The impact of medium acidity on the chronological life span of Saccharomyces cerevisiae – lipids, signaling cascades, mitochondrial and vacuolar functions



Because of its multifactorial nature, aging is one of the most complicated cell phenomena known. A systems biology approach, which aims to understand the organism as a whole rather than concentrate on the behaviors of individual genes, thus comprises a seamless tool for investigating the aging machinery, which arises mainly as a result of degeneration of the collaboration between signaling and regulatory pathways. In the present study, the effects of medium buffering on the chronological life span are investigated via transcriptome analyses and subsequent integration of the data obtained with the chronological aging network of yeast. The comparative inquiry of transcriptome data of young and old cells grown in buffered and unbuffered media reveals new roles for pH control (e.g. the re-organization of lipid metabolism and intracellular signaling cascades) that have beneficial consequences on chronological longevity. Integration of the transcriptome data onto the aging network, as well as validation experiments, suggest that Snf1p is a possible intermediate player in the interjunction of sphingolipid and ergosterol metabolisms with extracellular pH control with respect to regulation of the chronological life span. Consequently, a more detailed insight of the chronological aging mechanism of yeast is obtained. The results of the present study provide a solid basis for further research focusing on uncovering the agents that affect aging and age-related diseases in humans.


atorvastatin calcium salt


analysis of variance


colony-forming units


chronological life span


endoplasmic reticulum


mitogen-activated protein kinase




principal component analysis


pH control-significant gene


reactive oxygen species


synthetic dextrose complete


target of rapamycin


vacuolar ATPase


yeast extract peptone dextrose


The yeast chronological life span (CLS) is described as the proliferation capacity of quiescent yeast cells. Different from the replicative life span, it deals with the resumption of proliferation upon stimulation: quiescent cells are rendered ‘quiescent’ either by genetic manipulations or extracellular conditions such as depleted nutrients, and are in a ‘dormant’ state. Their ‘awakening’ capacity, which interestingly declines in a continuous way with increasing time, determines their so-called ‘CLS’. Accordingly, quiescence is a reversible process: upon stimulation, quiescent cells resume their growth, whereas ‘senescent’ cells cannot. Therefore, the choice between ‘senescence’ and ‘quiescence’ is the key factor governing the fate of a chronologically aged cell. This switch from quiescence to senescence is the reason for investigating that particular type of aging encountered in model organisms such as yeast: the very same switch is experienced in many diseases, one of which is cancer. Therefore, understanding the dynamics leading to a senescent phenotype may also be enlightening for gaining insight about aging and age-related diseases in humans.

One of the factors triggering this quiescence to senescence switch in yeast grown in media containing 2% glucose and excess amino acids is the production of acetic acid and the subsequent acidification of the medium to below pH 4 [1, 2]. Several hypotheses about the effects of this acidification have been proposed. One of them assumes that the acetic acid produced is the main cell-extrinsic factor determining CLS [2] and extracellular acidification encountered upon growth in batch cultures stimulates the accumulation of its dissociated form, rendering the cells more prone to its toxic effect [3-5]. Another opinion focuses on the potential use of acetic acid as a carbon source [6, 7] and thereby proposes that an overproduction of acetic acid may result in the inhibition of the beneficial effect of caloric-restriction. Nevertheless, the demonstrated positive effect of medium buffering on CLS [1, 2], along with the stimulation of senescence in human tumor cells by medium acidification [8], clearly indicates that extracellular acidification is a determinant in the preference of senescence over quiescence.

To investigate the effects of extracellular pH on CLS, Saccharomyces cerevisiae cells are aged in unbuffered and buffered (pH 4.5) media to represent the control and perturbed cases and extracellular metabolite analyses are performed in samples collected from these main cultures, whereas, for transcriptome analysis, a stimulation phase is added to the procedure: because the distinction between quiescence and senescence is only manifesting itself in response to stimulation, young and old cells collected from main cultures are ‘awakened’ in fresh media, where samples for transcriptome analyses are collected. The enrichment analyses of the significantly expressed genes suggest that medium buffering alters lipid metabolism, which in turn affects CLS by modulating mitochondrial and vacuolar functions.

Results and Discussion

pH control slows down chronological aging: the distinction between young and old cells is abolished in pH-controlled media

To investigate the effect of medium buffering on chronological aging, wild-type yeast cells were batchwise cultivated in fully controlled fermenters in synthetic dextrose complete (SDC) media, at 30 °C and 400 r.p.m., with an air supply of 1 vvm. For the pH-controlled case, the medium pH was kept at 4.5 (i.e. the initial pH of the fresh medium before the inoculation) by automatic addition of NaOH. For the uncontrolled case, no such buffering was adopted and the final pH of the culture was observed to be 3.5, from 72 h until the end of the experiment (Fig. S1a). Samples representing the ‘young’ and ‘old’ cells were taken at days 3 and 7, respectively (Fig. S1) and were transferred into fresh media until D600 of 0.2 was reached, before collection for mRNA extraction and the subsequent microarray analysis, whereas the original samples taken from the main cultures were tested for colony-forming units (CFU) and metabolite analyses. The collected cells either are in the early quiescent state (the beginning of post-diauxic phase, day 3) or have spent 4 days in that state (day 7).

As shown in Figs 1 and 2, pH control delays chronological aging in S. cerevisiae: the adaptation of the old cells grown in buffered culture is faster than those grown in unbuffered culture, although they still require more time to adjust to the new medium compared to the young cells (Fig. 1). This discrepancy between the old cells of the buffered and unbuffered cultures is more readily seen upon inspection of the survival values: the old cells grown in the pH-controlled medium preserve their colony-forming capacity (~ 90% of the young cells grown in the same culture) relative to those grown in the unbuffered medium, where the CFU decreases to ~ 50% compared to the values in young cells (Fig. 2). These results are in agreement with the reported beneficial effect of medium buffering on CLS [1].

Figure 1.

The growth curves of ‘stimulated’ cells. Growth of young (y) (A) and old (o) (B) cells collected from pH-controlled (pH) and uncontrolled (nopH) media. Subscripts 1 and 2 denote biological replicates.

Figure 2.

Survival values of young and old cells grown in pH-controlled (pH) and uncontrolled (nopH) media. The error bars represent the SDs of two biological and two technical replicates, respectively.

When the extracellular ethanol, glucose, acetate and glycerol levels (metabolites all reported to affect the CLS of yeast) [9] are analyzed, the long-living old cells of the buffered medium are observed to accumulate significantly higher glycerol levels in their extracellular medium (t-test, P < 0.01), whereas no statistically significant change is observed for glucose, ethanol or acetate concentrations between any other conditions (Fig. 3). Indeed, Fig. 3 shows that both young and old cells grown in pH-controlled and uncontrolled media are in the diauxic phase, a requirement for the entry into the quiescent state, because the extracellular glucose is completely depleted [10]. Although statistically nonsignificant, the relative drop in the ethanol and acetate concentrations of the young and old cells grown in the pH buffered medium, along with their increased glycerol levels, is quite similar to the phenotypic traits observed in the long-lived mutants with deficiencies in the TOR-Sch9/S6K and Ras-PKA pathways. These mutants also deplete ethanol, showing a reduced accumulation of extracellular acetic acid [11, 12], as well as activation of glycerol biosynthesis [12] .

Figure 3.

Extracellular metabolite levels of the young (y) and old (o) cells grown in pH-controlled (pH) and uncontrolled (nopH) media. The error bars and asterisk represent the SDs and significance of two biological and two technical replicates, respectively.

Aside from the phenotypic data, the life span-inducing effect of pH control also manifests itself at the transcriptome level. When the microarray data obtained from awakened young and old cells are analyzed by principal component analysis (PCA), it is observed that principal component 1 (PC1, 54.71% variance) allocates the experimental conditions with respect to pH control, whereas PC2 (22.97% variance) divides them according to the second factor, age (Fig. 4). Although the data for the young and old cells are well separated by PC2 if they stem from the unbuffered media, data belonging to the different age groups grown in pH-controlled media fail to do so and cluster together along PC2.

Figure 4.

PCA plot of the transcriptome data belonging to the young and old cells grown in pH-controlled and uncontrolled media. ‘nopH’ and ‘pH’ denotes unbuffered and buffered media, whereas ‘o’ and ‘y’ represents old and young cells, respectively. Numbers in parenthesis on the x- and y-axes represent the percentage variance captured by the respective principal component.

The convergence of the transcriptome data of the old and young cells grown in buffered media creates a notable effect on the differentially expressed gene analysis: because the pH control counteracts the aging effect, the group comprised of old and young cells cultivated in pH-controlled media behaves relatively similarly, with factor 1, pH control, dominating the results. This effect is investigated further below.

Transcriptome analyses reveal that pH control exerts its extending effect on life span by affecting a broad range of cellular processes

The statistical analysis identified a total number of 1038 genes (out of 5667), whose expression profiles are significantly altered in at least one condition (Fig. 5 and Table S1b). pH control-significant genes (PSG) have a statistically significant difference in their expression (q ≤ 0.01) between the two pH control levels: age-significant genes (ASG) showed significant differences in their expression between different age groups (q ≤ 0.1) and the interaction-significant genes (ISG) (q ≤ 0.07) are affected differently by pH control across the different age groups.

Figure 5.

Two-way analysis of variance. The blue, yellow, red and green circles contain gene numbers whose expressions are significantly altered by pH control, age, and interaction effect of pH control and age, respectively.

The dominance of F1 manifests itself in the results of the analysis of variance (ANOVA): more genes are affected by pH control compared to F2, age, or the interdependence between F1 and F2. No genes were affected solely by factor 2, age. This behavior is quite logical and expected when the similar transcriptome profiles (Fig. 4) and CLS (Fig. 2) of the young and old cells collected from the pH-controlled medium are taken into account. Thus, pH control exerts a pooling effect on the ‘young’ and ‘old’ cells, both at transcriptional and phenotypic levels.

The main process affected by pH control appears to be involved in biological regulation, as observed via the regulation of chromatin silencing, microtubule depolymerization, transcription and pseudohyphal growth terms (Fig. 5, 885 genes, blue box). Terms related to telomere maintenance and DNA repair emphasize the protective role of pH control on DNA maintenance. Proteasome assembly and misfolded protein catabolism terms, which are among the enriched processes of pH-controlled cells, suggest that S. cerevisiae grown in buffered medium has a better capacity to cope with the cellular stress created by damaged proteins. The co-emergence of proteasome assembly and cytokinesis terms also suggests a successful exit from mitosis in the cells collected from buffered medium because proteasomes are required for the proper directionality of the cell cycle [13], as well as cytokinesis, which is functional in the effective separation of the two daughter cells. The remaining enriched process terms such as lipid and polyamine metabolic processes (glycolipid biosynthetic process, ceramide catabolic process, polyamine catabolic process and β-alanine biosynthetic process) indicate that the main metabolic pathways affected by pH control are lipids and amino acids.

The interaction effect (Fig. 5, 155 genes, red box) of the two factors results in alterations in DNA integrity and checkpoint, as manifested by the ‘negative regulation of response to DNA damage stimulus’, ‘one-carbon metabolic’ and ‘deoxyribonucleotide biosynthetic’ processes, together with ‘signal transduction involved in meiotic recombination checkpoint’ terms. Aside from affecting DNA integrity and replication, the two factors, age and pH control, also behave antagonistically in regulating protein synthesis and maintenance, as revealed by the ‘cytoplasmic translation’, ‘oxidation-dependent protein catabolic process’, ‘de novo’ protein folding’ and ‘stress granule assembly’ terms. The co-occurrence of ‘oxidation-dependent protein catabolic process’ and ‘selenocysteine metabolic process’ among the enriched terms of ISG (with the latter generating an essential component of glutathione peroxidase and therefore protecting cells against oxidative damage) emphasizes the protective role of pH control against oxidative stress encountered upon advancing age. Terms belonging to cellular signaling machinery (‘protein dephosphorylation’ and ‘regulation of fermentation’), intracellular transport [‘endoplasmic reticulum (ER) to Golgi ceramide transport’, ‘proton transport’, and ‘retrograde vesicle-mediated transport, Golgi to ER’], component organization (‘Golgi organization’, ‘inner mitochondrial membrane organization’ and ‘mitochondrion degradation’) and lipid metabolism (‘ER to Golgi ceramide transport’ and ‘ergosterol biosynthesis’) constitute the remaining terms of the enrichment analyses of the ISG (Fig. 5, red box).

A preliminary result might be deduced from this global enrichment analysis: pH control affects predominantly the regulatory machinery of the cell and cooperates with factor 2, age, to alter organelle organization, intracellular transport and signaling mechanisms, ultimately leading to an effective cell cycle resumption. The simultaneous overrepresentation of the processes pertinent to sterol and sphingolipid metabolism among the significantly expressed genes suggests that this cooperation may be conducted via adaptations in cellular lipid metabolism. Indeed, sterols and sphingolipids have already been reported to take part in the branch point processes that are found to be affected by both pH control and age [14-17].

Maintenance of mitochondrial and vacuolar health appears to be the main route for pH control with respect to accomplishing effective cell cycle resumption

The results of experimental and statistical analyses suggest that pH control mainly affects regulatory machinery of the cell, which leads to an efficient exit from cell cycle arrest, whereas the main negative effect of unbuffered medium on CLS appears to exert itself on vacuoles and mitochondria. The specific responses of the up- and down-regulated genes constituting PSG and ISG have been analyzed in detail (Figs 6, 7, Figs S2–S3 and Tables 1 and 2). The self-organizing map method is adopted for the clustering purposes because it is a valuable tool for exploratory data analysis as a result of its unsupervised nature [18].

Figure 6.

Self-organizing maps of the transcriptome profiles of PSG around a 2 × 1 arrangement. ‘nopH’ and ‘pH’ denote unbuffered and buffered media, whereas ‘y’ and ‘o’ represent young and old cells, respectively. The cluster number and number of genes in each cluster are indicated in the top right corner and in the top center of each cell. The blue and red curves represent the centroids and the SD around the centroids respectively.

Figure 7.

Self-organizing maps of the transcriptome profiles of ISG around a 2 × 2 arrangement. ‘nopH’ and ‘pH’ denote unbuffered and buffered media, whereas ‘y’ and ‘o’ represent young and old cells, respectively. The cluster number and number of genes in each cluster are indicated in the top right corner and in the top center of each cell. The blue and red curves represent the centroids and standard deviation around the centroids, respectively.

Table 1. The filtered enriched GO process terms of the significantly expressed genes of clusters depicted in Fig. 6.
C1Proteasome regulatory particle assembly7.00 × 10−5
tRNA-type intron splice site recognition and cleavage2.54 × 10−3
Ceramide catabolic process7.74 × 10−3
Polyamine catabolic process2.54 × 10−3
Microtubule nucleation7.80 × 10−3
Cristae formation8.12 × 10−3
Intracellular sequestering of iron ion2.19 × 10−2
Mitochondria-associated protein catabolic process2.19 × 10−2
Methylglyoxal catabolic process to d-lactate2.19 × 10−2
Transcription from mitochondrial promoter2.19 × 10−2
Heteroduplex formation3.82 × 10−2
Threonylcarbamoyladenosine metabolic process3.82 × 10−2
Deoxyribose phosphate biosynthetic process4.12 × 10−2
Actin filament severing4.12 × 10−2
Proton-transporting V-type ATPase complex assembly4.12 × 10−2
Negative regulation of programmed cell death4.12 × 10−2
C2Negative regulation of chromatin silencing3.87 × 10−6
Exit from mitosis1.28 × 10−4
Pseudohyphal growth1.55 × 10−4
Regulation of transcription during mitosis2.87 × 10−4
Nuclear-transcribed mRNA catabolic process, deadenylation-dependent decay1.03 × 10−3
Regulation of transcription from RNA pol. II promoter in response to stress1.46 × 10−3
Regulation of cell size3.37 × 10−3
Positive regulation of inositol biosynthetic process4.65 × 10−3
Cytokinesis, completion of separation4.80 × 10−3
Phosphorelay signal transduction system5.41 × 10−3
Nitrogen catabolite activation of transcription5.41 × 10−3
Fungal-type cell wall organization8.12 × 10−3
Glycolipid metabolic process8.29 × 10−3
Establishment of cell polarity1.75 × 10−2
Amino acid transmembrane transport2.23 × 10−2
Signal transduction involved in filamentous growth2.64 × 10−2
Endocytosis3.36 × 10−2
Regulation of ergosterol biosynthetic process4.05 × 10−2
Telomere maintenance via telomerase4.33 × 10−2
Table 2. The filtered enriched GO process terms of the significantly expressed genes of clusters depicted in Fig. 7.
C1Ergosterol metabolic process7.22 × 10−4
One-carbon metabolic process4.51 × 10−3
Septin ring organization5.06 × 10−3
Double-strand break repair via nonhomologous end joining9.66 × 10−3
Protein kinase C-activating G-protein coupled receptor signaling pathway1.20 × 10−2
C2Meiotic recombination checkpoint2.24 × 10−5
Protein folding1.36 × 10−3
Modification-dependent protein catabolic process3.19 × 10−3
Response to oxidative stress2.96 × 10−2
C3Cytoplasmic translation9.34 × 10−4
Mitochondrion degradation8.84 × 10−3
C4Golgi to plasma membrane protein transport3.70 × 10−3
ATP hydrolysis coupled proton transport6.91 × 10−3
ER to Golgi ceramide transport8.49 × 10−3

Transcriptional profiles of PSG indicate disruptions of mitochondrial and vacuolar health as the main factors reducing CLS in unbuffered medium

Up-regulated processes when extracellular pH is uncontrolled

The simultaneous emergence of terms pertinent to the mitochondrial maintenance in the up-regulated processes of cells under uncontrolled pH (Fig. 6, Fig. S2, Table 1, C1 and Table S1c), such as ‘cristae formation’, ‘mitochondria-associated protein catabolic process’, ‘transcription from mitochondrial promoter’, ‘threonylcarbamoyladenosine metabolic process’ and ‘tRNA-type intron splice site recognition and cleavage’, suggest that cells grown in unbuffered medium tend to spend the available nutrients on mitochondrial maintenance after being transferred into stimulating medium. This is in agreement with the mitochondrial damage encountered upon oxidative stress during starvation, which yields mitochondrial incompetence as one of the main factors reducing the CLS [19-21]. Methylglyoxal, a glycating agent, is another inducer of mitochondrial dysfunction and reactive oxygen species (ROS) formation [22], so the need to up-regulate its catabolism supports the view that cells grown in unbuffered media have reduced mitochondrial function and elevated ROS, consistent with the preventive effect of medium buffering on ROS production [23]. This scenario supports the appearance of the enriched ‘heteroduplex formation’ and ‘deoxyribose phosphate biosynthetic process’ terms, indicating an effort to sustain DNA integrity in cells aged in the unbuffered medium: they may tend to recover from the mitochondrial DNA damage encountered upon oxidative stress when they are transferred into stimulating media. Furthermore, the two genes responsible for the ‘intracellular sequestering of iron ion’ term, NFU1 and SSQ1, are involved in the iron homeostasis in mitochondria [24, 25]; thus, they support the theory of a disrupted mitochondrial efficiency confronted in the unbuffered medium, which ultimately leads to a reduction in longevity [26, 27]. The up-regulation of the ‘proteasome regulatory particle assembly’ term may be indicative of the downstream effects of this mitochondrial damage because the ubiquitin-proteasome system is demonstrated to function in the mitochondrial quality control system of the cell [28-30]. The enriched ‘microtubule nucleation’ and ‘actin filament severing’ terms suggest that these cells try to reorganize their actin cytoskeleton upon transfer into fresh medium, another player that is tightly connected to ROS, mitochondrial health and longevity [31, 32]. The dynamical interplay between mitochondria and actin cytoskeleton is shown to affect the ‘nature’ of cell death; in other words, whether necrosis or apoptosis would take place [33, 34]. The remaining enriched terms such as ‘polyamine catabolic process’, which is demonstrated to be involved in necrotic cell death [34], and ‘negative regulation of programmed cell death’, as generated by FYV10 and STM1, which negatively regulate apoptosis [35, 36], suggest that cells aged in unbuffered medium tend to experience necrosis instead of apoptosis, with the latter shown to stimulate the survival of older cells in chronologically aged cultures[37, 38]. This assumption is further supported by the reduced CLS of the cells aged in unbuffered medium (Fig. 2).

The remaining two over-represented terms, ‘ceramide catabolic process’ and ‘proton-transporting V-type ATPase complex assembly’, highlight the participation of another crucial organelle, the vacuole (i.e. the yeast counterpart of the mammalian lysosome), in CLS regulation. Vacuolar ATPases are responsible for intracellular pH homeostasis and organelle acidification [39] and are also demonstrated to be functional in mitochondrial maintenance and the regulation of longevity [40, 41]. Moreover, external pH, sphingolipids and sterols modulate the activity of vacuolar ATPase (V-ATPase) [42-45], suggesting that cells grown in medium lacking external pH control have disturbed V-ATPase activity, which in turn impairs mitochondrial function.

Processes stimulated by external pH control

Unlike the genes in C1, both young and old cells grown in the buffered media (C2) are able to up-regulate genes involved in cell cycle progress and in the response to various stimuli necessary to cope with senescence (Fig. 6 and Table 1, C2; Fig. S2 and Table S1c) as observed via the ‘negative regulation of chromatin silencing ‘, ‘exit from mitosis’, ‘regulation of transcription during mitosis’, ‘regulation of transcription from RNA polymerase II promoter in response to stress’, ‘cytokinesis, completion of separation’, ‘establishment of cell polarity’ and ‘nuclear-transcribed mRNA catabolic process, deadenylation-dependent decay’ terms. These results reflect the indispensable trait of quiescent cells, the resumption of the cell cycle upon stimulation [46], which is the main distinguishing factor between senescent and quiescent phenotypes. In-depth investigation of the remaining enriched processes of the genes constituting C2, which co-emerge with the elements describing a successful cell cycle progress, might shed light on the routes leading to this ultimate beneficial effect of pH control on CLS.

One such route appears to comprise the efficient chromosomal maintenance machinery of the cells cultivated in pH-controlled media because some of the genes exhibiting an opposite trend in buffered and unbuffered media are involved in the negative regulation of chromatin silencing and telomere maintenance (Fig. 6 and Table 1). The up-regulation of these genes in response to pH control suggests that medium buffering results in the efficient maintenance of telomeres, in line with the direct relationship between telomere length preservation and longevity extension [47, 48]. Moreover, the increase in chromatin condensation encountered by intracellular acidification [49], which in turn leads to gene silencing and aging [37], infers that the antagonistic process of negative regulation of gene silencing in response to extracellular pH control is highly plausible.

Apart from chromosomal processes, several metabolic pathways involving lipids such as glycolipid biosynthesis, inositol and ergosterol metabolic processes and cell wall organization are among the activated machineries of the pH-controlled media compared to the unbuffered media. As a corollary, mechanisms involving the contribution of lipid metabolites such as cytokinesis, endocytosis and establishment of cell polarity, whose activities are demonstrated to increase the CLS span in yeast [50-53], are also induced. Therefore, one beneficial effect of pH control appears to involve cellular lipid metabolism.

The remaining processes that are over-represented by the up-regulated genes in pH-controlled media may be grouped under the signaling branch: the induction of mitogen-activated protein kinase (MAPK) (‘phosphorelay signal transduction system’, ‘regulation of transcription from RNA polymerase II promoter in response to stress’, ‘pseudohyphal growth’ and ‘signal transduction involved in filamentous growth’) and nitrogen sensing (‘nitrogen catabolite activation of transcription’ and ‘amino acid transmembrane transport’) machinery, processes that are both related to the target of rapamycin (TOR) signaling cascade [54]. Several downstream effectors of an active TOR signaling cascade such as GAT1, GLN3 [55], MDS3 [56] and RTG1 [57] are among the up-regulated genes in C2. Moreover, the up-regulation of processes governed by RNA polymerase II promoter (Table 1, C2), another TOR target regulating ribosome biosynthesis [58], suggests that cells grown in the buffered media are capable of proper TOR signaling activation upon encountering stimulating conditions. In other words, these cells are more competent with respect to sensing the abundance of amino acids in the stimulating medium, re-organizing their cellular state by activating the TOR cascade and thus resuming their arrested cell cycle. This competency in nutrient sensing machinery is not restricted to amino acids because these cells also have up-regulated MIG1, IRA1 and GPB1, which are genes involved in glucose repression and Ras-cAMP pathways [59-61], along with the up-regulated MAPK cascade, which is confirmed to be responsive to the extracellular glucose concentration [62]. On the other hand, although cells of the uncontrolled media have up-regulated KOG1 expression, the inducer of TOR cascade [63], they also have elevated levels of PPH22, the repressor of the same signaling machinery [64]. Together with the up-regulation of STM1 and REG2, two genes taking part in translation and glucose signaling processes, respectively, when nutrients are scarce [65, 66], it can be deduced that cells collected from unbuffered media fail to display coordinated nutrient sensing and processing machinery when they are stimulated by fresh medium. Thus, another possible route for the anti-aging role of pH control is the successful coordination of signaling mechanisms such as TOR and MAPK.

In summary, the simultaneous investigation of the enriched processes in C1 and C2 leads to the conclusion that extracellular acidification damages vacuolar and mitochondrial fitness, leading to a reduction in CLS. Medium buffering appears to prevent this detrimental effect via the successful resumption of the cell cycle as a result of the involvement of signaling machineries such as the TOR and MAPK cascades. Interestingly, both clusters possess terms related to lipid metabolism, especially those associated with ergosterol and sphingolipid biosyntheses, suggesting a role for these lipid sub-classes in the regulation of CLS. The inter-connection of sphingolipids and sterols with V-ATPase activity [44, 45], mitochondrial health [30, 67-70], TOR [71] and MAPK [16, 72] cascades further supports this hypothesis.

Specific responses of ISG in age and pH control groups: lipids, mitochondria, and vacuoles

Interaction-specific genes, which are genes that are affected by the interdependence of time and pH effects, display opposite behaviors across the levels of one factor when the level of the other one changes. In other words, they are affected differently by pH control across different age groups (Fig. 7, Fig. S3, Table 2 and Table S1d). The clustering of these genes yields four groups: C1 contains genes whose expression values are up-regulated in young cells, although no remarkable change in transcription is observed in old cells in response to pH control. In addition, the expression profiles of these genes increase with age in unbuffered media, whereas no such trend is observed for the pH-controlled case, and genes are similarly expressed in both young and old cells when they are collected from buffered media. Genes of C2 are also induced in young cells but repressed in old cells by pH control, similar to C1. However, although those genes are induced in the old cells compared to young cells grown in unbuffered media, controlling the medium acidity completely reverses this trend: the same group is repressed with age if the growth medium is buffered. C4 and C3 contain genes that display the antagonistic profiles encountered in C1 and C2, respectively.

Enrichment results of the genes belonging to C1 (Fig. 7, Fig. S3, Table 2 and Table S1d) reveal that cells cultivated in pH-controlled media exhibit relatively similar activities in ergosterol and one-carbon metabolic processes, DNA damage repair, septin-ring organization, and cell wall integrity machineries in both young and old cells, compared to those grown in unbuffered media, which display an increase in the same processes with age. Up-regulation of one-carbon metabolism, which is closely linked to the anti-oxidant glutathione [73], with increasing age may be a trait developed to overcome the oxidative stress endured upon aging as a result of the impairment of mitochondria [74]. Similarly, proper cytoskeleton organization, manifested by the ‘septin-ring organization’ term, is essential for mitochondrial heath and inheritance [28]. An increase in the intracellular ergosterol levels in the cells grown in buffered media may be related to the activating role of ergosterol in V-ATPase function, whereas the similar tendency of unbuffered cells to up-regulate ergosterol biosynthesis is probably related to its protective role from oxidative stress [75]. Finally, the terms related to DNA damage repair and cell wall integrity mechanisms are also involved in the response to oxidative stress [76, 77]; therefore, the remarkable up-regulation of these processes in the old cells grown in unbuffered media is expected. The interesting result, however, is the induced expression profiles of these genes in the cells of pH-controlled media, irrespective of age, compared to young cells collected from unbuffered medium. This profile suggests that cells of the buffered media exhibit a constant ROS generation, irrespective of age, and, apparently, this minor increase in ROS does not cause a reduction in CLS.

The enriched processes of C2 genes (Fig. 7, Fig. S3, Table 2, C2; Table S1d) might also be interlinked with mitochondrial functions and subsequently-generated ROS: ‘protein folding’, ‘modification-dependent protein catabolic process’ and ‘response to oxidative stress’ terms emphasize the detrimental effect of oxidative stress on protein structure that plays a crucial role in life span regulation [78]. The ‘meiotic recombination checkpoint’ term is also closely related to ROS signaling because ROS damages DNA and two of the three genes responsible for the emergence of this term, PSY2 and PPH3, are involved in DNA damage checkpoint and meiotic DNA replication mechanisms [79, 80]. Because the strains adopted in the experiments are haploid HO deletion mutants, they are unable to divide by meiosis, and so the induction of meiotic DNA replication may be an adaptive response to achieve recombinational DNA repair [81-83]. Therefore, the stimulation of these machineries in the old cells is again expectable, in agreement with a decreased mitochondrial capacity and the subsequent elevation of ROS, similar to terms emerging from the genes of C1. The same profile, in other words the induction of machineries counteracting detrimental ROS effects, is also observed in young cells of the buffered medium, as in the case of C1. The up-regulation of these age-related traits in young cells grown in pH-controlled cultures, along with the finding that medium buffering extends CLS, may initially appear contradictory, although recent discoveries regarding the beneficial role of a minor increase in ROS levels in extending CLS may be the reason for this profile [84-87]: small and/or transient amounts of reactive oxygen species elicit a protective stress response that may improve lifespan, although the exact machineries are not known. Relatively large and/or chronic amounts of the same species, however, cause cellular damage or death because they exceed the capacity of the oxidative stress response to maintain homeostasis. Therefore, pH control appears to extend CLS by stimulating an early stress response and activating the establishment of efficient homeostasis, as observed via the expression profiles of the genes of C1 and C2.

The enriched terms of C3 are related to mitochondrial degradation and translation, and genes constituting these groups are mainly induced in young cells grown in unbuffered medium, whereas a similar response is not seen in old cells (Fig. 7, Fig. S3 and Table 2, C3). The establishment of a retrograde signaling between the TOR cascade, which regulates translation, and mitochondria [88] suggests that pH control diminishes this machinery in the case of young cells, whereas it further aggravates this cross-talk in aged cells. Consistent with the enrichment results described above, this behavior appears to result in the efficient cell cycle resumption of old cells grown in buffered media, thus lengthening CLS.

Terms constituting the enriched processes of C4 are involved in intracellular transport and vacuolar acidification processes, and the genes constituting these groups are mainly induced in young cells grown in unbuffered medium. All of these terms are inter-related to sphingolipid metabolism [44, 89, 90], suggesting that the main effector for the reduced CLS of old cells grown in unbuffered medium may be this specific sub-class of lipids.

To determine whether the young and old cells grown in pH-controlled or uncontrolled media differ with respect to their mitochondrial and vacuolar health, as well as their competence to cope with the oxidizing agent hydrogen peroxide, sensitivity analyses using agents such as KNO3, NaN3 and H2O2 were performed. To determine the mitochondrial function before treatment, untreated cells were also grown on respiratory media (i.e. plates containing glycerol instead of glucose as the carbon source). As shown in Fig. 8, the cells grown in either media exhibit an increased sensitivity to these perturbing agents with increasing age. However, cells grown in pH-controlled media are more resistant to all of the distress caused by those agents compared to those cultivated in unbuffered media; in the case of both young and old cells, this supports the hypothesis of an altered redox potential together with disturbed mitochondrial and vacuolar functions with increasing age, which can be prevented to some extent via medium buffering.

Figure 8.

Resistance of young and old cells to perturbations in cellular redox state, vacuolar and mitochondrial functions. Percent fitness of (A) young and (B) old cells collected from pH-controlled (pH) and uncontrolled (nopH) media grown on glycerol, or treated with NaN3, KNO3, and H2O2. Error bars represent percentage deviations in CFU values, calculated from duplicate measurements.

In summary, the simultaneously enriched terms related to sterol and sphingolipid biosyntheses, which are all linked to vacuolar ATP-ase [44, 45] and mitochondrial [30, 67, 70] activity, support the proposed role of pH control in extensive mitochondrial damage prevention via alterations in lipid metabolism because these two organelles have already been shown to cross-talk in S. cerevisiae [40, 91-93].

Integration of transcriptome data onto the chronological aging network suggests that Snf1p is a possible player in the regulation of CLS by its cross-talk with extracellular pH control and lipid metabolism

Increasing evidence suggests that cell responses are usually organized as pathways or responsive gene modules consisting of a group of interacting genes at the molecular level [94]. Hence, to obtain a deeper understanding of the cellular reprogramming occurring upon chronological aging, determination of responsive gene modules rather than individual responsive genes has a higher potential for unraveling the cross-talk between the altered cellular machineries described above. Accordingly, to identify the possible players taking part in the longevity extension via extracellular pH control, an integrative approach is adopted: the P-values obtained from statistical analyses of transcriptome data were superposed onto the aging network tCAN [95], with the use of ‘jactivemodules’ plugin of cytoscape ( This integration yielded three active sub-networks that were specifically responsive to the individual perturbations: age (ASN), pH control (PSN) and interaction effect of pH control and age (ISN) (Fig. 9 and Table S2).

Figure 9.

The results of the sub-network analysis. Yellow, blue and red colored nodes and edges represent members of ASN, PSN and ISN, respectively. Green nodes and edges represent shared paths and genes between the three sub-networks.

Topological analysis of the ISN led to identification of CDC28, YCK1 and TPK3 as the nodes with the highest connectivity; in other words, the ‘hubs’. Cdc28p is the catalytic subunit of the main cell cycle cyclin-dependent kinase, whereas Yck1p and Tpk3p mediate the cellular response to nutrient levels, processes that cross-talk and regulate the CLS via their joint effect on DNA replication stress [96]. On the other hand, WHI3, TOR1 and PEX29 are the hubs of PSN, highlighting the interconnection of the extracellular pH control between the TOR cascade, cell cycle and peroxisome organization with respect to regulating longevity. Last, but not least, hubs of ASN are BRE5 and RPT5, which are genes taking part in protein degradation [97, 98], suggesting the involvement of proteasome function and autophagy as the main disconcerted machineries in aged cells.

The 15 genes common to ASN and ISN (ACS2, CYS3, DHH1, MAK16, NAS6, PAN1, RVS167, SAC6, SNF1, SSD1, SWM1, ERG13, RNQ1, SAH1 and THR4) are mainly involved in endocytosis (P = 3.14× 10−2), emphasizing the role of cytoskeleton and endocytotic machinery in longevity regulation [51]. Similarly, the 15 genes that are members of PSN and ISN (YHL018W, YPL229W, ATG29, CDC5, LAG1, LYS21, MPT5, NPR2, NST1, YPL150W, BUD31, ELM1, SEN1, TCB3 and YVH1) include key players in the previously mentioned TOR (NPR2), cAMP (YVH1) and osmotic stress (ELM1 and NST1) signaling, autophagy (ATG29) and sphingolipid (LAG1) pathways, which are all known to regulate CLS. The 12 genes shared by ASN and PSN (PRE8, BLM10, GET3, CDC16, RPN4, PDR5, MRH1, VAC8, RPT2, DEF1, PGK1 and SER3) have roles in mainly the protein catabolic process (P = 2.84 × 10−3), thereby hinting that pH control plays a part in the elimination of damaged and/or oxidized proteins, a process extending longevity [99]. The gene common to all three sub-networks is TID3, which is involved in chromosome segregation, emphasizing the protective role of extracellular pH control in chromosome integrity and maintenance, a process closely linked to aging and cancer [100].

Apart from the common genes, the three sub-networks have also common edges: SAC6-RVS167, SNF1-SSD1, ERG13-THR4 and RNQ1-SWM1 between ASN and ISN; BLM10-PRE8, RPN4-RPT2 and PDR5-MHR1 between ASN and PSN; and TID3-CDC5-YPL150W-LAG1, NST1-YPL229W and ELM1-NPR2 between PSN and ISN (Fig. 9). The fact that these edges are part of different responsive sub-networks suggests that the nodes constituting these ‘bridging’ edges play an important part in the information flow between ASN, PSN and ISN, thereby rendering the nodes constituting these edges as attractive cross-talk points between extracellular pH control and chronological aging.

The collocation of ELM1, SNF1, LAG1 and ERG13 among the genes constituting these shared edges is particularly interesting considering the induction of AMPK activity with extracellular medium buffering [101-103] and the enriched sphingolipid and sterol-related terms of PSN and ISN (Tables 1 and 2). This result suggests that the previously mentioned cross-talk of extracellular pH control and cellular lipid metabolism may be transmitted via the Snf1 kinase complex. The recent findings on the inter-connection of sphingolipids and Snf1p in the regulation of cellular response to ER and cell wall stresses, as well as ion homeostasis, which are processes all involved in longevity determination [104], together with the regulatory role of mammalian AMPK in cholesterol metabolism [105], suggests that this is a hypothesis worthy of testing.

Validation of the effects of ergosterol and sphingolipids on CLS

Joint effects of pH control and lipid metabolism on CLS regulation

To validate the proposal of an altered lipid metabolism followed by improved V-ATPase activity for the protective role of pH control on chronological longevity, the enriched lipid pathways in the comparative transcriptome analyses, the sphingolipid and ergosterol biosynthetic pathways, are disturbed with myriocin (MY) and atorvastatin calcium salt (AC), respectively. MY is an antibiotic that inhibits serine palmitoyltransferase activity, the first step in the sphingolipid biosynthesis [106]. AC is a statin type drug that inhibits HMG-CoA reductase, the enzyme catalyzing the rate-limiting step in ergosterol biosynthesis [107].

Considering the antagonistic roles of ergosterol and sphingolipids on the functioning of V-ATPases [44, 45], hampering mevalonate and sphingolipids pathway activities should result in reduced and improved V-ATPase activity, respectively. The survival curves obtained for the drug-treated and control cases, both in buffered and unbuffered media, are provided in Fig. 10.

Figure 10.

Survival curves of BY4742. Data collected from cells grown in buffered (dashed lines) and unbuffered (solid lines) SDC (blue), SDC + MY (green) and SDC + AC (red). Error bars represent percentage deviations in survival values, calculated from triplicate measurements.

MY and AC treatment cause a delay in initial cell growth and increases the time that it takes for BY4742 cells to enter stationary phase (~ 4 days), a fact determined via CLS assays and unnoticed otherwise, via the growth curves. Drug-treated cells grow slowly and, unlike the other cells, do not stop growing at 72 h (CLS day 0) but keep growing up to approximately the 100-h time point, as observed from the continuous increase in their CFU counts. To compensate for this extended growth phase, day 0 for the drug treated cells was assumed to be the 100-h time point in the experiments (compared to the 72-h time point for the other cultures) and the survival values of these cultures were assessed accordingly.

A dashed black line is plotted to visualize 50% survival in Fig. 10, and the time points corresponding to this level of survival for different growth cases are evaluated. MY treatment extends the mean CLS by ~ 73%, whereas administration of AC reduces it by ~ 15% (Table 3) in unbuffered medium, consistent with the previously stated antagonistic roles of sterols and sphingolipids in V-ATPase activity. Moreover, small but significant reductions and increases in extracellular pH values are observed in the case of AC and MY treatments, respectively (P = 1.06 × 10−3 and 1.85 × 10−6), when the growth medium is unbuffered.

Table 3. Final extracellular pH and mean chronological life span values of BY4742 cultures. SDC, SDC + AC and SDC + MY represent the unperturbed, AC-treated and MY-administrated experiments respectively. pHin indicates the initial pH.
  Mean CLS (days)Percentage change in mean CLSFinal pHPercentage change in pH
(pHin 4.6)SDC + AC4.875–15.222.55–0.87
SDC + MY10.2578.262.794.35
(pHin 4.6)SDC + AC9.01–43.684.35–1.52
SDC + MY15.990.064.481.30

In the buffered medium, inhibition of sphingolipids cause no further beneficial effect, whereas hindering the activity of the mevalonate pathway with AC results in a more pronounced reduction (~ 41%) in CLS (Table 3) compared to the unbuffered case. These results are in accordance with the percentage change in pH [(ΔpHcase − ΔpHcontrol)/pHin × 100] upon drug administration, when buffered and unbuffered media are compared: the percentage increase in the extracellular pH caused by MY treatment is less dramatic in the buffered growth medium (1.3 versus 4.35), whereas the medium-acidifying effect of AC intensifies with medium buffering (1.52 versus 0.87). These results support the proposed cooperation of sterols, sphingolipids and extracellular pH in CLS determination, probably via the maintenance of V-ATPase activity.

Snf1p is involved in CLS regulation by pH control via alterations in ergosterol and sphingolipid metabolisms

To investigate the involvement of Snf1p in regulating sterol and/or sphingolipid metabolism upon external pH, the same CLS experiments were performed with a strain lacking Snf1p.

The deletion of SNF1 results in an ~ 40% CLS reduction compared to wild-type (3.50 versus 5.75 days) in unbuffered medium (Figs 10 and 11 and Tables 3 and 4). A 0.89% decrease in the final extracellular pH is also observed for the mutant strain (2.55 versus 2.59). However, although MY treatment gives rise to an increase in both CLS and final pH in snf1Δ/snf1Δ, as it does for the wild-type case, albeit at different magnitudes, the trend observed for AC supplementation is reversed in the deletion mutant: the hindering of ergosterol metabolism increases both mean CLS and extracellular pH in the strain lacking Snf1p, whereas it has an opposite effect in the wild-type strain (Figs 10 and 11 and Tables 3 and 4). The results are more clearly shown in Fig. 12, where the CLS data for the wild-type and snf1Δ/snf1Δ grown in unbuffered media are plotted.

Figure 11.

Survival curves of snf1Δ/snf1Δ. Data collected from cells grown in buffered (dashed lines) and unbuffered (solid lines) SDC (blue), SDC + MY (green) and SDC + AC (red). Error bars represent percentage deviations in survival values, calculated from triplicate measurements.

Table 4. Final extracellular pH and mean chronological life span values of snf1Δ/snf1Δ cultures. SDC, SDC + AC and SDC + MY represent the unperturbed, AC-treated and MY-administrated experiments respectively. pHin indicates the initial pH.
  Mean CLS (days)Percentage change in mean CLSFinal pHPercentage change in pH
(pHin 4.6)SDC + AC7.13103.712.611.30
SDC + MY6.3882.142.897.39
(pHin 4.6)SDC + AC7.63–44.064.34–0.22
SDC + MY8.75–35.804.411.30
Figure 12.

Survival curves of snf1Δ/snf1Δ and BY4742 grown in unbuffered media. Blue, green and red lines denote SDC, SDC + MY and SDC + AC media, whereas the CLS curves of mutant strain are represented by dashed lines. Error bars correspond to percentage deviations in survival values, calculated from triplicate measurements.

On the other hand, in the buffered medium, the reduction in mean CLS encountered upon Snf1p deletion is ~ 15% compared to wild-type (Tables 3 and 4), fortifying the proposed role of Snf1p in longevity regulation by extracellular acidification. Moreover, when the medium is buffered, inhibition of ergosterol metabolism decreases CLS by ~ 44% both in the mutant and deletion strain, whereas hindering the activity of SPT creates an opposite response in snf1Δ/snf1Δ compared to wild-type: inhibition of sphingolipid synthesis does not affect CLS of the wild-type strain dramatically in buffered medium, whereas it causes a 36% reduction in the mean CLS of the strain lacking a functional Snf1p (Figs 10-13 and Tables 3 and 4). It is also worth noting that the positive correlation observed between the trends indicated by the change in final extracellular pH values and mean CLS is disrupted in MY-treated mutant cells: although the final pH increases with MY treatment, mean CLS decreases (Table 4).

Figure 13.

Survival curves of snf1Δ/snf1Δ and BY4742 grown in buffered media. Blue, green and red lines denote SDC, SDC + MY and SDC + AC media, whereas the CLS curves of mutant strain are represented by dashed lines. Error bars correspond to percentage deviations in survival values, calculated from triplicate measurements.

In summary, both AC and MY supplementation, when combined with SNF1 deletion, create a contradictory response in mean CLS across the two levels of factor 1 (pH control), which is a clear indication of an interaction effect. These conjoint antagonistic effects support the interplay between extracellular pH control, Snf1p and sphingolipid and ergosterol metabolisms in the regulation of longevity.


Controlling the pH of the medium extends the CLS [11, 108], although cellular mechanisms leading to this beneficial effect are not completely understood. Initially, a role for acetic acid as a cell-extrinsic aging factor was proposed as the main reason for aging [2], although the discovery of the involvement of diverse signaling pathways in the cellular responses to medium acidification and subsequent loss of viability suggests that a more intertwined mechanism takes part in acidification-induced aging [7, 109]. The similar acetic acid profiles of the young and old cells grown in either buffered or unbuffered media, despite their different proliferation capacities, supports the view that the positive effects of medium buffering on CLS cannot be attributed solely to extracellular acetate concentrations.

Indeed, the main result of this part of the present study is the fact that medium buffering phenocopies several long-lived mutants of signaling pathways, such as RAS2Δ, SCH9Δ and TOR1Δ, by exhibiting decreased levels of extracellular acetic acid and ethanol at the same time as accumulating extracellular glycerol. The effect of pH control is not only observable at the metabolite, but also at the transcriptome level: the old and young cells cultivated in buffered medium tend to cluster together, unlike those grown in acidified media, and, more interestingly, they are in close vicinity to the aged cells (rather than the young ones) of the uncontrolled media. This profile suggests that controlling the medium acidity does not produce an ‘overall rejuvenating’ effect but, instead, provides a more efficient metabolic reorganization, which, in the end, stimulates prolonged survival capacity. Therefore, the involvement of different signaling pathways is not only expectable, but also necessary for the beneficial effect of pH control on CLS.

The common response of the pH-controlled cells is comprised of the induction of cell cycle as well as lipid and regulation related processes, among which cellular transport and polarization machineries are pertinent. Elements of different signaling pathways, such as the MAPK and TOR cascades, are also present in the enrichment results of the genes grown in buffered media, emphasizing the important phenotypic trait of quiescent cells, which is the resumption of the cell cycle upon transfer into stimulating media by up-regulation of signaling pathways involved in growth.

Lipids, especially sphingolipids and sterols, constitute the shared processes of PSG and ISG along with mitochondrial, vacuolar and signaling-related terms. Apart from being involved in the intracellular transport systems, these two subclasses of lipids are also functional in mitochondrial health and intracellular signaling machineries [110]; thus, the simultaneous enrichment of mitochondria-related processes with those involving these lipids suggests that they may be key players responsible for the effect of pH control in extending the CLS. Taking into account the dynamic re-organization of sphingolipids upon diauxic shift (e.g. translocation of Isc1p to mitochondria where its presence is indispensable for the healthy functioning of this organelle) [70, 111], it is highly possible that pH control exerts an anti-aging effect via alterations in sphingolipid metabolism. A role for sterols and sphingolipids in cellular response to acid has already been established because of their connection to the ATPases responsible for the regulation of intracellular pH. The fact that V-ATPase activity is lost upon glucose depletion in cells grown in acidic medium, whereas only a reduction of 50% in activity is observed in the cells grown in high extracellular pH (the pH-controlled case in this study) [43], together with the enriched terms of PSN and ISN genes pertinent to V-ATPase activity (Table 1-C1 and Table 2-C4), and together with increased resistance of the cells grown in buffered medium towards KNO3 (Fig. 8), suggests that a regulatory loop in the reverse direction exists: medium buffering may induce V-ATPase activity by altering lipid metabolism.

Analyses of perturbed routes regulating CLS in response to the time spent in stationary phase, extracellular pH control, and the interaction of these two factors, confirm the participation of lipids not only in vacuolar functions, but also in intracellular signaling machineries: the pairwisely shared nodes and edges by the three sub-networks, ASN, PSN and ISN, include members of the sphingolipid (LAG1) and ergosterol (ERG13) metabolisms, together with those of the key signaling processes that are demonstrated to regulate longevity, such as TOR (NPR2), AMPK (ELM1, SNF1), cAMP (YVH1) and cell wall integrity (SSD1, MPT5) pathways, in addition to VAC8, the gene functional in the vacuolar fusion together with V-ATPases [112]. The emergence of SNF1 in the pool of the shared nodes and edges, together with its involvement in cellular responses to medium pH [110] and lipid biosynthesis [113], renders this kinase as an attractive candidate to be positioned at the inter-junction between extracellular pH control and lipid metabolism.

The CLS assays conducted with agents inhibiting the synthesis of sphingolipids and sterols also support the proposed involvement of these lipid subclasses in CLS. In line with the respective stimulatory and inhibitory roles of ergosterol and sphingolipids in V-ATPase activity, damaging ergosterol biosynthesis with AC reduces CLS, whereas hindering sphingolipid metabolism by MY extends it in the unbuffered medium. When the extracellular pH is maintained at 4.5, however, inhibition of sphingolipid synthesis does not affect CLS, whereas repressing ergosterol metabolism creates an even sharper decrease in CLS. As also revealed by the responsive sub-network analysis, the association of sphingolipids and sterols with intracellular signaling cascades is well established [71, 72] and so it would be improper to attribute the effects of MY and especially AC solely to V-ATPase activity. Yet, the positive correlation between the final extracellular pH values and mean CLS values supports the hypothesis of an altered V-ATPase activity and vacuolar function in response to drug-treatment. Considering the regulatory loop between extracellular pH and lipid metabolism proposed above, these results support the participation of altered lipid levels in the intracellular signaling dynamics encountered in the enriched processes. The presence of Snf1p in the bridging edges of responsive sub-networks, together with the antagonistic CLS profiles of snf1Δ/snf1Δ to MY and AC supplementation in buffered and unbuffered media, suggests that at least one route by which extracellular pH and lipids cross-talk involves Snf1p. The recent findings regarding the participation of Snf1p in the intracellular pH control [114] and on the cross-talk of this kinase with sphingolipid metabolism in the regulation of crucial cellular processes [104] fortifies this result. Interestingly, when the present study was being conducted, another group tested the role of sphingolipid metabolism in the regulation of life span, focusing mainly on their roles as signaling molecules [115], and reported similar results to those of the present study, for unbuffered media. By contrast to the results of the present study, however, they observed a comparable extension in CLS upon MY treatment in buffered medium. This discrepancy probably stems from the differences in the adopted media because their medium is supplemented with excess iron, another player interacting with both Snf1p and sphingolipid metabolism [104, 116].

In summary, the present study confirms that chronological aging is a multi-factorial process and that the alterations yielding a modification in CLS accomplish this ultimate effect via involvement of different machineries, such as the cooperation of extracellular pH, lipid metabolism and signaling cascades. One of the intermediate players in this collaboration appears to be Snf1p, although further experiments are needed to clarify the exact machinery by which the kinase conjoins these processes. Studies concentrating on the identification of the other possible participants in the reorganization of lipid metabolism upon extracellular pH control will also prove interesting because the results obtained may help to enlighten the dynamics of the cellular adaptation that plays a crucial role in the determination of cell fate.

Materials and methods

Yeast strains and growth conditions

The strain used in the main experiments performed in fermenters was ΔHO derived from a BY4742 background (Matα; his3Δ1; leu2Δ0; lys2Δ0; ura3Δ0; YDL227c::kanMX4) obtained from the EUROSCARF deletion collection ( to prevent any mating type switching in course of starvation and subsequent perturbations related to possible sporulation effects. Validation experiments were conducted with BY4742 (MATα; his3Δ1; leu2Δ0; lys2Δ0; ura3Δ0) and snf1Δ/snf1Δ (BY4743; Mat a/a; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; lys2Δ0/LYS2; MET15/met15Δ0; ura3Δ0/ura3Δ0; YDR477w:: kanMX4/YDR477w). All chemicals were supplied by either Sigma Aldrich, Inc. (St Louis, MO, USA) or Merck KGaA (Darmstadt, Germany).

For the main cultures, overnight cultures grown in SDC medium were diluted until a D600 value of 0.1 was reached and inoculated into 2.5-L fully controlled fermentors (Sartorius, Göttingen, Germany) with a working volume of 1 L of fresh SDC medium. The cultures were then grown batchwise at 30 °C and 400 r.p.m., with 1 vvm aeration. pH was maintained at 4.5 via automatic NaOH addition for pH-controlled case, whereas this control was turned off for the unbuffered experiments. Samples of the young and old cells for metabolite, longevity and sensitivity analyses were collected at day 76 (day 3) and day 168 (day 7), respectively, representing cells that are in the early quiescent state (the beginning of post-diauxic phase, day 3) and those that spent 4 days in that state (day 7). At the same time points, 1 mL of the main cultures of the young and old cells were transferred to stimulating media in micro-aerated 500-mL flasks containing 100 mL of fresh SDC medium. The growth on glycerol was determined by spreading the 1000-fold diluted samples directly onto yeast extract peptone glycerol plates, whereas, for the remaining sensitivity experiments, the collected samples were treated with either 0.2 mm NaN3, 50 mm KNO3 or 2 mm H2O2 for 30 min and were spread onto yeast extract peptone dextrose (YPD) plates. Growth was monitored after 3 days and sensitivity profiles were determined by normalizing the CFUs to those acquired from the untreated samples grown on YPD. All experiments were performed in two biological replicates.

Sampling and mRNA extraction

For metabolite analyses, samples of 1 mL from the main cultures were collected at the indicated time points, were centrifuged at 5870 g for 6 min and the pellets were discarded. The supernatants belonging to these samples were used for the extracellular metabolite determination. For mRNA extraction, 5-mL samples were collected from the cells grown in stimulating media at D600 of ~ 0.2, were frozen in liquid nitrogen immediately and stored at −80 °C before RNA extraction. RNA extraction was performed automatically with qiacube using RNeasy Mini Kit (Qiagen, Valancia, CA, USA) in accordance with the manufacturer's instructions.

Microarray data acquisition and processing

The qualitative and quantitative spectrophotometric analysis of RNA was carried out using a UV-visible spectrophotometer (NanoDrop ND-1000; Thermo Fisher Scientific Inc., Waltham, MA, USA). RNA samples to be used for microarray analysis were subjected to another quality check step. RNA integrity number values were checked using a microfluidics-based platform (Bioanalyzer 2100; Agilent Technologies, Santa Clara, CA, USA) using RNA6000 Nanokit (Agilent Technologies) and samples with RNA integrity number values in the range 7–10 were processed. Next, the microarray analysis steps comprising the synthesis of cDNA, the conversion of cDNA into a double-stranded DNA, transcription and synthesis of biotin-labeled aRNA from the double-stranded DNA, purification and fragmentation of aRNA, and final hybridization of aRNA were performed as described in the Affymetrix GeneChip®Expression Analysis Technical Manual (Affymetrix Inc., Santa Clara, CA, USA). The gene expression data are available on the ArrayExpress (accession number E-MEXP- 3972;

The raw data files were assessed with dchip [117] software for outliers at the array level, as well as at the probe-set level. Data without outliers were then processed in r using raw cell files via the ‘affy’ and ‘affycoretools’ packages of bioconductor [118-120]. RMA-normalized [121] and log2-transformed final expression values were adopted for the subsequent statistical analyses.

Statistical analysis

Log2 transformed expression values were analyzed via PCA and two-way ANOVA to determine the distribution of samples and significantly expressed genes, respectively, using the functions embedded in matlab, version 7.0 (MathWorks, Inc., Natick, MA, USA). The individual P-values obtained with two-way ANOVA were corrected for multiple comparisons with the ‘mafdr’ function of matlab (Table S1a). genecluster, version 2.0 [122] was used for clustering of significant transcripts via self-organizing maps. GO term analyses of the significantly expressed gene groups were performed with the web-based ‘ontologizer’ tool using the topology-weighted enrichment option [123].

To identify the genes affected by pH control and age, the gene expression data belonging to the young and old cells grown in the buffered and unbuffered media were analyzed using two-way ANOVA. The two factors of the experimental design are ‘pH control’ (F1) and ‘time spent in stationary phase’ (F2), which are both tested at two levels: ‘pH-controlled’ versus ‘pH-uncontrolled’ and ‘young’ versus ‘old’ cells. For the sake of simplicity, F2 is referred as ‘age’ in the proceeding sections. The P-values obtained by two-way ANOVA are then corrected for multiple testing and q-value thresholds for each condition are determined by histogram plots as q ≤ 0.01 for pH-control specific genes, q ≤ 0.1 for age specific genes and q ≤ 0.07 for interaction of pH-control and age specific genes.

Responsive sub-network analysis

Responsive sub-networks specific to each case were determined via ‘jactivemodules’ plugin of cytoscape [124, 125] with the annealing option and default parameters. Because this procedure relied on random sampling, it was repeatedly run to ensure that the networks were converging. The network was the tuned chronological aging network reconstructed elsewhere [95], whereas P-values belonging to the nodes of tCAN were those evaluated as described in the statistical analysis. The highest scoring network was the input for the second run of the algorithm, which was performed to obtain a more refined and specific sub-network for each case.

CLS determination and extracellular metabolite analysis

Samples of the main cultures collected at the indicated time points were serially diluted and spread onto four YPD plates (two biological and two technical replicates) for each case and time point. Colony formation was monitored after 2 days. CFU measurements were normalized to CFUs of day 3 (assuming 100% viability) to obtain survival data. Extracellular glucose, acetic acid, ethanol and glycerol concentrations of the same samples were determined enzymatically using Roche/Boehringer–Mannheim kits (r-biofarm AG, Darmstadt, Germany).

For the validation experiments, drug treatment cases and BY4742 cells were cultivated in liquid SDC medium [126] with 2% glucose. Overnight cultures grown in SDC were diluted until D600 of 0.1 was reached, and inoculated into erlenmayer flasks containing fresh SDC medium, maintaining a volume ratio of 1 : 5. In the case of pH control, buffering was achieved with sodium-succinic acid (200 mm, pH 4.6) buffer. The cultures were then incubated at 30 °C and 180 r.p.m. and were supplemented with either MY or AC to obtain a final drug concentration of 2 μm (i.e. a drug concentration that does not inhibit growth dramatically) [104] at their exit from lag phase (D600 of ~ 0.2). CLS was determined via CFU measurements, as described above.


The research was financially supported by Bogazici University Research Fund via Project No. 5681 and TUBITAK via Project No. 110M428. The authors would like to thank Professor Stephen G. Oliver from Cambridge University for providing the strains.