionophores that carry protons across the plasma membrane, while the 2-deoxyglucose and NaN3 deprive the cell from ATP and thus block energy-dependent pH maintenance. Each point represents data from 20 cells. (C) Data collected from 80 yeast cells during the process of replicative aging. The young group consists of data points from the first recorded ratio at time 0 hr, and the old group consists of the last recorded measurement before cell death. Colored crosses indicate the average and bold lines show the median, p=9.1E-18. (D) Single-cell profiles of the pH at different replicative ages. Cells are grouped according to lifespan (RLS <15, 15–20, 20–25, 25–30 and >30). Grey bar indicates the range of pH values measured in the first time point (young age) to illustrate that 36% of the cells remain within this range throughout their lifespan. (E) Cytosolic acidity (normalized ratiometric pHluorin read-out:F475/F390, top) and replicative age (bottom) as a function of time, for a cell that enters senescence at 26 hr (dashed blue line) after which a sharp acidification of the cytosol (dashed red line) follows. (F) The pH of single cells during aging plotted versus their age in hours. Each grey line represents a single cell. The orange line shows the average pH. The dark grey circles show the total proteome pI in aging estimated based on the PI and changes in abundance of 1229 proteins. Error bars represent the standard error of the mean (SEM).

To compare differences in pH between populations of young and aged cells, we grouped the first and the last measurements taken for each cell under two categories: ‘young’ and ‘old’. The ‘young’ category contains pH measurements in the first fluorescent image when cells are predominantly age 0, since newborn daughters are the most common group in an exponentially growing culture. These young cells have an average cytosolic pH of 7.7 and pH levels are comparable between individual cells (Figure 1C). Furthermore, the measured average cytosolic pH is higher than the previously reported pH of 7.3 (Orij et al., 2009), but similar to the pH reported from population-based measurements (Knieß and Mayer, 2016). The ‘old’ category contains the last pH measurement of a cell, taken less than 10 hr before its death. This category is composed of senescent and non-senescent cells of different replicative ages, where most cells had no divisions left to complete (n = 59, 100% of RLS), and others had up to five divisions to complete (n = 21, 78-96% of RLS). The average lifespan of this cohort is 23 divisions. We observe that the pH of individual old cells spans a wide range of pH values from 7.8 to 5.7, and 36% of the cells maintain a pH above 7.5, which is the lower boundary in young cells. Comparing the old and young cells, our results show that the average cytosolic pH significantly decreases by 0.5 pH units in old cells, compared to young (Figure 1C), corresponding with previous findings (Knieß and Mayer, 2016).

As our dataset is composed of single-cell life trajectories, we can analyze the intrinsic variability in cytosolic pH values within the population, the timing of changes in pH, and the correlation of these phenotypes to cell lifespan. As depicted in Figure 1D, we plot the pH of single-cell trajectories for different ranges of replicative lifespans. We observed a gradual decrease in pH already early in life in almost all cells, and interestingly this gradual decrease is followed by a substantial drop in pH in a subpopulation of cells that stop dividing and enter senescence (Figure 1D). The decrease in pH in the earlier divisions (through a replicative age of 15 ± 2) correlates to the cell lifespan (R2 = 0.19, p=0.0007185) (Figure 1—figure supplement 1C), but we find no relationship between the lifespan of cells and pH at a young age, or the pH at old age (Figure 1—figure supplement 1D–F). Life trajectories of the cells with a very low pH at young or old age (the population outliers of Figure 1C) are excluded from 1D (Figure 1—figure supplement 1A and B). While the outliers from the aged population have nothing in common, the seven young cells, which had a lower pH than 91% of the young cohort, all increase their cytosolic pH within the next 10 hr to levels above 7.5 (Figure 1—figure supplement 1B). We conclude that, apart from this subpopulation of young cells with low starting pH, a shared phenotype of all aging yeast cells is that the cytosolic pH drops gradually and modestly throughout the mitotic lifespan, and that when cells stop dividing but remain alive, the pH decreases steeply.

To investigate the precise timing of entry into senescence in relation to the sudden pH decrease in the cytosol, we generated a data set with a higher time resolution where fluorescent measurements were taken in intervals of 1 hr for the total duration of 50 hr. We collected and analyzed single-cell trajectories from 50 cells (see Materials and methods section). These cells have a shorter lifespan compared to the cells from the low time resolution (Supplementary file 3), possibly due to higher phototoxicity. We qualitatively determine three types of behavior. Firstly, we find that 36% of the population is comprised of cells that display an abrupt entry into senescence. These cells have the highest increase in cytosolic acidity (~2-fold increase, SD = 0.4), and their entry into senescence always precedes the sharp increase in acidity (Figure 1E, Figure 1—figure supplement 2A; Figure 1—figure supplement 3). Hence, cytosol acidification cannot be the cause of entry into senescence in these cells. Secondly, we observe that 50% of the cells have a gradual entry into senescence, where the cell cycle is significantly extended, but sporadic divisions occur. It is likely that these cells later on become post-mitotic, but this phase is not captured within the duration of the experiment. Throughout the experiment, cells with gradual entry into senescence exhibit smaller (1.6-fold increase, SD = 0.22) and a more gradual increase in cytosolic acidity (Figure 1—figure supplement 2B; Figure 1—figure supplement 3). Lastly, cells that actively divide throughout the duration of the experiment (14% of the total population) exhibit little changes in their cytosolic pH (1.4-fold increase, SD = 0.14) (Figure 1—figure supplement 2C; Figure 1—figure supplement 3). In all cells, the fold increase in cytosolic acidity with aging correlates to the time cells spend in their last division (R2 = 0.6; Figure 1—figure supplement 2D). Overall, these data show that the drop in pH is posterior to the entry into a post-mitotic state and the degree of cytosolic acidification is dependent on the time spent in post-mitotic state.

In all cells, more than one measurement could be performed during the same cell cycle, if its duration lasted longer than 2 hr, allowing us to observe cell cycle-induced pH fluctuations. We find that there are fluctuations in the pHluorin read-out with the different cell cycle stages, but also find these to be minor and generally overpowered by the larger changes occurring after entry into senescence (Figure 1—figure supplement 2A–C). Considering that the majority of cells collected in the aged population are senescent cells, we conclude that the increased heterogeneity in the old group (Figure 1C) is dependent on the senescent status of the cells and not on the cell cycle stage at which the last measurement was performed.

We next asked how changes in cytosolic pH compare to those measured in the vacuole and cell cortex. Using the pHluorin2 (Mahon, 2011) and the same microfluidic design Crane et al., 2014, Chen and colleagues (Chen et al., 2020) measured the changes in vacuolar pH allowing comparison with our datasets (Figure 1—figure supplement 2E). From both studies, it is clear that the vacuole and cytosol change their pH in opposite directions during aging where the vacuole decreases in acidity and the cytosol becomes more acidic. An early life increase in cortical pH was reported previously (Henderson et al., 2014), which is the opposite of a decrease in the cytosolic pH observed in our long-term experiments. This apparent dichotomy prompted us to ask whether changes in the aging proteome can account for the distinct behavior of the cytosolic pH in relation to the vacuole and cell cortex. Indeed, in addition to the activity of proton pumps, the strong buffering capacity of metabolites and amino acid side chains contributes to pH homeostasis (Moriyama et al., 1992). Because the concentration of amino acids with a physiologically relevant pKa at protein surfaces is orders of magnitude larger than the concentration of free protons at pH 7 (protein concentration is in the mM-range while pH 7 corresponds to 60 nM H+), the proteome represents a buffer for changes in pH. Thus, we assessed the proteome isoelectric point (pI) during yeast replicative aging.

We utilized available datasets for protein abundance during aging (Janssens et al., 2015), predicted isoelectric point (Saccharomyces Genome Database, SGD), and protein copy number (Ghaemmaghami et al., 2003). We used data from 1229 proteins and corrected our calculations for relative protein abundance, thus weighing the proteome pI for the copy number of each protein. We found that in young cells, the proteome pI is 7.1, which accounts mostly for the cytosolic part of the cell, according to the Panther database for gene ontology. In cells aged for 60 hr (average replicative age ~22 divisions), the proteome pI reaches as low as 6.7 (Figure 1F). While this analysis carries uncertainties, for example related to protein pI predictions and age-related aggregate formation, it is striking that the proteome pI roughly follows the pH of the cytoplasm during aging. This suggests that additional to changes at the level of proton pumps, changes at the level of the buffering capacity of the proteome may underlie the decrease in cytosolic pH.

Validation of FRET-based crowding probes during yeast aging

Because pH has previously been related to macromolecular organization (Joyner et al., 2016; Munder et al., 2016), we asked whether macromolecular crowding also changes in aging yeast cells. It is not trivial to quantitatively measure crowding in living organisms (Rivas and Minton, 2016), and most estimates of crowding were obtained from measuring dry cell mass (Cayley et al., 1991; Zimmerman and Trach, 1991), which does not necessarily reflect the actual in vivo levels of crowding. Additionally, crowding has not yet been addressed in aging. We previously developed a genetically encoded FRET-based probe that enables quantification of macromolecular crowding in vivo (Boersma et al., 2015). The sensor is genetically encoded, and it harbors a FRET pair, connected with a flexible linker. When placed in a crowded environment, the probe will obtain more compressed conformations, thus increasing the FRET efficiency. We utilized this FRET-based probe, named crGE, and harboring mCerulean3 as donor and mCitrine as acceptor (Liu et al., 2018). We also developed a new probe, named CrGE2.3 by exchanging the donor and acceptor for mEGFP and mScarlet-I, respectively (Figure 2A, left).

Figure 2 with 1 supplement see all
Crowding sensor with traditional CFP-YFP FRET pair is functional in yeast cells, but sensitive to pH variations; the sensor CrGE2.3 with mEGFP-mScarlet-I is aging-compatible.

(A) Left: Schematic representation of the crGE and CrGE2.3 sensors. Right: Yeast cells expressing crGE sensor, the DIC, mCerulean3, and FRET channels are shown. Ratiometric image shows changes in the crowding ratio upon osmotic shock. The scale bar is 5 μm. (B) CrGE sensor (left), expressed under a strong constitutive TEF1 promoter in yeast, shows an increase in FRET/CFP ratio upon osmotic shock with 1M NaCl versus control in sodium phosphate buffer at pH 7. Crosses indicate the average and bold lines show the median. The data is from n > 58 cells per condition; p=2.0E-25. Right: CrGE2.3 sensor, expression and osmotic shock experiment is the same as for the original CrGE sensor, the data is from 30 cells per condition, p=1.6E-24 (C) Left: Fluorescence intensities of mCerulean3 (blue line), and mCitrine (yellow line; directly excited) and FRET/CFP ratios (black circles). Right: Fluorescent intensities from mEGFP (green line) and mScarlet-I (magenta; directly excited) and FRET/mEGFP ratio (black circles) from permeabilized cells, expressing the crowding sensor CrGE (left) and CrGE2.3 (right), in buffers with pH ranging from 5 to 8. Each data point represents the average of 20 cells; error bars show standard deviation.

Through fluorescence microscopy, we determined the crowding read-out from the ratio of the intensity in the FRET channel and the donor channel (an example is shown in Figure 2A, right). When the crowding levels are higher, the fluorescence signal in the FRET channel also increases, leading to a higher FRET ratio and vice versa. To validate the read-out of the FRET probes, we induced a hyperosmotic shock by resuspending cells from an exponentially growing culture in 1M NaCl. Upon osmotic upshift, the water content and the cell volume should decrease, resulting in increased macromolecular crowding. Indeed, we observed a significant increase (p<0.001) in the crowding ratio of 18% from 0.54 to 0.64 from three independent experiments for crGE (Figure 2B, left). To determine whether in the new sensor, crGE2.3, functionality is retained after exchanging the fluorophores, we again induced hyperosmotic shock with 1M NaCl or 1.5M Sorbitol and observed significant increases in crowding ratio (p>0.001) of 23% from 1.19 to 1.54 for NaCl and to 1.46 for Sorbitol (Figure 2B, right and Figure 2—figure supplement 1I). These results show that both sensors are responsive to crowding changes in yeast in a reproducible manner.

In order to perform crowding measurements in aging cells, the sensor should be insensitive to age-related changes, such as proton concentration or fluorescent protein maturation. Because the cytosolic pH decreases with age, it could compromise read-outs from the crowding sensor. To estimate the pH sensitivity of the crGE and the crGE2.3 probes, we resuspended cells permeabilized with ionophores in buffers with pH ranging from 5 to 8. Our results show that read-outs from the sensor harboring the original FRET pair, mCerulean3-mCitrine, has a strong linear dependence on the intracellular pH (Figure 2C, left). When assessing the pH sensitivity of the new sensor, CrGE2.3, we found that this probe gives more stable read-outs within the pH range observed in aging (Figure 2C, right).

Fluorescent protein maturation also plays a role: Aging cells do not maintain the same division frequency, but transition to a slow and irregular division mode, called the senescence entry point (SEP) (Figure 2—figure supplement 1, top row) (Fehrmann et al., 2013). When the SEP occurs, the relative rates of sensor synthesis, degradation, and dilution through division are all disrupted, leading to an increase in fluorescence of the slower maturing fluorescent protein (Figure 2—figure supplement 1A). For aging studies, the FRET pair in crGE2.3, which has similar pH sensitivity and maturation kinetics between the two fluorescent proteins, is thus more suitable than the original sensor. To eliminate additional systematic errors, we corrected for an unequal number of donor and acceptor fluorophores by FRET normalization (NFRET) (Xia and Liu, 2001) as pH-induced fluorescence quenching leads to a different number of fluorescent proteins and the same accounts for the proportion of fully matured sensors. Indeed, determining the NFRET eliminates differences between cells treated with 1 μM cycloheximide (Figure 2—figure supplement 1A–D) and in permeabilized cells at pH levels of 7 and 7.5, which are physiologically relevant pH levels where mScarlet-I signal shows pH dependence (Figure 2—figure supplement 1E and F). Additionally, the normalization retains read-outs from crowding changes (Figure 2—figure supplement 1G–J). Thus, in crGE2.3 we have increased the robustness of the probe to make it suitable for challenging long-term aging experiments.

Crowding homeostasis is maintained during yeast replicative aging

To determine age-associated changes in macromolecular crowding, we constitutively expressed the optimized crowding probe in yeast cells immobilized in the ALCATRAS microfluidic device (Crane et al., 2014). As with the pH-sensor , we observe that the crowding probe localizes in the cytosol and nucleus, but it is excluded from other membrane-enclosed organelles (Figure 3A). We performed three independent experiments (Figure 3B, Figure 3—figure supplement 1B and C). From the first experiment, we collected data from 80 cells in the time course of 70 hrs and determined NFRET and FRET/mEGFP ratios (Figure 3—figure supplement 1A). Our measurements show heterogeneity in crowding levels in young cells and more so in old cells (Figure 3B, Figure 3—figure supplement 1B and C). Overall, we find that macromolecular crowding is maintained in the course of aging, with average ratios of 0.51 for both young and old cells, where, as in Figure 1, the ‘young’ group reflects the first and the ‘old’ group the last crowding measurements (Figure 3B). Two additional independent experiments showed similar trends in aging, albeit with small variation in absolute crowding levels (Figure 3—figure supplement 1B and C).

Figure 3 with 1 supplement see all
Crowding remains remarkably stable in aging.

(A) Yeast cells expressing the crGE2.3 sensor, trapped in the aging chip. Images are from the same cell at the beginning of the experiment and the last measurement taken before cell death. Fluorescent images are taken once every 10 hr. Boxes indicate a cytoplasmic region. The scale bar is 5 μm. (B) Boxplot, showing normalized crowding ratios in young and old cells. For the young population, the first ratios recorded at the start of the experiment were taken. For the old population, the last ratio recorded before cell death was taken. The old population contains cells from different ages with a median lifespan of 18 divisions, n = 80. The ratio recorded for the cell displayed in Figure 3—figure supplement 1A is shown with a red dot. (C) Single-cell trajectories of cells with indicated replicative lifespan-ranges and an indicated number of cells in each category. Grey boxes indicate the range of crowding ratios at the end of the replicative lifespan (RLS) of each age group. (D) The fold change in crowding plotted against the replicative lifespan of cells, R2 = 0.22, p=1.3E-05. (E) The average first and last recorded NFRET values of cells with an RLS shorter (red) or longer (blue) than the median RLS of 18 divisions. There is no difference in crowding between young and old cells from the long-lived population. However, in the short-lived population, the old cells have significantly higher crowding than the young cells (p=0.007), and the old cells from the long-lived population (p=0.003).

Plotting single-cell trajectories for cells that reach a replicative lifespan of 10, 10–15, 15–20, 20–25, or larger than 25 shows that the shortest-lived cells tend to increase the crowding levels during their lifespan, while the longer lived cells tend to have more stable crowding levels (Figure 3C). Indeed, there is a weak correlation (R2 = 0.14, p<0.001) between lifespan and old age crowding levels (Figure 3—figure supplement 1E), and the fold change in NFRET ratios in aging shows a weak correlation with lifespan (R2 = 0.22, p<0.001) (Figure 3D). In support of the relationship between crowding and aging, we observe that cells that live shorter than the average lifespan of 18 divisions have significantly higher ratios in aging (p<0.01), compared to long-lived cells (Figure 3E). It seems that it is the maintenance of crowding homeostasis, rather than the absolute crowding levels, which has an association with lifespan, as lifespan does not correlate to the crowding ratios in young cells (Figure 3—figure supplement 1F).

Volume distribution of cellular compartments changes disproportionally in yeast replicative aging

Cell volume increases in aging and the increase per division is predictive for the lifespan of cells (Janssens and Veenhoff, 2016b). Because macromolecular crowding is directly linked to cell volume, we aimed to determine how the volume of the cytosol changes in aging and assessed aged cells on the ultrastructural level. To explore the ultrastructure of aged cells exclusively, we labeled the cell wall of cells in log-phase with Alexa-488, which were then allowed to age over 20 hr. As the dye remains with the aging mother cells (Smeal et al., 1996), we can specifically identify aged cells within the population using correlative light (Figure 4A) and electron microscopy (CLEM) (Figure 4B). Aged cells make up only a minor fraction of the population; at the 20 hr time point used here, each aged (and labeled) cell is outnumbered by approximately 8000 daughter cells. Although the exact age of each cell was not determined, based on population doubling times we estimate that the Alexa-488 labeled cells will have performed on average 13 divisions, with a significant spread due to cell-to-cell differences in doubling time (Janssens et al., 2015). Fourteen tomograms of aged cells and 10 tomograms of young cells were acquired and segmented to define the plasma membrane, nucleus, vacuole, lipid droplets (LD), endoplasmic reticulum (ER), multivesicular bodies, and mitochondria (Figure 4—figure supplements 1 and 2). The aged cells show diverse phenotypes in terms of, e.g., numbers and size of vacuoles, or numbers of lipid droplets. The lipid droplets become more prevalent in aged cells, confirming previous findings (Beas et al., 2020; Figure 4—figure supplement 3). In almost all cells, especially the vacuoles take up a larger fraction of the cell volume than they do in young cells.

Figure 4 with 4 supplements see all
Inter-organellar crowding increases in aged cells.

(A) Identification of aged cells (20 hr, replicative age ~13) by CLEM. Overlay of low magnification (225x) electron micrograph and fluorescence image and zoom-in of old cells with Alexa-488 labeling (boxed). Scale bar is 10 µm. Boxed cells I and II are shown at higher magnification (8900x) in lower panels with a scale bar set at 1 µm. (B) Single slices of tomograms without (left panel) and with an overlay to emphasize organelles (middle panel). 3D isosurface rendering (right panel) of tomograms of young or old cells. Nuclei (orange), vacuoles (blue), lipid droplets (yellow), mitochondria (red), ER (magenta), and plasma membrane (green). Scale bars are 500 nm. (C) The relative cell volume occupied by vacuoles increases in old cells. The scatterplot shows volume fractions of nuclei and vacuoles in 14 aged cells and 10 young cells. The volume fraction of the cell section minus the volume fractions of the nucleus and vacuole is used to estimate of the cytosol volume fraction. The values for the example shown in B are boxed. Black lines denote the median. (D) Membrane surface areas of nucleus and vacuole as a fraction of the plasma membrane surface area in young (left) and aged (right) cell sections. Each point represents a single nucleus or the sum of all vacuole surface areas within a single cell. The boxes indicate values from cells in B; linear correlations, R2 values, are indicated. (E) Method to measure the distance between membranes in young and aged cells. Right is a scatter plot of measured distances from represented cells (F) Left is a scatter plot of inter-organelle distances between membrane-bound organelles in young and old cells. The black line demarks the median. Right is a histogram of inter-organelle distance distribution. Values were grouped into 200 nm width bins. n = 305 from 10 young cells, n = 778 from 14 old cells.

To quantify organellar volume in the tomograms, we focused on the vacuole and nuclei, which are the largest compartments. We calculated their volume and membrane surface areas relative to the total cell volume and plasma membrane surface area in each tomogram (Figure 4C). We find that the vacuoles take up 3% to 24% v/v in young cells and a higher volume fraction, namely 16% to 66% v/v, in old cells. The nuclei take up between 8% and 37% v/v in young cells, which is similar to old cells (7–30% v/v). Interestingly, we find that in old cells, vacuoles can occupy up to 66 % v/v, this is comparable to the theoretical maximum volume of 64% v/v that randomly packed spheres can occupy in a container. Subtracting the nuclear and vacuolar volumes from the total volume provides an estimate of the cytosolic volume in the analyzed tomograms. The cytosol volume takes up 54–82% v/v in the young-cell sections compared with 22–70 % v/v in those from old cells. Consistent with the changes in organelle volumes in aging, we find that the surface area of the vacuole and nuclear membranes relates to the surface area of the plasma membrane in the young cell sections. In contrast, the relation between organellar and plasma membrane surface areas is lost upon aging (Figure 4D). The loss of this relationship provides the correspondingly wide distribution in occupied organellar volumes.

To assess the consequences of the changes in organellar crowding, we determined the average distance between organelles (Figure 4E), which would be a measure for cytoplasmic confinement of larger particles induced by organelles. We find a striking decrease in the most common distance from ~500 nm in young cells to ~100 nm in old cells, albeit with a wide distribution. Given a diffusion coefficient of 0.1 µm²/s previously determined in yeast cells for a particle with a diameter of 40 nm, it would take 0.005 s versus 0.3 s to reach a membrane with Brownian diffusion alone, assuming similar viscosity (Materials and methods; supplementary information, Figure 4—figure supplement 4).

Our analysis of the ultrastructure of aged cells further suggests that even though the total cell size increases in replicatively aging cells (Crane et al., 2014; Fehrmann et al., 2013; Janssens and Veenhoff, 2016b; Lee et al., 2012), the cytoplasmic volume fraction does not increase in aging, but rather remains stable or even decreases as the vacuolar volume fraction increases. Moreover, we conclude that crowding at the length scales of organelles, which we coin organellar crowding, strongly increases with aging.

Discussion

Here, we provide an analysis of the progressive change during aging for several parameters that define a cell’s intracellular environment; namely, cytosolic pH and crowding on the scale of macromolecules and organelles, all impinging on the hallmarks of aging. We find that the largest changes arise from organellar crowding, exemplified by the average membrane-to-membrane distance in aged cells being >2 times smaller in aged cells than in a young cell, while macromolecular crowding homeostasis is mostly maintained. The cytosolic pH shows a progressive decline that follows the pI of the proteome throughout much of the lifespan, while a steeper decrease in pH is observed at senescence. Below, we discuss possible causes of this changed intracellular environment and implications on molecular processes in an aging cell.

pH homeostasis in aging

In general, pH has far-reaching consequences on cell physiology. Among others, the pH influences protein folding, enzyme activity, phosphorylation of metabolites and proteins, protein solubility/phase separation, interactions between the molecules, redox potential, proton gradients, and proton-dependent transport of nutrients (Orij et al., 2011). Even pH variations of 0.5 pH as we measure here, cause various enzymes to lose activity (Ju et al., 2004; Talley and Alexov, 2010) or induce liquid-liquid phase separations of proteins in cells (Triandafillou et al., 2018).

Because the drop in cytosolic pH has limited predictive value for lifespan (Figure 1—figure supplement 1F), we conclude that it is not a general early cause of aging. Significantly, we find that the sharp acidification of the cytosol only occurs after the cells become senescent. This suggests that senescence entry is not driven by cytosol acidification and there is possibly another underlying cause for both entry into senescence and cytosol acidification. Additionally, the degree of acidification of the cytosol, and therefore the increased heterogeneity in the old population, is related to the time spent in the senescent state. We speculate that a low availability of energy in the aging cell is a potential driver of senescence entry and cytosolic acidification.

Previously pH has been linked to aging in studies that showed both the vacuole and cell cortex become more basic with age (Henderson et al., 2014; Hughes and Gottschling, 2012) and the cytosol more acidic (Knieß and Mayer, 2016). Our findings add to this work, providing new insights only possible from the single-cell level. We show that cytosolic acidity strongly increases only after entry into senescence and we do not observe drastic changes in early lifespan. The changes in cytosolic pH are thus decoupled from those in the vacuole, which occur very early in the replicative lifespan (Chen et al., 2020; Hughes and Gottschling, 2012) and have no established connection to the timing of the senesce entry point. Comparing with previous measurements of the pH at the cell cortex with a plasma membrane-anchored pH sensor (Henderson et al., 2014), our measurements with a freely diffusing cytosolic pH sensor suggest that the plasma membrane sensor reports a local pH distinct from the cytosol. At the cortex, the pH was shown to alkalize already in age 3 cells, while the cytosolic pH acidifies mostly so after the cells enter senescence. Previously, an elegant model was proposed that explained why the pH of the vacuole and at the cell cortex alkalize in aging cells, which entailed competition for protons between the vacuolar and plasma membrane ATPases (Henderson et al., 2014). From the distinct timing of changes in vacuole and cytosol, plus the distinct directions of the changes in the cytosol and cell cortex, we conclude that this model alone cannot fully explain the acidification of the cytosol. We suggest that pH homeostasis in the cytosol relies on additional mechanisms, and present data to support that the proteome itself could provide buffering capacity (discussed below). To explain the alkalinization of the cell cortex in aging cells, we speculate that the high concentrations of proton pumps in the plasma membrane contribute to creating an alkaline microenvironment. On the other hand, the alkalization of the vacuoles could, besides the previously proposed competition for protons (Henderson et al., 2014), be driven by the loss of functional vacuolar ATPases. The latter relation is supported by the observations of loss of the stoichiometry of the ATPase subunits in aging (Janssens et al., 2015) and the partial recovery of the stoichiometry after overexpression of the Vma1 component, which subsequently increases lifespan in yeast (Hughes and Gottschling, 2012). In future studies, the measurement of the pH at all three subcellular locations simultaneously in single cells and at high time resolution, combined with interventions at the proteome level plus the activities of the ATPase, would test this model.

Can the decreasing pI of the proteome, observed in aging be a source for the drop in cytosolic pH? Cellular proteins have weakly acidic or basic residues, which collectively act as a buffer of cellular pH. It is proposed that yeast proteins evolved to have a pI that is similar to the pH of the compartment where they usually reside in Brett et al., 2006, although this is not described in human cells (Garcia-Moreno, 2009). However, a cell must regulate its pH away from the pI to maintain protein solubility and prevent protein phase separation or aggregation. Thus, the yeast proteome may provide the basal pH of a cell compartment, while energy-dependent mechanisms regulate it away from the pI. This principle has been demonstrated in vitro, where isolated lysosomes maintain their acidity through a Donnan-type equilibrium (Moriyama et al., 1992). Hence, if a cell’s energy state does not allow proper pH regulation, the pH would follow the pI. Therefore, the drop in pI of 7.1 to 6.7 could be responsible for the modest drop in pH during replicative aging, and when cellular energy is lost upon senescence, a more considerable drop ensues. In young cells, the vacuolar pH is around 5.5 (Plant et al., 1999), while its pI is 6.5 (Brett et al., 2006). Unfortunately, the vacuolar pH sensor read-outs were not calibrated to yield actual pH values (Chen et al., 2020; Hughes and Gottschling, 2012), but it would be interesting to know whether pH levels in the vacuoles also approach values equal to the pI, as in the cytosol. Such observations raise the question of whether in old cells, the proteome becomes not only the main buffering mechanism, but also a regulatory mechanism for cellular pH. The cell-to-cell variation in pH that we observe during aging could, in turn, reflect variation in proteome composition or energy state. Additional mechanisms could add to the changes that we observe, such as altered proton transport activity, accumulation of acidic metabolites, or polynucleotide content. Nevertheless, the qualitative similarities between the pI and pH are striking and suggest the proteome contributes to the cytosolic pH in aging.

Crowding homeostasis in aging

Physicochemical parameters such as pH and macromolecular crowding can determine biochemical organization in yeast (Joyner et al., 2016; Munder et al., 2016) and potentially cause disease (Patel et al., 2015). Changes in physicochemical homeostasis could explain the multifactorial nature of the aging process since they will have ubiquitous and diverse effects on all cellular processes (Alberti and Hyman, 2016). We investigated macromolecular crowding as an intrinsic cell property, which can modulate, for example, phase separation and transition in vivo (Delarue et al., 2018; Joyner et al., 2016). Considering the significant increase in volume the cell undergoes with age, it is at first somewhat surprising that we find crowding levels to be very stable in old yeast cells. This stability may be a consequence of a combination of relatively small changes in cytoplasmic volume fraction (discussed below) and efficient mechanisms to maintain macromolecular crowding homeostasis.

To date, the only way to change crowding levels in the cells considerably and rapidly is to apply an osmotic upshift in the medium. The cell responds immediately by uptake of potassium ions to regain its volume and crowding (Granados et al., 2018). Even in a potassium-deprived medium, the crowding effects are observed for a short period due to other response mechanisms and/or efficient uptake of trace amounts of potassium. These short response times indicate that crowding levels are crucial. Crowding could be regulated by an array of mechanisms that prevent drift in crowding over more extended periods. These include, for example, (1) uptake of counter ions upon biopolymer synthesis, inducing an osmotic pressure over the membrane resulting in cell growth and reduction of the biopolymer concentration (Basan et al., 2015), (2) carbon catabolite repression to reduce the space taken up by metabolic enzymes (Zhou et al., 2013), or (3) altering the ribosome/tRNA concentration (Delarue et al., 2018; Klumpp et al., 2013). These, and yet to be identified mechanisms could regulate crowding in aging. The robust sensor developed here will provide a valuable tool to identify the genes that maintain macromolecular crowding at the nanometer scale.

Are there consequences of crowding during aging? We find that cells with crowding levels that remain relatively unchanged during aging have a longer lifespan, while cells increasing in crowding tend to be average or shorter-lived. Young cells display a natural variability of crowding between cells that is independent of lifespan. Therefore, retaining initial crowding levels could be beneficial for old cells. A too-large drop in crowding would reduce cell viability: It was recently suggested that dilution of the cytoplasm could evoke cell cycle arrest and lead to senescence, through a variety of mechanisms (Neurohr et al., 2019). Possibly, an optimum macromolecular crowding exists, from both a physicochemical and biochemical viewpoint, within a window of less optimal but viable crowding levels.

Organellar crowding in aging

Live-cell imaging studies of yeast expressing GFP-tagged organelle markers have highlighted how several organelles change in abundance or shape in aging yeast cells. Amongst them are the increase in vacuolar size (Lee et al., 2012) and fragmentation of nucleoli (Crane et al., 2019; Kennedy et al., 1997) and mitochondria (Scheckhuber et al., 2007). However, because aged cells are scarce in exponentially growing cultures, the detailed ultrastructural properties of aged cells had not yet been researched using EM analysis. We use CLEM to reveal that aged cells have an altered ultrastructure. Particularly striking is that the available space for the cytoplasm can become minimal in aged cells and is enclosed by a large surface area of organellar membranes. The average membrane-to-membrane distance in aged cells is >2 times smaller in aged cells than in a young cell: The average distance between organelles decreases from ~1000 to <500 nm. The distribution is, however, strongly tailing and these averages correspond to a change in the most common distance from 400–600 to 0–200 nm. Already from the frequency distribution of the measured inter-organelle distances that are smaller than 80 nm, which is ~12% in aged cells and ~2% in young cells, one could deduce that contact sites are possibly expanded in aged cells. The implications of an increase in membrane contact sites can be widespread. Moreover, the enlarged compartmentalization must affect the movement of larger structures such as ribosomes and induce confinement on similar-sized particles in the 40 nm range.

The organellar crowding should have several effects that are highly dependent on the local distance between the membranes and the size of the particle, as demonstrated by the time required to diffuse to the membrane (supporting information, Figure 4—figure supplement 3). The proximity to the membrane also increases the likelihood of interactions with membranes. The particles will also suffer an entropic cost by sacrificing translational degrees of freedom in the inter-organellar spaces. Hence, larger particles may be crowded out of regions with high organellar crowding leading to a size-dependent spatial sorting. On a technical note, given the extreme dependence on particle size, particles that are much smaller than the distance between the organelles would notice less of this confinement, which would include, for example, the macromolecular crowding sensor, which is polymer-like with a radius of ~5 nm. In contrast, fluorescent particles with radii of ~20 nm should experience confinement when present in the typical 100 nm confinements. Therefore, the behavior of such a particle will be dependent on where it is inside the cytoplasm. From a biological perspective, future studies should address how altered spatial sorting of large biological relevant particles, such as ribosomes, RNPs, or protein condensates or aggregates, would affect the physiology of aged cells.

An integrated view of physicochemical homeostasis in aging

It is well established that aged cells are derailed in molecular aspects, such as the loss of protein homeostasis, genome stability or metabolic state, and that these molecular changes affect the cell’s physiology. However, the identity of cells and organelles is not only defined on a molecular composition level but is also defined by physical and physicochemical properties. The data presented here highlight that these are also drifting in aging. First, the physical property of organelle size drifts in aging: where the cytosol in young cells occupies most volume, followed by the nucleus and vacuole (Figure 4D left), in old cells the order is opposite: here the vacuoles are largest, and the cytosol represents the smallest volume fraction. In addition, the physicochemical property of pH changes in aging: where the pH of the vacuoles is kept much lower than the cytosol in young cells, their values come closer together as the cytosol acidifies (this study and Knieß and Mayer, 2016) and the vacuole loses acidity in aged cells (Chen et al., 2020; Hughes and Gottschling, 2012). Therefore, both in terms of size and pH, the vacuole and cytosol lose aspects of their compartmental identity. Lastly, crowding on the scale of organelles, which we term organellar crowding, sharply increases with aging. Organellar crowding likely influences phenomena that are on the 100 nm to µm length scale, such as long-range diffusion of larger particles and organelles, condensate formation, organellar shape, RNA translation, and cytoskeletal dynamics. Besides, the increased surface area presented by the organelles may give additional opportunity for adsorption or increased membrane contact sites. However, crowding at the length scale of a single protein, that is, ~10 nm, changes little, and these proteins do not experience a direct effect of organellar crowding. By analogy, an ant would not notice if there were a fence around a field, but an elephant does. We speculate that while in young cells, micrometer length structures are mostly hindered in their diffusion throughout the cytosol by cytoskeletal structures, in aged cells, the high occupancy of intracellular organellar membranes provides the major obstacle. In the context of these significant changes in the physical and physicochemical properties of aged cells, it is remarkable that macromolecular crowding on the scale of single proteins is rather stable in aging, and instigates future studies to identify what regulates these crowding levels.

Materials and methods

Plasmid construction and yeast strains

Request a detailed protocol

All yeast strains (Supplementary file 1) were constructed in the BY4741 genetic background (his3Δ1, leu2Δ0, met15Δ0, ura3Δ0) (Brachmann et al., 1998) and were transformed with pRS303 yeast integrative plasmid, harboring the respective sensor gene with a TEF1 promoter and CYC1 terminator. A complete list of primers used to construct the strains can be found in Supplementary file 2.

To construct SMY008, the yeast codon-optimized gene of the crGE-NLS sequence was amplified from pYES2 vector (GeneArt, Invitrogen) together with pTEF1 and CYC1T by PCR with the forward primer F1_SM and reverse primer R1_SM, introducing SalI restriction site downstream of the terminator and removing the NLS localization signal. The sequence was subcloned into a pRS303 yeast integrative vector in between SpeI and SalI sites. The construct was then used to obtain chromosomal integration of the sensor sequence in the HIS locus.

For the generation of SMY015, the yeast codon-optimized mEGFP-crGE-mCherry (GeneArt, Invitrogen) was amplified using primer F3_SM to introduce a HindIII site and the primer R3_SM to introduce a stop codon and a downstream XbaI site. The PCR product was subcloned in pYES2-TEF1 between HindIII and XbaI. The resulting TEF1-mEGFP-crGE-mCherry construct was amplified with F1_SM an R1_SM and subcloned in pRS303, as described above. The gene encoding for Gamillus-crGE-mScarlet-I (GeneArt, Invitrogen) was amplified with primers F4_SM and R4_SM to introduce a stop codon, as well as XmaI and XbaI restriction sites after the mScarlet-I. The resulting PCR product was digested with NcoI and XmaI to isolate the mScarlet-I gene. The mCherry sequence in the pRS303-mEGFP-crGE-mCherry was then substituted with mScarlet-I, by subcloning between the NcoI and XmaI restriction sites. The resulting construct of mEGFP-crGE-mScarlet-I in pRS303 was used for chromosomal integration into the HIS locus of BY4741.

To construct SMY012, the pHluorin gene was amplified from pYES2-ACT1-pHluorin (Diakov et al., 2013) by PCR, using primers F2_SM, introducing HindIII restriction site and AAAAAA for enhanced expression in front of the start codon, and R2_SM, introducing a stop codon and XmaI and XbaI restriction sites. The PCR product was subcloned in the pYES2-TEF1 vector between HindIII and XbaI restriction sites. The TEF1-pHluorin-CYC1T construct was then amplified by PCR with primers F1_SM and R1_SM and integrated into the pRS303 with SpeI and SalI restriction sites. All pRS303 constructs were sequenced.

Media and growth conditions

Request a detailed protocol

Yeast cells were grown at 30°C, 200 rpm in Synthetic Dropout medium without histidine (SD-his), supplemented with 2% (w/v) glucose. Cells from an overnight culture are diluted 100 × in 10 mL of SD-his, 2% glucose. After 7 hr of incubation, appropriate dilutions were made to obtain cultures in the exponential growth phase on the following day (OD600 = 0.4–0.7).

Microscopy

Request a detailed protocol

All in vivo experiments were performed at 30°C. Images were acquired using a DeltaVision Elite imaging system (Applied Precision (GE), Issaquah, WA, USA) composed of an inverted microscope (IX-71; Olympus) equipped with a UPlanSApo 100× (1.4 NA) oil immersion objective, InsightSSI solid-state illumination, ultimate focus, and a PCO sCMOS camera.

Excitation and emission were measured with the following filter sets, respectively, and in the indicated order: crGE: CFP 438/24 and 475/24 nm, YFP 513/17 and 548/22 nm, FRET 438/24 nm and 548/22 nm. crGE2.3: FITC: 475/28 and 525/48 nm, A594: 575/25 and 625/45 nm, FRET: 475/28 and 625/45. pHluorin: DAPI: 390/18 and 435/48 nm, FITC: 475/28 and 525/48 nm. For crGE and pHluorin, 32% transmission power and for crGE2.3 2% for the FITC channel and 32% for the A595 and the FRET channel. For the aging experiments, stacks of 3 or 4 images with 0.7 μm spacing were taken, and for other experiments, stacks of 30 images with 0.2 μm spacing were taken at an exposure time of 25 ms for all experiments.

Aging experiments

Request a detailed protocol

The microfluidic chips were used as described previously (Crane et al., 2014). DIC images of the cells were taken every 20 min to follow the number of divisions for each cell and determine the replicative lifespan. Fluorescent images were collected every 10 hr. Z-sections were taken in both DIC and fluorescence imaging with three or four slices of 0.7 μm thickness. The experiment was left to continue up to 80 hr and only cells that were trapped from the beginning of the experiment and died in the device within the time course of the experiment were included in the analysis.

Tracing of cytosolic acidity at high time frequency

Request a detailed protocol

Images were acquired using a DeltaVision Elite imaging system (Applied Precision (GE), Issaquah, WA, USA) composed of an inverted microscope (IX-71; Olympus) equipped with a UPlanSApo 100× (1.4 NA) oil immersion objective, InsightSSITM Solid State Illumination, excitation and emission filters DAPI: 390/18 and 435/48 nm, FITC: 475/28 and 525/48 nm, ultimate focus and a CoolSNAP HQ2 camera (Photometrics, Tucson, AZ, USA). Exposure time was 0.1 s at 32% transmission. Fluorescent images were collected every 1 hr and DIC images were collected every 20 min for the whole duration of the experiment of 50 hr. Only cells that were present in the microfluidic device from the beginning of the experiment were analyzed. All cells were analyzed regardless of whether they died, became senescent, or were mitotically active until the end of the experiment. After entering senescence, some cells begin to lose their fluorescence signal reaching levels close to those of the background. In these cases, the measurements were further processed only if the fluorescent signals from both channels were at least twice the background to ensure reliable read-outs from the pHluorin.

Image analysis

Request a detailed protocol

Processing of all images was performed using Fiji (ImageJ, National Institutes of Health). For each image, the z-stack with the best focus was selected. pHluorin and the crowding sensors localize in the cytosol and nucleus, and appear to be excluded from the vacuole and probably also from other membrane-enclosed cytoplasmic organelles. We determined the fluorescence in each channel for the entire cell and subtracted the background from a region outside the cell. The respective ratios were subsequently calculated.

Determining NFRET

Request a detailed protocol

Fluorescence signals from the donor (ImEGFP), acceptor (ImScarletI), and FRET (IFRET) channels after background subtraction were used to calculate the normalized FRET (NFRET) (Xia and Liu, 2001). We did not correct for the donor bleedthrough in the FRET channel and the acceptor cross-excitation because these contributions were minimal.

NFRET=IFRETIdonor×Iacceptor

Statistical analysis

Request a detailed protocol

Statistical parameters, including the number of cells analyzed, are reported in the figures and corresponding figure legends. The significance of changes was determined with a two-tailed Student’s t-test; linear regression analysis (R2) was done in Excel, and Spearman's rank correlation coefficient (rs) were calculated in Matlab.

Relation between pHluorin ratios and pH values

Request a detailed protocol

As described in Munder et al., 2016, 2 mL of exponentially growing culture with OD600 of 0.5 were centrifuged at room temperature for 3 min at 3,000 g in a tabletop centrifuge. The cells were then resuspended in 200 μL calibration buffer (50 mM MES, 50 mM HEPES, 50 mM KCl, 50 mM NaCl, 200 mM NH4CH3CO2) at pH 5, 5.5, 6, 6.5, 7, 7.5, and 8. 75 μM monensin, 10 μM nigericin, 10 mM 2-deoxyglucose, and 10 mM NaN3 (final concentrations) to each buffer. The cells were then loaded in microfluidic chips (see below), and their fluorescence was determined (Figure 1B).

pH sensitivity of crGE and crGE2.3 crowding sensors

Request a detailed protocol

Cells from exponentially growing cultures at OD600 of 0.5 were harvested and resuspended in the same calibration buffers titrated to pH of 5, 5.5, 6, 6.5, 6.7, 7, 7.3, 7.5, 8 as the pHluorin calibration. The FRET/CFP and FRET/mEGFP ratios were determined from cells on a glass slide.

Cycloheximide treatment

Request a detailed protocol

A 20 mL exponentially growing culture was split into two cultures of 10 mL. To one of the flasks, a final concentration 1 μM cycloheximide was added from a 1000 × stock solution in DMSO. As a control, 10 μL DMSO was added to the other culture. Samples were collected immediately after the addition of cycloheximide or DMSO and imaged. Both cultures were then incubated at 30°C, shaking at 200 rpm for 90 min. Samples were collected after 90 min from both treatment and control cultures and imaged as described before.

Osmotic shock

Request a detailed protocol

2 mL of exponentially growing culture was collected by centrifugation at 3000 g for 3 min. The cells were resuspended in 200 μL low osmolality buffer (50 mM NaPi, pH 7 for isotonic conditions) or high osmolality buffer (50 mM NaPi, 1 M NaCl or 1.5 M Sorbitol) to induce the osmotic upshift. Cells were then placed immediately on the glass slide and imaged.

Proteome isoelectric point

Request a detailed protocol

To calculate the overall isoelectric point of the aging proteome, we used published datasets for the age-related change in protein abundance (Janssens et al., 2015), the protein copy numbers in young cells (Ghaemmaghami et al., 2003) and the computed isoelectric points (pI) (Saccharomyces Genome Database, SGD, https://www.yeastgenome.org/). We excluded 264 proteins from the aging dataset for which a copy number was not available. Out of the remaining 1229 proteins, 1071 proteins belong to the GO term ‘cytoplasmic component’ (Panther Gene Ontology; www.pantherdb.org), indicating the aging proteome mostly reflects (highly abundant) cytosolic proteins. For each time point in aging, the contribution of and individual protein to the overall pI was calculated by multiplying its pI with its relative abundance in the proteome. The total proteome pI was then derived as a sum of all weighed isoelectric points of the proteins in the dataset.

Ultrastructural analysis of aged cells

Request a detailed protocol

WT cells were grown to mid log phase in SD medium supplemented with 1% glucose. 8 × 107 cells were collected by centrifugation, washed twice with PBS, and resuspended in 0.5 mL of PBS. 4 mg of EZ-link Sulfo-NHS-LC-Biotin (ThermoFisher Sci.) was dissolved in ice-cold H2O and immediately mixed with the cell slurry, which was incubated for 20 min at RT. Biotin-labeled cells were collected by centrifugation and washed twice with PBS. After resuspension in 100 μL of PBS, 5 μL of 5 mg/mL streptavidin-labeled with Alexa Fluor 488 (ThermoFisher Sci.) was added for 30 min. After washing in PBS, 40,000 cells were used to inoculate 30 mL of SD with 1% glucose, which was grown to an average age of 13 before processing for CLEM as follows. Cultures were concentrated into a thick slurry by gentle centrifugation and aspiration of excess media. These samples were high-pressure frozen in a Leica HMP100, freeze-substituted in a Leica Freeze AFS with 0.1% uranyl acetate in dry acetone, and infiltrated with Lowicryl HM20 resin. The polymerized resin block was cut into ~200 nm thick sections onto 135 mesh H15 patterned copper/rhodium grids (Labtech). Fluorescence imaging was carried out as previously described in Kukulski et al., 2011. Fluorescent micrographs were acquired using a DeltaVision widefield microscope (Applied Precision/GE Healthcare) equipped with UPlanSApo 60x (1.64 NA) and 100x (1.4 NA) oil immersion objectives (Olympus), solid-state illumination and CoolSnapHQ2 CCD camera (Photometrics). Bright-field and fluorescent images of grid squares with cells were acquired at both 60x and 100x magnification to facilitate alignment to electron micrographs in subsequent steps.

Grids were post-stained with lead citrate and labeled on both faces with 15 nm gold fiducials. Tilt series from −60° to 60° of selected cells were acquired with a magnification of either 8900x or 13300x on an FEI F20 fitted with an FEI Eagle CCD camera (4k × 4 k) and using Serial EM (Mastronarde, 2005). Tomograms were reconstructed in IMOD (Kremer et al., 1996) using 15 nm gold fiducials for alignment. Low magnification (225x-440x) electron micrographs were also acquired to facilitate alignment with fluorescence micrographs and tomograms.

Correlation between light micrographs and electron micrographs was completed in Icy (de Chaumont et al., 2012) using the ec-CLEM plugin (Paul-Gilloteaux et al., 2017). Corresponding points in EM and light microscopy images were selected based on cellular features distinguishable in both the light microscopy and EM.

Analysis of electron tomograms

Request a detailed protocol

Manual segmentation of tomograms was performed in IMOD/3DMOD (Version 4.9.8, Kremer et al., 1996) with contours drawn every 30 nm or less in z. The surface area and organelle volume were calculated from uncapped meshed models in IMOD. The cytosolic volume was estimated by subtracting the calculated volume of nuclei and vacuoles from the volume encapsulated by the plasma membrane.

Inter-organelle distances were calculated in an unbiased manner by overlaying horizontal lines spaced every 200 nm over the midplane of the reconstructed tomogram using the stereology tool in IMOD. Distances between organelles on overlaid lines were then drawn as contours and measured. As cortical ER was not faithfully visible in each cell, all ER membranes were excluded from the analysis. Graphs were compiled in Prism (GraphPad). The linear correlation coefficients R2 were calculated in Prism (GraphPad).

Dependence of particle size on diffusion time to a membrane

Request a detailed protocol

The weighted mean distance d a particle travels by Brownian diffusion depends on the diffusion coefficient D and the time t following the equation (Phillips et al., 2012):

(1) d=2nDt

Here n is the dimensionality, which we set to 1. In reality, the particle can move in 3 dimensions. We set d as half the distance between the organelles, which is the distance the particle has to travel to the walls. The resulting t provides the time a particle diffuses on average to hit the wall. We thus assume the particle starts in the middle, and therefore the time estimates are longer than when it would start at another point.

This formula is derived from the chance c a particle has traveled to position x, which is at the organelle:

(2) cx,t=N4πDte-x²4Dt

The chance can be set arbitrarily to determine the time it requires for a particle to interact with the organelle. This formula will result in the same trends as Equation 1 for observing the relative effect of the change in organelle distance upon aging. We, therefore, continue with Equation 1.

We take the distance the particle has to travel as d = L/2, with L the distance between the organelles. To take into account the size of the diffusing particle, we set d = L/2 r, with r = radius of the particle. Inserting into Equation 1 gives for the average time for a particle to diffuse to the organelle:

(3) t=(L2-r)²2D

To compare old cells with young cells, we take the ratio t(young)/t(old) that provides the fold-decrease in time required to diffuse to the organelle membrane.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
    Senescence-associated (beta)-galactosidase reflects an increase in Lysosomal mass during replicative ageing of human endothelial cells
    1. DJ Kurz
    2. S Decary
    3. Y Hong
    4. JD Erusalimsky
    (2000)
    Journal of Cell Science 113:3613–3622.
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
    The cellular biology of proton-motive force generation by V-ATPases
    1. N Nelson
    2. N Perzov
    3. A Cohen
    4. K Hagai
    5. V Padler
    6. H Nelson
    (2000)
    The Journal of Experimental Biology 203:89–95.
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
    Physical Biology of the Cell (Second Edition)
    1. R Phillips
    2. J Theriot
    3. J Kondev
    4. H Garcia
    (2012)
    Garland Science.
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76

Decision letter

  1. Weiwei Dang
    Reviewing Editor; Baylor College of Medicine, United States
  2. Jessica K Tyler
    Senior Editor; Weill Cornell Medicine, United States
  3. Vyacheslav M Labunskyy
    Reviewer; Boston University School of Medicine, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

As a Tools and Resources paper, this study provides a novel set of experimental data on the aspects of yeast aging that are not previously investigated. These datasets include cytosolic pH measurements, cellular crowding levels, and organellar volumes in replicatively aged yeast cells. Reviewers believe that these new datasets will provide an initial framework and important technical resources/methodology that could significantly accelerate studies addressing heterogeneity of the aging process.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "A physicochemical roadmap of yeast replicative aging" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Vyacheslav M Labunskyy (Reviewer #3).

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

Although reviewers have pointed out some novel aspects of this manuscript, especially reporting the pH and crowding measurements in the context of aging, during discussion the reviewers expressed a consensus view that (1) the manuscript in its current form lacks significant novelty for publication in eLife; (2) there are some common major concerns raised by reviewers, especially regarding the rigor and robustness of the measurements due to low cell numbers in many analyses; and (3) as a Tools and Resources paper, the manuscript should include a more comprehensive set of data measurements, especially for the CLEM data. The reviewers concluded that a significant amount of additional work will be needed in order to address all the major concerns raised by the reviewers, which lead to the decision to reject your submission. However, if you feel that these concerns can be effectively addressed in a major revision, we welcome an appeal detailing your plans.

Reviewer #1:

In the manuscript titled "A physicochemical roadmap of yeast replicative aging", authors Mouton et al. measured cytosolic pH, cellular crowding levels, and organellar volume in replicatively aged yeast cells, compared to young ones. Using a novel cellular pH sensor, the authors reported reduced pH in the cytosol in aged cells. Using a set of FRET-based crowding probes, the authors found no apparent change in cellular macromolecular crowding during aging. Using correlative light and electron microscopy, the authors measured the volume of vacuole and nuclei. This analysis showed both increased volume for vacuole and nuclei and increased inter-organellar crowding.

Overall, the authors used novel technologies and measured a number of cellular physiochemical changes during yeast replicative aging, providing a set of experimental data that are not previously available. However, there are a few key weaknesses even when considering this manuscript as a Tools and Resources paper without investigating in the mechanistic details.

1) Several assays only measured a very limited number of cells, especially for the cytosolic pH, where only 30 cells were analyzed. Importantly, as the authors noted, around 50% of these cells were able to maintain a pH in the normal range of young cells. This percentage appears to be even higher for cells that have relatively longer replicative lifespan. Given the broad variation in this phenotype, more cells should be included in this analysis.

2) Vacuole acidity is much reduced in aged cells due to imbalance in proton pumps between vacuolar membranes and plasma membranes (Hughes and Gottschling, 2012). Could this imbalance be the cause of cytosol acidification? Does cytosol acidification and vacuole deacidification concur at the single-cell level during aging? VMA1 overexpression has been shown to at least partially correct the proton pump imbalance resulting in extended lifespan. Does it also ameliorate the cytosol acidification during aging? Many of these questions can be experimentally tested.

3) CLEM analysis is another assay that was only performed for a very small number of cells (14 aged cells and 10 young cells). Although hundreds of section images were analyzed, they still only represent a very few selected cells. Given the general huge variation in cellular morphology, these small numbers are not sufficient to make a convincing case.

4) For the CLEM experiment, the authors mentioned that the cells were aged "over 20 hours to achieve an estimated replicative age of 13 divisions". Since these cells were imaged by fluorescent microscopy, would it be possible to precisely determine their replicative age by staining and counting bud scars? It seems to make sense to count bud scars for these 14 old cells and 10 young cells.

5) The authors stated that "14 tomograms were acquired and segmented to define the plasma membrane, nucleus, vacuole, lipid droplets, ER, multivesicular bodies, and mitochondria". Since these fine cellular structures, other than vacuole and nucleus, are clearly defined, why didn't the authors measure and compare them? Age-associated morphological changes to plasma membrane, lipid droplets, ER, vesicles, and mitochondria at the EM resolution are likely provide more insights to the physiology state of aging cells and support their conclusions based on the analysis of vacuole and nucleus.

Reviewer #2:

In this study, Mouton et al. use longitudinal tracking of single dividing yeast cell to follow the dynamics of several physiological markers using fluorescence microscopy in aging cells. Using the ratiometric probe phluorin, they report a gradual loss of pH in aging cells, that is not predictive of cell death. Then, they use a modified version of a molecular crowding sensor and report that crowding homeostasis is maintained during aging. Last, they perform cellular EM tomographic analyses to monitor the evolution of organelles size with age and they show that inter-organelles crowding increases with age.

Major concerns:

– The message conveyed by the paper is somewhat blurry, knowing that pH measurements do not appear to be conclusive: the authors confirm previous results showing a decrease in pH with age, but they show that it does not predict the evolution of cell fate. Also, they do not provide evidence that this observation is connected to any other well-described physiological marks of age (e.g. protein aggregates, ERCs, etc.). Similarly, they show that molecular crowding does not seem to be a marker of replicative aging. Therefore, even though some observations are novel in the context of aging (measurements of molecular and organelles crowding), the paper is lacking insights of high biological significance regarding the mechanism that controls the entry into replicative senescence.

– One weakness of the paper is that it appears as an assembly of disparate observations that are wrapped as a “physicochemical roadmap”. This term sounds quite obscure, and a bit oversold, or sophistic, knowing that it is only supported by a limited set of readouts that provide mostly inconclusive results.

– pH measurements in aging cells have already performed, leading to the same conclusion that pH decreases with age (using population-based measurements, Kniess et al., 2016, as acknowledged by the authors). However, another study (Henderson et al., 2014, also cited in the present manuscript) also reported measurements of “cortex-proximal cytosolic pH” in aging cells. In that paper, an increase in pH was observed. Whether the cortex-proximal cytosol has an identical pH as the “overall” cytosol is unclear at this point.

Therefore, knowing that the main merit of the paper is to attempt to confirm previous measurements of cytosolic pH, one would expect the authors to clearly address this controversy using in-depth complementary experiments. Instead, interpretations of results are made in such a way to give the impression that the results obtained by the authors is in line with all the literature, which is misleading.

In this context, this statement in the Discussion: “Here, we provide the first (to our knowledge) roadmap for the progressive decline in aging in several parameters defining a cell's intracellular environment, namely, pH, crowding, and volume, all of which impinge on the hallmarks of aging” is inappropriate, knowing that both pH and volume have already been documented in aging cells.

Reviewer #3:

In this manuscript, Mouton et al. have used a combination of single cell imaging and electron microscopy to survey heterogeneity in age-dependent physiological changes, including cytosolic pH, macromolecular crowding, and organellar size, during replicative aging in yeast. This manuscript addresses an important and interesting topic that has not yet been thoroughly investigated by previous studies. The authors made an interesting observation that in replicatively-aged yeast cytosolic pH gradually decreases early in life, but undergoes a significant drop when the cells stop dividing. However, at the single cell level old cells demonstrate incomplete penetrance phenotype, with only ~ 50% of the old cells showing this significant pH drop. Strikingly, despite significant increase in cell volume with age and increased organellar density, which authors refer to as "organellar crowding", macromolecular crowding remained very stable through the cell lifespan. This study is rigorously designed and of good technical quality. Although the exact mechanism for the incomplete penetrance remains unclear, this paper reports an initial framework and important technical resources/methodology that could significantly accelerate studies addressing heterogeneity of the aging process. I would support publication of the manuscript in eLife as Tools and Resources article. However, several points need to be further clarified by the authors.

1) One of the concerns is the low number of cells that were used for analysis of the pH and FRET reporters using microfluidics. Several hundreds of cells can be typically trapped and monitored though their replicative lifespan in the microfluidic device. Were any of the cells excluded from the analysis? If yes, what were the criteria for exclusion? An increase in the number of cells analyzed would help to better assess the proportion of cells that experience a drop in pH and strengthen their conclusions.

2) Another limitation of this study is that time intervals selected to collect fluorescent images for aging experiments (every 10 hours) do not allow assessing short-term physiological fluctuations. For example, does pH change in individual cells between different phases of the cell cycle? This issue needs at least to be discussed.

3) The Introduction section could be improved by logically putting the findings of this paper in the context of the published studies. Does the cell size (organellar volume) affect intracellular (organellar pH)? In several places, authors' statements are not supported by proper citations. E.g.: "In healthy young yeast cells, a "volume hierarchy" is observed where the cytosol represents the largest compartment, followed by the nucleus and vacuole." And "V-ATPase pumps protons from the cytosol into the lumen of various organelles and regulates their pH".

4) Another issue is related to the interpretation of the results. Although the data presented in this manuscript is not sufficient to assess pH changes in different cellular compartments (only in cytosol), authors should discuss whether the loss of vacuolar acidity with aging (Hughes and Gottschling, 2012) can potentially precede and/or be causative for the observed drop in cytosolic pH in old cells. Moreover, the decrease of cytosolic pH with aging (Knieß and Mayer, 2016) seems to contradict with previously reported data showing increased pH in the cell cortex of old mother cells (Henderson et al., 2014). What is the author interpretation of this discrepancy? This issue needs to be discussed in more detail.

5) Ultrastructural analysis of aged cells. Authors state that cells were "grown to an average age of 13" before processing for CLEM. How the age of cells was determined or confirmed?

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for sending your article entitled "A physicochemical perspective of aging from single-cell analysis of pH, macromolecular and organellar crowding in yeast" for peer review at eLife. Your article is being evaluated by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation is being overseen by Jessica Tyler as the Senior Editor.

Reviewer #1:

In this revised manuscript, the authors have addressed numerous concerns raised by reviewers, most notably, the lack of cell numbers in the analysis and the obscure title. The authors also noted additional technical difficulties for addressing all the concerns experimentally but have made appropriate revisions in the text related to these concerns. I feel that the manuscript is in an appropriate state for publication by eLife.

Reviewer #2:

I understand that is a revised version of the manuscript that I evaluated in February 2020. The manuscript was rejected based on the consensus of all three reviewers and it seems that the authors made an appeal to submit a revised version.

Please find below my new detailed assessment of the new manuscript.

This revised version features an extensive rewriting of the text as well as an additional experiment that is reported in Figure 1E.

I thank the reviewer for taking our comments into account by rewording extensive parts of the manuscript. Specifically, the discussion about how their data fit in with the existing literature is better than in the original version.

However, not surprisingly, the rewriting somehow tones down the significance of the results. It also highlights further the necessity to perform additional experiments, as was requested in the original review. We regret that the major concerns have not been addressed experimentally, as requested. Instead, the authors have argued that these concerns were irrelevant or fell outside of the scope of the paper. Therefore, we still have major reservations regarding the insights of this study, knowing that overall these results lead to weak conclusions and/or demonstrate that pH and macromolecular crowding are (1) likely to not to play any causative in the aging process, (2) do not appear as strong hallmarks of physiological impairments in aging, unlike other well described biological processes, and (3) require further experiments (see below) to validate the existing data sets.

The only notable exception concerns Figure 1, in which the addition of longitudinal single cell analysis with higher sampling rate provides some interesting complementary results that refine previous measurements. I think that expanding this kind of analysis, which truly exploits the capabilities of this single-cell approach would be the way to go to make this study more convincing, provided that pH/crowding are indeed demonstrated to play a role in aging. But this would require extensive additional analyses.

Last, the Introduction now suffers quite a bit from sporadic edits added during the revision, and that makes it difficult to read by now, due to number of redundancies (see specific point below).

Specific points:

1) pH measurements

– In Figure 1C, cells are grouped according to their replicative lifespans. Why is such binning achieved, since this is not exploited further (e.g. representation of the average pH drop in each group as a histogram)? Does it make sense to discriminate between 10-15 and 15-20 for instance, knowing that cells don't seem to display any different behavior?

Knowing that the average RLS of WT cells in the literature is around 25 divisions, it is surprising that there is only one bin for cells that do more than 25 divisions. Related to this, no RLS curve is displayed as a control (e.g. as supplementary figure) that shows the distribution of lifespans to prove that cells are doing well in the microfluidic device. The authors claim that cells are undergoing phototoxic damages in the experiment in Figure 1E, therefore we need to know how physiological these experiments are. This problem encountered by the authors is surprising, knowing that many other groups have reported lifespans assays using microfluidic devices that are comparable to the well-established microdissection assays.

– “We observed a gradual decrease in pH already early in life in almost all cells, and interestingly in a fraction of the cells this gradual decrease is followed by a substantial drop in pH in the subpopulation of cells that stop dividing and enter senescence (Figure 1D).”

Since the plot shows pH as a function of the number of divisions, it is impossible to assess whether pH drops substantially after cells stop dividing, unless this assessment corresponds to the very last point on each single cell trace. If so, this should be properly quantified, because it seems to only be apparent in a minority of the cells. Unlike other observations in this figure, this one is not quantified.

– “We conclude that, apart from this subpopulation of young cells with low starting pH, a shared phenotype of all aging yeast cells is that the cytosolic pH drops gradually and modestly throughout the mitotic lifespan, and that when cells stop dividing but remain alive, the pH decreases steeply.” (Emphasis added.)

This is neither correctly reported (see comment above) nor quantified (by averaging over a number of cells). As is, this result looks anecdotal, yet this is the main conclusion of the single cell longitudinal approach (the other result, namely, that pH drops modestly, is a recap of Kniess and Mayer, 2016).

– Figure 1—figure supplement 1C: Y label is incorrect “Fold change at age 15+/-2”. pH fold change?

If we assume that there is indeed a linear model that links pH fold change to RLS, the fit indicates that there is 5% change in pH to be expected between short lived (~10 div) and long lived (~30 div) cells. This suggests that pH is not a parameter here (hence a low correlation coefficient).

– Figure 1E: “Cytosolic acidity” is not defined. The legend says: "Normalized ratiometric pHluorin read-out (F475/F390, top) and replicative age (bottom) as a function of time for a cell that enters senescence at 26 hours (dashed blue line) after which a sharp acidification of the cytosol (dashed red line) follows."

It is normalized pH? If so, then it should be displayed as pH fold change or absolute pH values (in which case the curve would go down and not up).

Alternatively, does a doubling in acidity correspond to a 0.3 pH units decrease (as one would expect with a log10 scale)? This readout should be consistent with other pH measurements.

– Overall, it seems that pH is drop is posterior to the entry into a post-mitotic state. If true (even if partially), drawing such conclusion would be more specific than: “there is a direct link between the time spent in a slow-dividing or post-mitotic state and”, because it would imply some causality or at least a temporal ordering of events. Such temporal order could not only be done with the 36% of the cells that do arrest their cell cycle , but also with those that only slow down (most cells seem to slow down in the 3 subpopulations of cells displayed on Figure 1—figure supplement 3).

Figure 1E displays a quite sharp increase in acidity that starts after the cell cycle arrest, whereas 1F show a progressive decrease in pH. Therefore the conclusion from Figure 1 is confusing: is the evolution of pH progressive or sharp during the lifespan? As is, it is impossible to have an opinion. Since the merit of this paper it to do a longitudinal analysis, it would necessary to use a data analysis framework that clarifies this apparent discrepancy.

– Results paragraph six: 2 questions are raised but not sorted out in the Results section. Those points should be raised in the Discussion if they are not experimentally addressed in the Results section.

– The fact that pH drops in aging cells does not mean that pH homeostasis is abolished. By definition, pH homeostasis is abolished if cytosolic pH is and medium pH are equal. Here, the set point of the homeostatic system may be changed and pH homeostasis still be functional. This may sound a bit semantic, but there are many situations in biology in which it is the set point that is affected, not the homeostatic system itself, and the experiments in this manuscript do not allow to distinguish between the two. Therefore, the statement should considerably toned down, especially knowing that pH only drops by 0.5 units.

2) "Crowding" measurements:

"Plotting single-cell trajectories for cells that reach a replicative lifespan of 10, 10-15, 15-20, 20-25, or larger than 25 shows that the shortest-lived cells tend to increase the crowding levels during their lifespan, while the longer-lived cells tend to have more stable crowding levels (Figure 3C)."

This statement is far from obvious when looking at single cell data. Figure 3D and 3E show that, even if statistically significant, this effect is tiny and probably irrelevant biologically: for instance, it would be impossible to predict cell replicative age or cell lifespan based on a given NFRET measurement. Therefore, I think this analysis is distracting or even useless.

"It seems that it is the maintenance of crowding homeostasis, rather than the absolute crowding levels, which has an association with lifespan"

As currently measured, the magnitude of crowding only varies by a few % throughout the lifespan. I think it is misleading to give the impression that crowding homeostasis may be impaired at all during aging. There are so many other biological processes that have reported to be massively impaired and directly connected to aging in yeast that this analysis sounds very anecdotal in comparison. A clear statement that crowding is not affected during aging would be much more clear-cut and fairer to the actual data.

In the manuscript, the authors report a change in pH from 7.5 and 7 throughout the lifespan (Figure 1). In Munder et al., 2016, decreasing the pH down to 6 or below is required to induce a transition to a gel-like cytoplasmic state. It is therefore not surprising that the authors do not observe any change in crowding.

Also, it would have been relevant to use a complementary marker of crowding , similar to what other people have used in this field (e.g. mobility measurements using microNS-GFP, Gln1-GFP, etc.) , in order to check the results obtained with the FRET sensor, especially knowing that the obtained results are mostly negative. In particular, these complementary measurements would somehow make the link between the macromolecular crowding and organelle crowding which lead to opposing conclusions in the present study.

By the way, “crowding” somehow refers to a functional property: its measurement allows one to assess whether diffusive and transport processes are impaired. Here, the authors use a slightly different meaning, for instance by measuring distance between organelles. One may question whether a decrease in organelles distance necessarily leads to functional impairments. This would need to be assessed using mobility assays (as mentioned above).

3) Introduction issues/redundancies:

The Introduction lacks structure: background information are mixed with justification of the choices made to study pH/crowding in the study. There are several redundancies in the Introduction that make it confusing and superficial, because the same arguments/information are repeated over and over.

– “we select crowding and pH”

Macromolecular crowding should be precisely defined and justified with relevant references, especially if the paper is intended to be read in part by aging experts.

– Introduction paragraph three: using both “alkalization” and “acidification” makes this paragraph quite confusing. Specifically, since this paragraph somehow points out the controversy about the evolution of pH during replicative aging, instead of: “the pH needs to be considered when defining a physicochemical roadmap of cellular aging” , I would rather write that measurements of pH during aging lead to opposite conclusions, therefore, additional data are required to sort out the controversy.

“what is currently missing is a single cell perspective on cytosolic pH in yeast replicative ageing”: this statement still lacks justification: why is it essential to perform complementary measurements? This is not explained… Instead, the authors the authors use the word “physicochemical roadmap” which does not explain the underlying controversy.

These 3 sentences are redundant:

“both properties, in addition to affecting the function of individual key molecules, also interplay to impact intracellular organization”

“pH have the potential to drive profound changes to intracellular organization (reference to Munder et al., 2016)”

“The pH also influences macromolecular organization in yeast:” (reference to Munder et al., 2016 again…)

4) Discussion

I think the additional discussion is fair, i.e. the fact that the modest cytosol acidification is likely to be a consequence of upstream impaired biological processes.

“We show that cytosolic acidity strongly increases only after entry into senescence and we do not observe drastic changes in early lifespan.” I'm ok with this, yet Figure 1F still yields an opposite conclusion. This would have to be clarified further (see point 1) above).

5) Comments on the rebuttal letter

In the Highlights section:

"the finding that crowding on the scale of macromolecules is well maintained in ageing implying that is must be strictly regulated”

I'm afraid that this is a purely rhetorical argument, that does not provide a very strong argument in favor of the paper. The fact that X or Y is not impaired during aging does unfortunately not help understand the mechanisms driving aging.

"the first single cell time course measurements of cytosolic pH in ageing revealing that the timing and direction of pH changes in cytosol, vacuoles, and cortical pH are distinct"

This is misleading because the paper does not properly address this question: only cytosolic measurements are performed in this study, thus a direct comparison to vacuolar and cortical pH is impossible based on the proposed datasets. It is only through a comparison to previous literature, which makes this conclusion much less strong.

Therefore, the controversy about vacuolar (apparently an early-life pH increase) versus cytosolic (presumably a late-life pH decrease) pH dynamics cannot be addressed based on the data reported in this manuscript. This means that point #2 from reviewer #1, point #3 from reviewer #3, and point #4 from reviewer #3 are left unaddressed.

Regarding the link between either pH or crowding and aging, the magnitude of effects observed in young versus old cells is very small and tend to support a model in which these physicochemical parameters are either not involved in the entry into senescence, or enslaved to upstream biological processes that truly affect cellular fitness (monitored as cell cycle arrest/slow down). Therefore, knowing that there is not technical or conceptual breakthrough in this paper, one may really question whether it adds any relevant insight towards a better understanding of the aging process.

"Organellar crowding is strongly increased. We highlighted the impact on diffusion times of cellular structures such as ribosomes and RNP granules, and the possible increase in membrane contact sites. These are important findings"

The argument about diffusion times is purely theoretical and would require an experimental validation (cf. comment about microNS-GFP above) to make the point.

Reviewer #3:

All my concerns have been addressed by the authors in the revised manuscript. I support publication in its current form.

https://doi.org/10.7554/eLife.54707.sa1

Author response

[Editors’ note: The authors appealed the original decision. What follows is the authors’ response to the first round of review.]

Thank you for reviewing our work now entitled "A physicochemical perspective of aging from singlecell analysis of pH, macromolecular and organellar crowding in yeast” and the invitation to submit our manuscript for revision. We have carefully considered the valuable comments by the reviewers. The possibilities to perform experiments are currently limited due to the global pandemic, but we have managed to perform all experiments that contribute to the thoroughness of the publication, and have, where appropriate, adjusted the text to account for the remaining concerns. We hope you agree that the manuscript in its current form is suitable for the special edition on aging in eLife. The highlights are:

– a valuable high resolution EM dataset showing for the first time that organellar crowding increases dramatically in ageing,

– the finding that crowding on the scale of macromolecules is well maintained in ageing implying that is must be strictly regulated,

– a newly developed crowding sensor that will aid to dissect the above mechanism in future studies,

– the first single cell time course measurements of cytosolic pH in ageing revealing that the timing and direction of pH changes in cytosol, vacuoles, and cortical pH are distinct and pointing out a possible role in pH homeostasis for the proteome.

Reviewer #1:

In the manuscript titled "A physicochemical roadmap of yeast replicative aging", authors Mouton et al. measured cytosolic pH, cellular crowding levels, and organellar volume in replicatively aged yeast cells, compared to young ones. Using a novel cellular pH sensor, the authors reported reduced pH in the cytosol in aged cells. Using a set of FRET-based crowding probes, the authors found no apparent change in cellular macromolecular crowding during aging. Using correlative light and electron microscopy, the authors measured the volume of vacuole and nuclei. This analysis showed both increased volume for vacuole and nuclei and increased inter-organellar crowding.

Overall, the authors used novel technologies and measured a number of cellular physiochemical changes during yeast replicative aging, providing a set of experimental data that are not previously available. However, there are a few key weaknesses even when considering this manuscript as a resources paper without investigating in the mechanistic details.

1) Several assays only measured a very limited number of cells, especially for the cytosolic pH, where only 30 cells were analyzed. Importantly, as the authors noted, around 50% of these cells were able to maintain a pH in the normal range of young cells. This percentage appears to be even higher for cells that have relatively longer replicative lifespan. Given the broad variation in this phenotype, more cells should be included in this analysis.

We agree that the numbers of cells analyzed were somewhat on the low side and have increased the number to 80 for the single cell pH and the crowding measurements (adjusted Figures 1C,D,F and 3B-E and Figure 1—figure supplement 1 and Figure 3—figure supplement 1). Increasing the dataset has not changed the numbers such as population averages, means, the statistics and correlations nor the conclusions drawn.

2) Vacuole acidity is much reduced in aged cells due to imbalance in proton pumps between vacuolar membranes and plasma membranes (Hughes and Gottschling, 2012). Could this imbalance be the cause of cytosol acidification? Does cytosol acidification and vacuole deacidification concur at the single-cell level during aging? VMA1 overexpression has been shown to at least partially correct the proton pump imbalance resulting in extended lifespan. Does it also ameliorate the cytosol acidification during aging? Many of these questions can be experimentally tested.

We agree and we have now addressed these issues better. We included new data to elucidate the precise timing of the cytosolic changes (Figure 1E and Figure 1—figure supplement 2 and 3 and text in subsection “Yeast replicative aging leads to acidification of the cytosol, especially after entry into senescence”) and we discuss on how this related to those in the vacuole and cell cortex (subsection “pH homeostasis in aging”). Whether one should expect that VMA1 overexpression would be sufficient to ameliorate the cytosol acidification is not so obvious to us considering that the timing of pH changes is rather different in cytosol and vacuole. While it would indeed be an interesting experiment, a practical concern is that our microfluidic chips are designed for haploid cells while the overexpression of VMA1 should be done in diploids as the haploids are sick and grow very slow.

3) CLEM analysis is another assay that was only performed for a very small number of cells (14 aged cells and 10 young cells). Although hundreds of section images were analyzed, they still only represent a very few selected cells. Given the general huge variation in cellular morphology, these small numbers are not sufficient to make a convincing case.

We respectfully disagree that the presented data are insufficient to make our point, which is, that organellar crowding increases dramatically in aged cells. This is a new finding that can only be obtained from the kind of (low throughput) EM data presented here. The cells chosen for analysis were selected randomly and the effects are statistically significant and unambiguous. While we agree that there is a high degree of cell-to-cell variation for some age-associated phenotypes which requires larger cohort sizes, for this particular phenotype the effect size is sufficiently large and the variation is sufficiently small that it is easily detectable with the number of cells analyzed here.

The CLEM analysis of aged cells is very challenging and not high throughput as aged cells make up only a minor fraction of the exponentially growing population. At the 20 hours time point used here, each aged (and labeled) cell is outnumbered by thousands of daughter cells. In the current experimental set, we chose to age the yeast cells in normal exponential growing cultures. This setup has minimal interventions (no attachments of beads for enriching aged cell, no genetic programs to enrich for aged cells, etc.) and hence we can accurately and uniquely report the ultrastructure of ageing cells.

4) For the CLEM experiment, the authors mentioned that the cells were aged "over 20 hours to achieve an estimated replicative age of 13 divisions". Since these cells were imaged by fluorescent microscopy, would it be possible to precisely determine their replicative age by staining and counting bud scars? It seems to make sense to count bud scars for these 14 old cells and 10 young cells.

This is unfortunately not possible. In our previous work we addressed the spread in the ages of cells cultured for 20 hours: we had modelled (by considering the distribution of division times) and experimentally determined (performing bud scar counts) the age-distributions and find that the ages in the population are approximately normally distributed with a half maximum peak width of 10 divisions (Janssens et al., 2015, Figure 1—figure supplement 2). These studies have shown us that we can estimate the age distribution at 20 hours based on the doubling time of the cells. The current studies were performed in rich medium where the doubling times are faster, hence the average of 13. Also, in the microfluidic setup we observe a similar distributing of ages after 20 hours.

In the current text we have made the uncertainty of the exact ages of the represented cells explicit in the Results section.

5) The authors stated that "14 tomograms were acquired and segmented to define the plasma membrane, nucleus, vacuole, lipid droplets, ER, multivesicular bodies, and mitochondria". Since these fine cellular structures, other than vacuole and nucleus, are clearly defined, why didn't the authors measure and compare them? Age-associated morphological changes to plasma membrane, lipid droplets, ER, vesicles, and mitochondria at the EM resolution are likely provide more insights to the physiology state of aging cells and support their conclusions based on the analysis of vacuole and nucleus.

The importance of simultaneous visualization of all organellar membranes enabled the quantification of the distances between organelles, and hence the organellar crowding. Quantification of individual features like surface area or volume of the ER, Mitochondria and MVB’s is not possible considering that we could quantify only few of each type. We did include an analysis of the number, and surface area of lipid droplets, showing lipid droplets increase in ageing (Figure 4—figure supplement 3).

Reviewer #2:

In this study, Mouton et al. use longitudinal tracking of single dividing yeast cell to follow

the dynamics of several physiological markers using fluorescence microscopy in aging cells. Using the ratiometric probe phluorin, they report a gradual loss of pH in aging cells, that is not predictive of cell death. Then, they use a modified version of a molecular crowding sensor and report that crowding homeostasis is maintained during aging. Last, they perform cellular EM tomographic analyses to monitor the evolution of organelles size with age and they show that inter-organelles crowding increases with age.

Major concerns:

– The message conveyed by the paper is somewhat blurry, knowing that pH measurements do not appear to be conclusive: the authors confirm previous results showing a decrease in pH with age, but they show that it does not predict the evolution of cell fate. Also, they do not provide evidence that this observation is connected to any other well-described physiological marks of age (e.g. protein aggregates, ERCs, etc.). Similarly, they show that molecular crowding does not seem to be a marker of replicative aging. Therefore, even though some observations are novel in the context of aging (measurements of molecular and organelles crowding), the paper is lacking insights of high biological significance regarding the mechanism that controls the entry into replicative senescence.

We are sorry to read we have not conveyed the novelty of our paper well enough. We have made several textual changes to explain this better in the current manuscript, namely:

i) The single-cell data showing the relationship between cytosolic pH and replicative age is new and not merely confirmatory to previous data which were based on population-level data. It is this single-cell perspective that reveals several aspects that remain invisible on the population levels (the timing of changes, the variation within the population, correlations with lifespan). The novelty and implications of our cytosolic pH data is now better discussed in subsection “pH homeostasis in aging”.

ii) Macromolecular crowding is well maintained throughout ageing and future studies should be aimed at finding what regulates this homeostasis. This is completely uncharted territory and outside the scope of this paper. The newly developed crowding sensor will provide a valuable tool identifying the genes that regulate crowding

iii) Organellar crowding is strongly increased. We highlighted the impact on diffusion times of cellular structures such as ribosomes and RNP granules, and the possible increase in membrane contact sites. These are important findings, which imply that out of the available methods to study crowding, which are tracer particles and crowding sensors, results obtained through tracer particles have to be carefully interpreted, because of confinement artifacts. These observations also warrant future investigations that fall outside the scope of this paper.

We respectfully disagree that additional evidence is needed showing that homeostasis of pH and crowding are connected to other well-described physiological marks of age (e.g. protein aggregates); we reference many papers that establish how volume regulation, pH homeostasis and crowding impact biology. We consider establishing how they are connected beyond the scope of this paper.

– One weakness of the paper is that it appears as an assembly of disparate observations that are wrapped as a “physicochemical roadmap”. This term sounds quite obscure, and a bit oversold, or sophistic, knowing that it is only supported by a limited set of readouts that provide mostly inconclusive results.

Our revised manuscript highlights the many clear connections between the separate observations throughout the text; the observations are intimately related with each other. However, we do not want our title to sound obscure or come across as oversold and thus have adjusted it to the more precise “A physicochemical perspective of aging from single-cell analysis of pH, macromolecular and organellar crowding in yeast”.

– pH measurements in aging cells have already performed, leading to the same conclusion that pH decreases with age (using population-based measurements, Kniess et al., 2016, as acknowledged by the authors). However, another study (Henderson et al., 2014, also cited in the present manuscript) also reported measurements of “cortex-proximal cytosolic pH” in aging cells. In that paper, an increase in pH was observed. Whether the cortex-proximal cytosol has an identical pH as the “overall” cytosol is unclear at this point.

Therefore, knowing that the main merit of the paper is to attempt to confirm previous measurements of cytosolic pH, one would expect the authors to clearly address this controversy using in-depth complementary experiments. Instead, interpretations of results are made in such a way to give the impression that the results obtained by the authors is in line with all the literature, which is misleading.

We realize we might have left a wrong impression for the merits of our manuscript. The aim of our work was to show the different physicochemical environment that aged cytosol of yeast cells have compared to young cells (now stated in the Introduction). We did not aim to resolve long standing questions in the field regarding pH of different compartments and we apologize if we did not express our intentions better. In order to reliably measure crowding in ageing cells we needed single-cell data on the evolution of the cytosolic pH, which was not yet available.

We have now adjusted our Discussion to address better how our findings relate to previously published pH measurements at the cell cortex. Please see our answers to reviewer 1 point 2.

In this context, this statement in the Discussion: “Here, we provide the first (to our knowledge) roadmap for the progressive decline in aging in several parameters defining a cell's intracellular environment, namely, pH, crowding, and volume, all of which impinge on the hallmarks of aging” is inappropriate, knowing that both pH and volume have already been documented in aging cells.

We have removed the sentence and replaced it with the more precise “Here, we provide an analysis of the progressive change during aging for several parameters that define a cell’s intracellular environment; namely, cytosolic pH and crowding on the scale of macromolecules and organelles, all impinging on the hallmarks of aging.”

Reviewer #3:

In this manuscript, Mouton et al. have used a combination of single cell imaging and electron microscopy to survey heterogeneity in age-dependent physiological changes, including cytosolic pH, macromolecular crowding, and organellar size, during replicative aging in yeast. This manuscript addresses an important and interesting topic that has not yet been thoroughly investigated by previous studies. The authors made an interesting observation that in replicatively-aged yeast cytosolic pH gradually decreases early in life, but undergoes a significant drop when the cells stop dividing. However, at the single cell level old cells demonstrate incomplete penetrance phenotype, with only ~ 50% of the old cells showing this significant pH drop. Strikingly, despite significant increase in cell volume with age and increased organellar density, which authors refer to as "organellar crowding", macromolecular crowding remained very stable through the cell lifespan. This study is rigorously designed and of good technical quality. Although the exact mechanism for the incomplete penetrance remains unclear, this paper reports an initial framework and important technical resources/methodology that could significantly accelerate studies addressing heterogeneity of the aging process. I would support publication of the manuscript in eLife as Tools and Resources. However, several points need to be further clarified by the authors.

1) One of the concerns is the low number of cells that were used for analysis of the pH and FRET reporters using microfluidics. Several hundreds of cells can be typically trapped and monitored though their replicative lifespan in the microfluidic device. Were any of the cells excluded from the analysis? If yes, what were the criteria for exclusion? An increase in the number of cells analyzed would help to better assess the proportion of cells that experience a drop in pH and strengthen their conclusions.

Indeed, many cells can be trapped in the device, although the absolute numbers vary per experiment. However, the number of cells that remain in the device from the start until their death is significantly lower compared to the total number of trapped cells. At each imaging position we have included all cells that are imaged from start to death and have not excluded cells from the analysis.

As mentioned above, we have increased the numbers of cells analyzed for both cytosolic pH and macromolecular crowding to 80 cells. Further increase in these numbers seems unnecessary, since none of our conclusions or their significance have changed after the additional analysis.

2) Another limitation of this study is that time intervals selected to collect fluorescent images for aging experiments (every 10 hours) do not allow assessing short-term physiological fluctuations. For example, does pH change in individual cells between different phases of the cell cycle? This issue needs at least to be discussed.

Imaging at higher time density runs the risk of increased phototoxicity and as our aim was to follow age-related phenotypes in pH, we chose to not measure short-term fluctuations in physiology. We have now generated a new dataset in which we follow cytosolic pH every hour in order to address the timing of pH drop better (see Figure 1E and Figure 1—figure supplement 2 and 3 and Supplementary file 3 confirming the mild impact of higher frequency imaging on lifespan). We agree that the issue of pH changes in the cell cycle should have been discussed and we now do so in Results paragraph four. The new data shows that variations in pH as function of the cell cycle are minor compared to those in aging.

3) The Introduction section could be improved by logically putting the findings of this paper in the context of the published studies. Does the cell size (organellar volume) affect intracellular (organellar pH)? In several places, authors' statements are not supported by proper citations. E.g.: "In healthy young yeast cells, a "volume hierarchy" is observed where the cytosol represents the largest compartment, followed by the nucleus and vacuole." And "V-ATPase pumps protons from the cytosol into the lumen of various organelles and regulates their pH".

Thank you for pointing this out. We have added the references and clarified connections to the existing literature in the revised Introduction and Discussion.

4) Another issue is related to the interpretation of the results. Although the data presented in this manuscript is not sufficient to assess pH changes in different cellular compartments (only in cytosol), authors should discuss whether the loss of vacuolar acidity with aging (Hughes and Gottschling, 2012) can potentially precede and/or be causative for the observed drop in cytosolic pH in old cells. Moreover, the decrease of cytosolic pH with aging (Knieß and Mayer, 2016) seems to contradict with previously reported data showing increased pH in the cell cortex of old mother cells (Henderson et al., 2014). What is the author interpretation of this discrepancy? This issue needs to be discussed in more detail.

We have indeed added a discussion related to these previous findings. Please also refer to our answer to reviewer 1 point 2.

5) Ultrastructural analysis of aged cells. Authors state that cells were "grown to an average age of 13" before processing for CLEM. How the age of cells was determined or confirmed?

Please see our answer to reviewer 1 point 4.

[Editors’ note: what follows is the authors’ response to the second round of review.]

Reviewer #2:

[…]

Specific points:

1) pH measurements

– In Figure 1C, cells are grouped according to their replicative lifespans. Why is such binning achieved, since this is not exploited further (e.g. representation of the average pH drop in each group as a histogram)? Does it make sense to discriminate between 10-15 and 15-20 for instance, knowing that cells don't seem to display any different behaviour ?

Knowing that the average RLS of WT cells in the literature is around 25 divisions, it is surprising that there is only one bin for cells that do more than 25 divisions. Related to this, no RLS curve is displayed as a control (e.g. as supplementary figure) that shows the distribution of lifespans to prove that cells are doing well in the microfluidic device. The authors claim that cells are undergoing phototoxic damages in the experiment in Figure 1E, therefore we need to know how physiological these experiments are. This problem encountered by the authors is surprising, knowing that many other groups have reported lifespans assays using microfluidic devices that are comparable to the well-established microdissection assays.

– “We observed a gradual decrease in pH already early in life in almost all cells, and interestingly in a fraction of the cells this gradual decrease is followed by a substantial drop in pH in the subpopulation of cells that stop dividing and enter senescence (Figure 1D).”

Since the plot show pH as a function of the number of divisions, it is impossible to assess whether pH drops substantially after cells stop dividing, unless this assessment corresponds to the very last point on each single cell trace. If so, this should be properly quantified, because it seems to only be apparent in a minority of the cells. Unlike other observations in this figure, this one is not quantified.

– ““We conclude that, apart from this subpopulation of young cells with low starting pH, a shared phenotype of all aging yeast cells is that the cytosolic pH drops gradually and modestly throughout the mitotic lifespan, and that when cells stop dividing but remain alive, the pH decreases steeply.” (Emphasis added)

This is neither correctly reported (see comment above) nor quantified (by averaging over a number of cells). As is, this result looks anecdotal, yet this is the main conclusion of the single cell longitudinal approach (the other result, namely, that pH drops modestly, is a recap of Kniess and Mayer, 2016).

– Figure 1—figure supplement 1C: Y label is incorrect “Fold change at age 15+/-2”. pH fold change?

If we assume that there is indeed a linear model that links pH fold change to RLS, the fit indicates that there is 5% change in pH to be expected between short lived (~10 divv) and long lived (~30 div) cells. This suggests that pH is not a parameter here (hence a low correlation coefficient).

– Figure 1E: “Cytosolic acidity” is not defined. The legend says: "Normalized ratiometric pHluorin read-out (F475/F390, top) and replicative age (bottom) as a function of time for a cell that enters senescence at 26 hours (dashed blue line) after which a sharp acidification of the cytosol (dashed red line) follows."

It is normalized pH? If so, then it should be displayed as pH fold change or absolute pH values (in which case the curve would go down and not up).

Alternatively, does a doubling in acidity correspond to a 0.3 pH units decrease (as one would expect with a log10 scale)? This readout should be consistent with other pH measurements.

– Overall, it seems that pH is drop is posterior to the entry into a post-mitotic state. If true (even if partially), drawing such conclusion would be more specific than: “there is a direct link between the time spent in a slow-dividing or post-mitotic state and”, because it would imply some causality or at least a temporal ordering of events. Such temporal order could not only be done with the 36% of the cells that do arrest their cell cycle , but also with those that only slow down (most cells seem to slow down in the 3 subpopulations of cells displayed on Figure 1—figure supplement 3).

Figure 1E displays a quite sharp increase in acidity that starts after the cell cycle arrest, whereas 1F show a progressive decrease in pH. Therefore the conclusion from Figure 1 is confusing: is the evolution of pH progressive or sharp during the lifespan? As is, it is impossible to have an opinion. Since the merit of this paper it to do a longitudinal analysis, it would necessary to use a data analysis framework that clarifies this apparent discrepancy.

– Results paragraph six: 2 questions are raised but not sorted out in the Results section. Those points should be raised in the Discussion if they are not experimentally addressed in the Results section.

– The fact that pH drops in aging cells does not mean that pH homeostasis is abolished. By definition, pH homeostasis is abolished if cytosolic pH is and medium pH are equal. Here, the set point of the homeostatic system may be changed and pH homeostasis still be functional. This may sound a bit semantic, but there are many situations in biology in which it is the set point that is affected, not the homeostatic system itself, and the experiments in this manuscript do not allow to distinguish between the two. Therefore, the statement should considerably toned down, especially knowing that pH only drops by 0.5 units.

2) "Crowding" measurements:

"Plotting single-cell trajectories for cells that reach a replicative lifespan of 10, 10-15, 15-20, 20-25, or larger than 25 shows that the shortest-lived cells tend to increase the crowding levels during their lifespan, while the longer-lived cells tend to have more stable crowding levels (Figure 3C)."

This statement is far from obvious when looking at single cell data. Figure 3D and 3E show that, even if statistically significant, this effect is tiny and probably irrelevant biologically: for instance, it would be impossible to predict cell replicative age or cell lifespan based on a given NFRET measurement. Therefore, I think this analysis is distracting or even useless.

"It seems that it is the maintenance of crowding homeostasis, rather than the absolute crowding levels, which has an association with lifespan"

As currently measured, the magnitude of crowding only varies by a few % throughout the lifespan. I think it is misleading to give the impression that crowding homeostasis may be impaired at all during aging. There are so many other biological processes that have reported to be massively impaired and directly connected to aging in yeast that this analysis sounds very anecdotal in comparison. A clear statement that crowding is not affected during aging would be much more clear-cut and fairer to the actual data.

In the manuscript, the authors report a change in pH from 7.5 and 7 throughout the lifespan (Figure 1). In Munder et al., 2016, decreasing the pH down to 6 or below is required to induce a transition to a gel-like cytoplasmic state. It is therefore not surprising that the authors do not observe any change in crowding.

Also, it would have been relevant to use a complementary marker of crowding , similar to what other people have used in this field (e.g. mobility measurements using microNS-GFP, Gln1-GFP, etc.) , in order to check the results obtained with the FRET sensor, especially knowing that the obtained results are mostly negative. In particular, these complementary measurements would somehow make the link between the macromolecular crowding and organelle crowding which lead to opposing conclusions in the present study.

By the way, “crowding” somehow refers to a functional property: its measurement allows one to assess whether diffusive and transport processes are impaired. Here, the authors use a slightly different meaning, for instance by measuring distance between organelles. One may question whether a decrease in organelles distance necessarily leads to functional impairments. This would need to be assessed using mobility assays (as mentioned above).

3) Introduction issues/redundancies:

The Introduction lacks structure: background information are mixed with justification of the choices made to study pH/crowding in the study. There are several redundancies in the Introduction that make it confusing and superficial, because the same arguments/information are repeated over and over.

– “we select crowding and pH”

Macromolecular crowding should be precisely defined and justified with relevant references, especially if the paper is intended to be read in part by aging experts

– Introduction paragraph three: using both “alkalization” and “acidification” makes this paragraph quite confusing. Specifically, since this paragraph somehow points out the controversy about the evolution of pH during replicative aging, instead of: “the pH needs to be considered when defining a physicochemical roadmap of cellular aging”, I would rather write that measurements of pH during aging lead to opposite conclusions, therefore, additional data are required to sort out the controversy.

“what is currently missing is a single cell perspective on cytosolic pH in yeast replicative ageing”: this statement still lacks justification: why is it essential to perform complementary measurements ? This is not explained… Instead, the authors the authors use the word “physicochemical roadmap” which does not explain the underlying controversy.

These 3 sentences are redundant:

“both properties, in addition to affecting the function of individual key molecules, also interplay to impact intracellular organization”

“pH have the potential to drive profound changes to intracellular organization (reference to Munder et al., 2016)”

“The pH also influences macromolecular organization in yeast:” (reference to Munder et al., 2016 again…)

4) Discussion

I think the additional discussion is fair, i.e. the fact that the modest cytosol acidification is likely to be a consequence of upstream impaired biological processes.

“We show that cytosolic acidity strongly increases only after entry into senescence and we do not observe drastic changes in early lifespan.” I'm ok with this, yet Figure 1F still yields an opposite conclusion. This would have to be clarified further (see point 1) above).

5) Comments on the rebuttal letter

In the Highlights section:

"the finding that crowding on the scale of macromolecules is well maintained in ageing implying that is must be strictly regulated”

I'm afraid that this is a purely rhetorical argument, that does not provide a very strong argument in favor of the paper. The fact that X or Y is not impaired during aging does unfortunately not help understand the mechanisms driving aging. IN addition, there are likely many processes that

"the first single cell time course measurements of cytosolic pH in ageing revealing that the timing and direction of pH changes in cytosol, vacuoles, and cortical pH are distinct"

This is misleading because the paper does not properly address this question: only cytosolic measurements are performed in this study, thus a direct comparison to vacuolar and cortical pH is impossible based on the proposed datasets. It is only through a comparison to previous literature, which makes this conclusion much less strong.

Therefore, the controversy about vacuolar (apparently an early-life pH increase) versus cytosolic (presumably a late-life pH decrease) pH dynamics cannot be addressed based on the data reported in this manuscript. This means that point #2 from reviewer #1, point #3 from reviewer #3, and point #4 from reviewer #3 are left unaddressed.

Regarding the link between either pH or crowding and aging, the magnitude of effects observed in young versus old cells is very small and tend to support a model in which these physicochemical parameters are either not involved in the entry into senescence, or enslaved to upstream biological processes that truly affect cellular fitness (monitored as cell cycle arrest/slow down). Therefore, knowing that there is not technical or conceptual breakthrough in this paper, one may really question whether it adds any relevant insight towards a better understanding of the aging process.

"Organellar crowding is strongly increased. We highlighted the impact on diffusion times of cellular structures such as ribosomes and RNP granules, and the possible increase in membrane contact sites. These are important findings"

The argument about diffusion times is purely theoretical and would require an experimental validation (cf. comment about microNS-GFP above) to make the point.

We are pleased to read that reviewers #1 and #3 have no further concerns and support publication of the manuscript in its current state. After careful reading of the review of reviewer #2 we have made the following changes.

1) As we interpret it, the reviewer suggests in the main text that we obtain even more single-cell data at a higher time resolution (alike the new data in Figure 1E). This is not interesting to do for the measurements of macromolecular crowding as the levels are relatively stable in aging, and, it is not feasible to do so for the organellar measurements as previously explained: CLEM is unsuited for this.

2) Related to Figure 1. The text suggestions have been incorporated. Other comments are explained below:

1) Why is data presented in age groups?

To avoid overcrowded panels, we distribute the data over multiple panels; grouped to RLS is the most logical way.

2) Figure 1E plots as "acidity instead of pH"

This is because the vacuolar data from (Chen et al., 2020) had not been calibrated to pH but only reports the direct read-out of the pH sensor.

3) What is RLS in this chip design?

There is no difference between the replicative lifespan determined in this microfluidic device or by microdissection; the reviewer can find this specific information in Figure 2E, published by (Crane et al., 2014), also confirmed in our hands in (Janssens, et al., 2016; Rempel et al., 2019).

4) It is confusing that Figure 1E displays a sharp increase in acidity, whereas 1F shows a progressive decrease in pH.

In Figure 1E, we show cytosolic acidity on the y-axis as in (Chen et al., 2020), and in Figure 1F, we show average cytosolic pH over the whole population, therefore the appearance of gradual changes.

5) Sharp decrease in pH of post-mitotic cells should be quantified “by averaging over a number of cells”.

In fact, we had already quantified the variable the reviewer is asking for (Figure 1C, specifically, and all raw data provided in Figure 1). Because the decrease in pH before senescence is only minor (as can be seen from Figure 1D), comparing the first and last measurement points of all cells (as done in 1C) is sufficient.

3) Related to Figure 3. We have included a sentence to the Abstract that crowding is rather stable in aging. Other comments are explained below:

1) Figure 3D and 3E show that, even if statistically significant, the relation between crowding and lifespan is tiny and probably irrelevant biologically.

We agree that crowding remains rather stable in aging. We have stated that numerous times throughout the manuscript, but we have added it back again to the Abstract as well. However, we respectfully disagree that the presentation of the details in Figure 3D and E are unimportant. The changes displayed in Figure 3D and E are within 4%. Considering that maximum change of crowding under severe osmotic shock is ~20%, and recovers within minutes, 4% cannot be considered minor or biologically irrelevant.

2) The reviewer suggests complementing the macromolecular crowing measurements and the organellar crowding measurements with measurements of mobility (single-particle tracking with μNS) or self-assembly of Gln1.

μNS particles are 10 times bigger than the crowding sensor, thus probing crowding on a completely different scale than reported in our study. Performing such experiments would not confirm or contradict any of our findings, since crowding effects are size-dependent (Delarue et al., 2018). Self-assembly of Gln1 could be a good complementary experiment, but also here, such experiments would not confirm or contradict any of our findings, but rather probe an additional parameter.

3) Figure 4—figure supplement 4, referred to in the Discussion requires experimental validation with μNS particles.

Given the data presented in Figure 5, the mobility of fluorescent nanoparticles particles must be decreased due to organellar crowding: Such enormous obstacles would hamper diffusion.

4) Textual issues related to the Introduction and Discussion have been addressed.

https://doi.org/10.7554/eLife.54707.sa2

Article and author information

Author details

  1. Sara N Mouton

    European Research Institute for the Biology of Ageing, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
    Contribution
    Conceptualization, Formal analysis, Investigation, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9429-3788
  2. David J Thaller

    Department of Cell Biology, Yale School of Medicine, New Haven, United States
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3577-5562
  3. Matthew M Crane

    Department of Pathology, School of Medicine, University of Washington, Seattle, United States
    Contribution
    Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6234-0954
  4. Irina L Rempel

    European Research Institute for the Biology of Ageing, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
    Contribution
    Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4655-5311
  5. Owen T Terpstra

    European Research Institute for the Biology of Ageing, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
    Contribution
    Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8767-4061
  6. Anton Steen

    European Research Institute for the Biology of Ageing, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
    Contribution
    Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1064-6038
  7. Matt Kaeberlein

    Department of Pathology, School of Medicine, University of Washington, Seattle, United States
    Contribution
    Supervision
    Competing interests
    Reviewing editor, eLife
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1311-3421
  8. C Patrick Lusk

    Department of Cell Biology, Yale School of Medicine, New Haven, United States
    Contribution
    Supervision
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4703-0533
  9. Arnold J Boersma

    DWI-Leibniz Institute for Interactive Materials, Aachen, Germany
    Contribution
    Conceptualization, Supervision, Writing - original draft, Writing - review and editing
    For correspondence
    boersma@dwi.rwth-aachen.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3714-5938
  10. Liesbeth M Veenhoff

    European Research Institute for the Biology of Ageing, University of Groningen, University Medical Center Groningen, Groningen, Netherlands
    Contribution
    Conceptualization, Supervision, Writing - original draft, Writing - review and editing
    For correspondence
    l.m.veenhoff@rug.nl
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0158-4728

Funding

Nederlandse Organisatie voor Wetenschappelijk Onderzoek (737.016.016)

  • Liesbeth M Veenhoff

Nederlandse Organisatie voor Wetenschappelijk Onderzoek (723.015.002)

  • Arnold J Boersma

National Institutes of Health (RO1 GM105672)

  • C Patrick Lusk

National Institutes of Health (P30 AG013280)

  • Matt Kaeberlein

National Institutes of Health (R01 AG056359)

  • Matt Kaeberlein

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We acknowledge The Netherlands Organization for Scientific Research: Vidi grant 723.015.002 to AJB and BBoL grant 737.016.016 to LMV, and the NIH RO1 GM105672 to DJT and CPL, and R01 AG056359, and P30 AG013280 to MK for financial support.

Senior Editor

  1. Jessica K Tyler, Weill Cornell Medicine, United States

Reviewing Editor

  1. Weiwei Dang, Baylor College of Medicine, United States

Reviewer

  1. Vyacheslav M Labunskyy, Boston University School of Medicine, United States

Publication history

  1. Received: December 23, 2019
  2. Accepted: September 28, 2020
  3. Accepted Manuscript published: September 29, 2020 (version 1)
  4. Version of Record published: October 14, 2020 (version 2)

Copyright

© 2020, Mouton et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,015
    Page views
  • 184
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

  1. Further reading

Further reading

    1. Biochemistry and Chemical Biology
    2. Neuroscience
    Damien Lemoine et al.
    Research Article

    Glutamate delta (GluD) receptors belong to the ionotropic glutamate receptor family, yet they don't bind glutamate and are considered orphan. Progress in defining the ion channel function of GluDs in neurons has been hindered by a lack of pharmacological tools. Here we used a chemo-genetic approach to engineer specific and photo-reversible pharmacology in GluD2 receptor. We incorporated a cysteine mutation in the cavity located above the putative ion channel pore, for site-specific conjugation with a photoswitchable pore blocker. In the constitutively-open GluD2 Lurcher mutant, current could be rapidly and reversibly decreased with light. We then transposed the cysteine mutation to the native receptor, to demonstrate with high pharmacological specificity that metabotropic glutamate receptor signaling triggers opening of GluD2. Our results assess the functional relevance of GluD2 ion channel and introduce an optogenetic tool that will provide a novel and powerful means for probing GluD2 ionotropic contribution to neuronal physiology.

    1. Biochemistry and Chemical Biology
    Marija Liutkute et al.
    Research Article

    Nascent polypeptides begin to fold in the constrained space of the ribosomal peptide exit tunnel. Here we use force-profile analysis (FPA) and photo-induced energy-transfer fluorescence correlation spectroscopy (PET-FCS) to show how a small α-helical domain, the N-terminal domain of HemK, folds cotranslationally. Compaction starts vectorially as soon as the first α-helical segments are synthesized. As nascent chain grows, emerging helical segments dock onto each other and continue to rearrange at the vicinity of the ribosome. Inside or in the proximity of the ribosome, the nascent peptide undergoes structural fluctuations on the µs time scale. The fluctuations slow down as the domain moves away from the ribosome. Mutations that destabilize the packing of the domain’s hydrophobic core have little effect on folding within the exit tunnel, but abolish the final domain stabilization. The results show the power of FPA and PET-FCS in solving the trajectory of cotranslational protein folding and in characterizing the dynamic properties of folding intermediates.