A glucose-starvation response regulates the diffusion of macromolecules

1Eukaryotic cells expend a significant amount of energy to establish and maintain a high degree of intracellular organization in order to regulate and orchestrate their complex metabolism. This includes the active enrichment of macromolecules into organelles, which in turn allows for the separation of biochemical pathways and enhances their efficiency by increasing local concentrations of enzymes and metabolites. Similarly, active transport by motor proteins on the cytoskeleton enables large eukaryotic cells to overcome the limits of Brownian motion in which the diffusion time scales with the square of the distance. Nevertheless, many cellular pathways and biochemical interactions are dependent on Brownian diffusion. Much research has therefore been performed to characterize diffusional processes that occur within eukaryotic cells. These studies have revealed that the very complex and extremely crowded cytosol of eukaryotic cells cannot be viewed as an ideal liquid but, instead, can be modeled as a polymer gel or soft colloidal glass. Still, a comprehensive understanding of the biophysical properties of eukaryotic cells is lacking, and it remains poorly understood how these characteristics influence macromolecular movement. Furthermore, little is known about the biological determinants of intracellular diffusion including whether cells functionally regulate their diffusional properties in response to changes in growth conditions or the environment.

2A particularly well-studied model of intracellular movement is the motion of chromatin in the nucleus. Various groups have analyzed chromatin mobility, but a coherent mechanistic understanding of this process has yet to emerge (reviewed in ). For example, it was shown that chromosomes wiggle in a manner consistent with constrained diffusion and it was proposed that chromosome movement results from Brownian motion rather than from active motility. In contrast, other studies have suggested that chromatin movement is ATP dependent, and it was hypothesized that chromatin movement is driven by a multitude of ATP-dependent processes along the length of the chromosome. Similarly, the role of the cytoskeleton in chromatin movement has remained controversial. For example, both microtubule-dependent and -independent movement was reported, and several recent studies have also implicated the actin cytoskeleton in chromatin mobility.

3To better characterize the diffusional properties of eukaryotic cells, we have analyzed here the movement of chromatin and mRNPs in budding yeast under changing growth conditions. Our results demonstrate that cells dramatically change their biophysical properties in response to glucose starvation, which causes a confinement of macromolecules and affects the mechanical properties of the cell. This effect cannot be explained by changes in ATP levels or pH but can be induced by a loss of cell volume without any change in cell mass. This response seems to be conserved as bacteria similarly restrict macromolecular mobility in response to starvation, and we show here that starvation also induces a volume loss in bacterial cells and severely affects the rigidity of fission yeast cells. Our results suggest a novel mechanism by which cells regulate their biophysical properties in order to adapt to environmental stress.

4To examine the effects of changes in growth conditions on nuclear chromatin dynamics in budding yeast, we analyzed the movement of various gene loci. LacO repeats were integrated at the POA1 locus on chromosome II, the URA3 locus on chromosome V, and on a centromeric plasmid (pLacO). Co-expression of LacI-GFP allowed us to visualize these three loci and track their mobility over minute-long sequences. Whereas several changes in growth conditions, including growth in different carbon sources or nitrogen starvation, had no obvious effect on chromatin mobility (data not shown), acute glucose starvation induced a dramatic cessation of chromatin movement (Figure 1A). This suggests that chromatin mobility is regulated by the presence of glucose.

5To quantify the dramatic changes in chromatin mobility, we calculated ensemble-averaged mean square displacements (MSDs) for the chromatin loci (n = 183–1172 trajectories each) (Figure 1B and C; Figure 1—figure supplement 1A; Figure 1—figure supplement 2A). These plots express the magnitude of diffusion for a given particle, quantifying the average displacement per unit time and are used to compute their effective diffusion coefficients. We find that the confinement of chromatin upon glucose starvation (Figure 1B and C; Figure 1—figure supplement 2) leads to an approximately three-fold reduction of the apparent diffusion coefficient (K): for instance, KPOA1 decreased from 5.7 x 10–3 µm2/s to 2.3 x 10–3 µm2/s upon starvation (Table 1). The change in mobility at this time scale was not caused by a change in the anomaly of the diffusion process as the anomalous diffusion exponent (α), which is given by the slope of the curves in the MSD log-log plot, is not affected (see also Table 1).

6To analyze whether glucose starvation uniquely affects chromatin dynamics in the nucleus, or whether it also influences the mobility of other macromolecules, we imaged the movement of cytoplasmic mRNPs, which can be conveniently tracked as single particles. 24-PP7 stem-loops were integrated into the 3’ UTR of GFA1 and FBA1, essential genes involved in distinct processes, and the movement of individual mRNPs was examined upon co-expression of the coat-binding protein, CP-PP7-3xYFP. Cumulative track projections revealed substantially higher mobility for mRNPs than chromosomal loci in glucose, which is expected given the significantly smaller size of mRNPs compared to chromosome fibers (Figure 1D). Yet, upon glucose starvation, GFA1 and FBA1 mRNPs also exhibited a dramatic reduction in their mobility (Figure 1E and F; Figure 1—figure supplement 1B). Removal of glucose led to a three- to four-fold decrease in the diffusion coefficient of both GFA1 (KGFA1reduced from 4.2 x 10–2 µm2/s to 1.3 x 10–2 µm2/s) and FBA1 (KFBA1reduced from 5.0 x 10–2 µm2/s to 1.4 x 10-2 µm2/s) mRNPs without affecting the anomalous exponent (α). These values are similar to the relative change that we observed in the diffusion of chromatin (Table 1). Depletion of glucose by growth into quiescence also confined the mobility of mRNPs; therefore, the effects we observe are not a consequence of our cell washes (Figure 1—figure supplement 2B and C). The decrease in mobility of both chromosomal loci and mRNP particles suggests that glucose starvation causes a confinement of macromolecules in the nucleus as well as the cytoplasm.

7To begin to understand the nature and mechanism of the starvation-induced reduction in chromatin and mRNP mobility, we first focused on the cytoskeleton. In eukaryotes, the movement of many macromolecules is directly influenced by the dynamics of the cytoskeleton. For instance, mRNPs can be actively transported by motor proteins along the cytoskeleton and both the actin cytoskeleton and microtubules were reported to influence chromatin dynamics. Interestingly, actin filaments were also shown to rapidly depolymerize upon starvation. We therefore wanted to explore whether the starvation-induced confinement of macromolecular mobility resulted from changes in cytoskeletal dynamics. Indeed, our starvation condition lead to ablation of actin filaments in nearly all cells (Figure 2A) and a nearly four-fold reduction of microtubule elongation events in G1 cells as visualized by Tub1-GFP (Figure 2B).

8If the changes in the mobility of chromatin or mRNPs upon glucose withdrawal were due to the reduction of cytoskeletal dynamics, we would predict that drug-induced inhibition of actin and/or microtubule polymerization mimics the starvation response. In order to test this hypothesis, we treated cells with the actin depolymerizer, latrunculin A (LatA), and/or the microtubule depolymerizing drug, nocodazole. Indeed, treatment with either drug reduced chromatin movement, although the effect was insufficient to mimic the confinement of chromatin during starvation (Figure 2C; Figure 2—figure supplement 1). Concurrent treatments with LatA and nocodazole reduced chromatin mobility in an additive manner to substantially reduce chromatin mobility (Figure 2C; Figure 2—figure supplement 1). Moreover, log-log MSD plots indicate that the effect of inhibiting the cytoskeleton is most apparent at longer time intervals and may augment the anomalous behavior of chromatin motion (αPOA1-DMSO = 0.56 versus αPOA1-Noc+LatA = 0.48; Figure 2C; Figure 2—figure supplement 1; Table 1).

9The actin and microtubule cytoskeleton could either act on chromatin itself or affect chromatin mobility indirectly by influencing overall nuclear motion. To differentiate between these two possibilities, we tracked the distance between two labelled chromosomal loci (POA1, chromosome II and PES4, chromosome VI) over time (Figure 2—figure supplement 1). This approach allows for the exclusive quantification of intranuclear movements because translational changes in the positioning of the nucleus affect both loci identically and thus, do not influence intranuclear distance measurements. This two-locus analysis confirmed the LatA and nocodazole-induced reduction in chromatin mobility as the interchromosomal distance between the two loci still decreased considerably with simultaneous drug treatment (Figure 2—figure supplement 1). Of note, however, this result cannot differentiate between cytoplasmic or nuclear cytoskeletal dynamics in modulating chromatin mobility (e.g. via deformations of the nucleus). In conclusion, actin and microtubule dynamics independently contribute to the mobility of yeast chromatin.

10In contrast to chromatin mobility, mRNP mobility was only moderately affected by perturbation of cytoskeletal dynamics (Figure 2D; Figure 2—figure supplement 1) suggesting that the mobility of the GFA1 and FBA1 mRNPs is largely independent of the cytoskeleton. Overall, our results show that glucose starvation restricts cytoskeleton-independent mobility as well as the mobility of macromolecules influenced by the cytoskeleton.

11Our results so far could be explained by two alternative models: 1) starvation impacts macromolecular diffusion through multiple, distinct mechanisms, or 2) a singular, starvation-induced pathway restricts the mobility of macromolecules, and leads to both the collapse of cytoskeletal dynamics and the restriction of mRNP mobility. The acute withdrawal of glucose in fermenting yeast cells is expected to have dramatic consequences on cellular physiology. For example, the cellular ATP concentration drops and the intracellular pH decreases in starved cells. We therefore tested whether these global changes in cellular physiology lead to the observed changes in macromolecular mobility.

12First, we investigated the changes in intracellular ATP concentration during starvation. Upon glucose starvation, the ATP concentration rapidly decreased by ~70%. Remarkably, after this initial drop, ATP levels were relatively stable at ~30% of the initial concentration for the remainder of the experiment (Figure 3A). Of note, the maintenance of this reduced ATP level required oxidative phosphorylation as the cellular ATP concentration quickly dropped to nearly undetectable levels when cells deficient in mitochondrial function were starved (cbp2∆; ) (Figure 3—figure supplement 1). In contrast, when cells washed in sugar-free medium were back-diluted into media containing glucose, ATP levels recovered to the initial concentration within 30. Thus, glucose starvation induces a rapid ~70% reduction of intracellular ATP levels in agreement with previously published results.

13To determine whether a ~70% reduction in intracellular ATP levels is sufficient to confine cytoskeleton-independent macromolecular mobility as well as mobility influenced by the cytoskeleton, we reproduced this decrease in ATP concentrations in the presence of glucose. This was achieved by back-diluting washed cells into glucose media containing 0.02% sodium azide (NaN3), an inhibitor of oxidative phosphorylation (Figure 3A). After 0.02% azide treatment, ATP levels dropped to the same level as in glucose-starved cells, but the mobility of chromatin and mRNPs did not decrease to the same degree as in starvation (Figure 3B and C). We therefore conclude that a ~70% reduction of intracellular ATP, as observed in glucose-depleted cells, is insufficient to fully explain the confinement of macromolecular mobility during glucose starvation.

14Next, we investigated the contribution of intracellular pH to macromolecular confinement upon starvation. The intracellular pH (pHi) of budding yeast was reported to decrease from pH ~7.3 to pH ~6.4 in a time frame consistent with the observed confinement of macromolecular mobility upon glucose deprivation. Using the pH-sensitive GFP analog, pHluorin, as a pHi biosensor, we confirmed in our experimental conditions that glucose starvation causes a reduction in pHi from ~7.4 to ~6.4 (Figure 4A). We then manipulated intracellular proton concentrations by the addition of the weak acid potassium sorbate (K+Sorbate). K+Sorbate traverses the plasma membrane, releases a proton, and reduces pHi in a concentration-dependent manner. Varying the extracellular concentration of K+Sorbate from 0 mM to 8 mM enabled us to titrate pHi from pH 7.4 to pH 5.9, thus covering more than the pHi range observed in cells grown either in (+) glucose or glucose conditions, with little effect on intracellular ATP levels (Figure 4A and B). Remarkably, exposing cells to increasing concentrations of K+Sorbate induced macromolecular confinement, and both chromatin and mRNP mobility could be titrated with decreasing pHi (Figure 4C–F; Figure 4—figure supplement 1).

15Importantly, however, treatment with the highest concentration of K+Sorbate (8 mM) was necessary to fully recapitulate the starvation-induced confinement of both chromatin and mRNPs (Figure 4C–F). Under this condition, pHi dropped to 5.9, which is below the physiologically observed pHi of 6.4 in glucose-starved cells (Figure 4A). When pHi was lowered to 6.4 (by the addition of 2 mM K+Sorbate), the movement of both chromatin and mRNPs was only moderately reduced (Figure 4C–F). Furthermore, starvation in glucose media adjusted to pH 7.4, which prevents a drop in intracellular pH, failed to inhibit the confinement of chromatin and mRNPs upon glucose starvation (Figure 4G and H). Thus, the starvation-induced reduction of pHi is neither sufficient nor necessary and cannot fully explain the observed confinement of macromolecular mobility.

16In the course of our experiments, we found that the nuclear volume decreased upon glucose withdrawal (Figure 5A). Since nuclear volume and cellular volume are generally tightly linked, we utilized a Coulter Counter to examine whether cell size changes upon glucose starvation. Indeed, the median cell volume decreased from 101.6 fL (σ = 54.6) in glucose to 86.1 fL (σ = 45.6) in starved cells, corresponding to a volume reduction of ~15% (Figure 5B). In addition, we observed that the yeast vacuole, an organelle involved in various processes including protein degradation and metabolite storage, swelled in size under glucose starvation conditions (Figure 5C). In non-starved cells, the vacuole-to-cell volume ratio was 0.25 ± 0.02, whereas for starved cells this ratio increased to 0.40 ± 0.01 (mean ± standard error for three independent experiments) (Figure 5D). In combination, this vacuolar volume expansion together with the decrease in total cell and nuclear volume, reduces the cytoplasmic space that is available for diffusing molecules upon glucose starvation by ~30% ('Materials and methods').

17We hypothesized that such a large reduction of the accessible cellular volume may be sufficient to induce the observed macromolecular confinement, for example, by an increase in macromolecular crowding or changes in cellular viscosity. If the hypothesis was correct that a decrease of cell volume was the singular response that confines macromolecular mobility, two predictions could be made: (1) lowering the pHi by the addition of K+Sorbate might induce a decrease in cellular volume since high concentrations of potassium sorbate can replicate the confinement of both chromatin and mRNPs during starvation, and (2) manipulation of cellular volumes, for example, by changes in osmolarity should recapitulate the effects of glucose starvation on cytoskeleton-independent mobility (mRNPs) and on mobility influenced by the cytoskeleton (chromatin).

18To test our first prediction, we explored the effect of pHi changes on cell volume. Remarkably, lowering pHi by the addition of K+Sorbate also induced an increasing reduction in cell volume. Strikingly, addition of 8 mM K+Sorbate, which led to the same mobility decrease as glucose starvation, reduced the median cell volume to 87.1 fL (σ = 46.3), a level virtually identical to that of glucose-starved cells (Figure 5—figure supplement 1A and Figure 5B). Although intracellular acidification led to a decrease in cell volume, it was not required for this decrease since starvation in glucose media adjusted to pH 7.4, preventing an intracellular pH drop, also reduced the cell volume (Figure 5—figure supplement 1B). Thus, the effects of K+Sorbate are better explained in terms of volume decrease than pH changes. These data suggest a strong relationship between cell size and macromolecular mobility (Figure 4C–F; Figure 5—figure supplement 1).

19To further test our model, we manipulated the cellular volume by changing the extracellular osmolarity. We treated cells with increasing concentrations of NaCl and monitored the effects on chromatin and mRNPs. Notably, addition of NaCl had little effect on pHi or cellular ATP levels, allowing us to uncouple the effects of cell volume from these factors (Figure 5—figure supplement 2). As shown in Figure 5E and F and Figure 5—figure supplement 2C and D, both chromatin and mRNP mobility were progressively confined by increasing the magnitude of osmotic shock, consistent with recent results demonstrating that osmotic shock affects intracellular diffusion. Therefore, a reduction of cell volume is sufficient to explain the starvation-induced confinement of macromolecular mobility.

20We next explored how decreased cell volume could impede the movement of macromolecules. In the very densely packed environment of the cell, a reduction in the volume accessible to diffusing molecules can alter protein-protein interactions and induce complex phase changes such as liquid-to-gel-like or liquid-to-glass-like transitions. To test whether glucose starvation causes an increase in molecular crowding, we examined whether the observed volume loss is compensated by a reduction in the total biomass of the cell. We performed lyophilization mass measurement experiments in which starved and non-starved yeast cells were freeze-dried to remove their water content, and their masses were measured and normalized against the total number of cells (Figure 6A). We saw no significant difference between the non-starved and starved samples (11.5 ± 0.5 pg/cell and 11.9 ± 0.4 pg/cell, respectively; mean ± standard deviation for three independent experiments). Importantly, yeast cells grown continuously in raffinose, which are known to be smaller than cells grown on glucose and thus served as a positive control, showed a 23% reduction in cell mass (8.9 pg/cell) compared to cells grown in glucose. Thus, acute glucose starvation does not induce a decrease in cell mass despite a significant reduction in the accessible cellular volume. This increase in mass density suggests that the intracellular environment experiences increased molecular crowding conditions, which may lead to broad changes in the material properties of the cell.

21To examine whether the starvation-induced changes in macromolecular mobility are accompanied by changes in the viscoelastic properties of cells we measured the effective stiffness of budding yeast upon acute glucose starvation using atomic force microscopy (AFM). Our setup used wedged AFM cantilevers for parallel plate compression to characterize yeast cell morphology. These live-cell micro-compression experiments provide measurements of cellular viscoelasticity and cell stiffness under various conditions. After removal of the cell wall, glucose-starved and non-starved yeast spheroplasts were subjected to AFM measurements. Force-response curves were recorded while compressing single cells and analysis showed that glucose starvation induced a change in the cells’ mechanical properties increasing their mean stiffness from 18 to 22 nN·µm–1 as well as increasing the cell-to-cell variability in stiffness (Figure 6B). Consistent with the cell size measurements described above (Figure 5B), these AFM experiments also suggest that starved yeast cells are smaller than the corresponding control cells and that there appears to be a negative correlation between cell size and stiffness (Figure 6—figure supplement 1).

22In a separate approach, we also investigated the rigidity of the cytoplasm by examining cell morphology after removal of the cell wall. Without an intact cell wall, the cytoplasmic turgor pressure causes the yeast spheroplast to ‘ball up’ into a sphere. Since budding yeast cells are naturally round to begin with, we treated cells with the mating pheromone α-factor to generate shmoos, mating projections that cause the cell to become substantially elongated. Spheroplasted shmooed cells which were not starved of glucose behaved as expected and formed spherical spheroplasts (Figure 6C, Video 1). Intriguingly, starved shmooed spheroplasts remained predominantly in an elongated form (~four-fold less spherical spheroplasts in starved versus non-starved conditions) (Figure 6C, Video 2). This resistance to the internal turgor pressure suggests that the starved yeast cytoplasm is significantly more rigid than that of non-starved yeast, consistent with our AFM measurements. Furthermore, this stiffening is reversible as re-addition of glucose to the starved spheroplasts returned them to a spherical form (Video 3), suggesting that sensory feedback pathways may exist to actively regulate this stiffening mechanism. Taken together our results demonstrate that budding yeast cells undergo dramatic biophysical changes upon acute glucose starvation affecting their viscoelastic properties in addition to their intracellular macromolecular diffusion.

23To examine whether similar biophysical changes can also be observed in other unicellular eukaryotes upon glucose withdrawal, we monitored the starvation-response of the fission yeast, Schizosaccharomyces pombe. Budding and fission yeast separated evolutionarily several hundred million years ago, and they are as different from each other as either is from metazoans. Since S. pombe is naturally rod-shaped, we performed analogous spheroplasting experiments as described in Figure 6C. Upon digestion of the cell wall, S. pombe cells round up and become spherical (Figure 6D, Video 4). Conversely, glucose-starved cells maintain their rod-shape, reminiscent of the shmooed budding yeast cells, again demonstrating that glucose-withdrawal leads to a remarkable stiffening of the cytoplasm (Figure 6D, Video 5). As for budding yeast, the stiffening is also reversible as addition of glucose to starved S. pombe spheroplasts returns them to a spherical shape (Video 6).

24It was recently reported that the bacterial cytoplasm has glass-like properties and can undergo liquid-like to solid-like transitions upon starvation or treatment with 2,4-Dinitrophenol (DNP), an uncoupler of oxidative phosphorylation. This condition results in confinement of macromolecules and disproportionally affects large particles above ~30 nm. The underlying mechanism for this cytoplasmic transition into a solid-like state in bacteria remains unexplored, but suggested that metabolic activity is needed to keep the bacterial cytoplasm fluid. Although the cytoplasmic organization of prokaryotic and eukaryotic cells have dramatic differences, the similarity between these observations in bacteria and our findings in yeast prompted us to test whether there is a common underlying mechanism and whether bacteria also use a reduction in cell volume to increase crowding and confine macromolecular mobility. Strikingly, a 30 min treatment of E.coli with 2 mM DNP caused a significant reduction in the volume of E.coli, inducing a ~9.5% modal volume contraction (SD = ± 4.1%) (Figure 6E). It should be noted, however, that the treatment utilized here, for technical reasons, is longer than the DNP-incubation time reported by. Nevertheless, this result suggests that the reported liquid-to-solid like phase transition of the bacterial cytoplasm may also be triggered by cellular volume loss and suggests that the starvation-induced reduction of cell size could be a deeply conserved starvation response.

25Our results identify a novel cellular response that severely alters the biophysical properties of cells and restricts intracellular diffusion in response to glucose starvation. The dramatic macromolecular confinement observed upon starvation of budding yeast cannot be explained by a reduction of intracellular ATP (Figure 3), by a drop in pH (Figure 4) or inhibition of global metabolism, as respiration persists upon glucose depletion (Figure 3—figure supplement 1). Instead, we propose that increased macromolecular crowding induced by an acute volume loss fundamentally alters the mechanically properties of cells leading to macromolecular confinement. Our results further suggest that both eukaryotic and prokaryotic cells regulate the level of crowding in order to control intracellular diffusion and macromolecular interactions. In this context, it is intriguing that metazoan cells also seem to regulate their cellular volume as it was recently reported that tissue culture cells rapidly swell during cell division, which might aid with the mitotic separation of large chromosomes.

26At this point, it is unknown how glucose starvation induces a reduction in cellular volume and more work is needed to elucidate the signaling pathways and mechanisms by which glucose starvation causes cell size contraction and vacuolar volume increase. Upon acute glucose starvation, budding yeast cells undergo a notable cytosolic acidification lowering their intracellular pH to ~6.4, presumably by the rapid inhibition of the major plasma membrane proton pump Pma1 (Figure 4;;; ). Whereas this acidification might contribute to macromolecular confinement and enhance changes of biophysical properties within cells, pH drop alone is neither necessary nor sufficient to recapitulate the macromolecular confinement, and the volume loss observed in glucose starved cells can occur in the absence of acidification (Figure 4 and Figure 5—figure supplement 1). An increase of the intracellular proton concentration below 6 by the addition of increasing concentrations of K+Sorbate is able to induce a complete confinement of macromolecules (Figure 4C–F) but also significantly reduces the available cytoplasmic volume. Interestingly, inhibition of cellular metabolism by 2-deoxyglucose and antimycin also provokes an enhanced drop of intracellular pHto 5.5, which in turn triggers macromolecular confinement and leads to profound changes in the mechanical properties of the cell consistent with a sol to gel-like transition of the cytoplasm (Munder et al., personal communication).

27Our data also support a mechanism of chromatin mobility in which the movement of yeast chromatin is strongly influenced by the joint dynamics of both actin and microtubules. Precedence for such a mechanism is evidenced by previous work describing similar microtubule-dependent motion and reported interactions between microtubules, spindle pole bodies, and centromeres throughout the cell cycle. Moreover, the actin cytoskeleton is known to propel the movement of meiotic chromosomes and to influence the diffusion of other large macromolecules, suggesting a central function throughout the life cycle of yeast. The long-range directional movement of chromosomal loci is also reported to be dependent on the actin cytoskeleton. This is consistent with our data, as the cytoskeletal inhibitors latrunculin-A and nocodozale appeared to more strongly affect the mobility of chromatin at longer timescales and have minor effects at shorter ones (Figure 2; Figure 2—figure supplement 1). Finally, recent work detailing the existence of actin filaments in the nucleus substantiates the possibility of a network of nuclear actin. However, because latrunculin A depolymerizes all actin, our work cannot currently differentiate between the effects of cytoplasmic and nuclear actin in modulating chromatin mobility.

28Paradoxically, previous experiments exploring the ATP dependence of chromatin mobility reached opposing conclusions in that treatment with sodium azide had only minor effects on the mobility of chromosomes in one study but drastically reduced their mobility in another. By tracking both chromatin and mRNPs in conditions that closely replicated the perceived differences in the two published protocols, we were indeed able to duplicate the distinct results (Figure 4—figure supplement 2). Importantly, however, we also observed that azide treatment lowered pHi and that the distinct treatments reduced intracellular pH to dramatically different extents (Figure 4—figure supplement 2). Thus, a pleiotropic effect of azide treatment on pHi might be sufficient to reconcile the observed differences in the confinement of macromolecules. These results also show that the azide treatment used here (Figure 3 (wash); Figure 4—figure supplement 2) replicates both the decrease in intracellular ATP concentrations and the drop in pHi in glucose-starved cells. This indicates that changes in ATP and pHi do not have additive effects and are neither alone nor in combination sufficient to explain the effects on macromolecular mobility that we observe in glucose starvation.

29As expected, our single particle-tracking experiments show that both chromatin and mRNPs display anomalous subdiffusion at the timescales that we investigated. Interestingly, while glucose starvation dramatically decreases the diffusivity of the tracked objects, it does not change the anomaly of the diffusion. Anomalous subdiffusion is usually considered to be a transient phenomenon that occurs in complex, disordered media such as the intracellular environment. At short timescales, objects explore a pure viscous environment thus displaying normal Brownian motion. Yet, at intermediate timescales, the motion of the objects is hindered by static structures, leading to anomalous subdiffusion. Finally, at long timescales, that is, when the traveled distances are much larger than the sizes of the structures that impede movement, diffusion is again pure Brownian motion. The fact that glucose deprivation affects diffusivity but not anomaly suggests that starvation induces a dramatic increase of the 'microscopic' viscosity without affecting the higher order cellular structures that impede macromolecule motion. Based on our results and according to the classical molecular crowding theory this change of microscopic viscosity is probably the consequence of an increase in crowding upon cell size contraction.

30As the cell cytosol becomes more viscous and restrictive to diffusional motion under glucose starvation, we posit that it must also become more resistive to external perturbations on the cytoplasm. Indeed, we find that upon starvation the cytoplasm of budding and fission yeast undertake an astonishing transition leading to an increase in the viscoelastic stiffness of the cell (Figure 6). Our observations are reminiscent of recently published observations in bacteria describing a glassy liquid-to-solid transition upon glucose starvation or DNP treatment. Furthermore, we demonstrate that bacteria also undergo a volume loss similar to starved yeast (Figure 6E). This observation suggests that molecular crowding also triggers the reported phase transition in bacteria, and similar mechanisms might be activated in both pro- and eukaryotes in response to glucose starvation.

31Notably, this modulation of intracellular diffusion represents a mechanism to globally regulate the biochemistry of cells. As described for bacteria, we envision glucose starvation as a dormant state which might hold cells in stasis until provisions in the environment improve. Consequently, we propose that cells functionally regulate intracellular diffusion in order to establish a cellular organization that is restrictive to some molecular interactions, perhaps thereby conserving energy, while simultaneously permissive to essential biochemistry.

32The conservation of cell contraction between prokaryotes and eukaryotes strongly suggests a vital function during starvation. On the one hand, the reduction in macromolecular mobility may allow for the rapid inhibition of excessive energy expenditure by diminishing unnecessary molecular interactions. This is consistent with previous observations describing the rapid inhibition of principal ATP consumers upon starvation. For example, Pma1p, required for regulation of intracellular pH and thought to use upwards of 20% of cellular ATP, is immediately inhibited upon glucose starvation. Ribosomes also disengage from mRNA, resulting in a global collapse of polysome profiles. Additionally, we found that depolymerization of the actin cytoskeleton may function, in part, to conserve ATP, as treatment with latrunculin A rapidly increases intracellular ATP concentrations (Figure 6—figure supplement 2). Because osmotic shock promptly inhibits translation initiation and depolymerizes the actin cytoskeleton, it is conceivable that the starvation-induced cytoskeletal breakdown and collapse of polysome profiles, as well as other starvation-associated phenotypes, could be partly driven by molecular crowding. Therefore, the confinement of intracellular diffusion through molecular crowding may prove vital for economical biochemistry and represent a very upstream event upon carbon starvation.

33On the other hand, molecular crowding-enhanced aggregation and confined macromolecular mobility may promote sustained interactions between complexes during starvation. In agreement with this hypothesis, starvation is known to induce the formation of compartmentalized biochemistry and potential pockets of liquid-liquid unmixing. For example, P-bodies, macromolecular complexes composed of mRNPs and mRNA decay machinery, form in the absence of glucose and potentially aggregate through unstructured regions of the component proteins. Moreover, stationary phase induces the formation of foci and filaments for large numbers of proteins. This aggregation was hypothesized to enhance interactions between members of the same pathway and promote more efficient biochemistry but was more recently also shown to cause enzymatic inactivation in some instances. As a consequence, prolonged starvation and confinement of macromolecular mobility may also pose unique challenges to cells, as complexes with elongated stretches of unstructured regions may prove more prone to extensive aggregation. It will now be interesting to explore whether other cell types, including metazoan cells, respond similarly to starvation or growth factor depletion and if limited diffusion indeed stimulates aberrant protein aggregation, a process which has been implicated in a wide variety of human diseases.

34Yeast strains used in this study are of the W303 strain-background, except for strains containing labeled mRNPs which are heterozygous diploids of W303 and S288c backgrounds. The E. coli strain utilized for volume measurements is wild type BW25113. Genotypes for every strain are listed in Table 2; plasmids are described in Table 3. Unless otherwise indicated, yeast cells were grown at room temperature in synthetic complete (SC) media at pH 5.0 (titrated with 1 M HCl) containing 2% dextrose (SCD). Specific dropout media was used to maintain plasmids when necessary. E. coli were grown at 30ºC in LB.

35Time-lapse epifluorescence imaging was performed using an inverted Nikon TE2000 microscope equipped with an Andor CCD camera, a motorized stage (Applied Scientific Instrumentation, Inc., Eugene, OR), 60x/1.49NA APO TIRF oil immersion objective, and controlled with Metamorph® software (Molecular Devices, LLC, Sunnyvale, CA). Cells were stabilized by coating the glass-bottom of MatTek® dishes (PG35G-1.5-14-C; MatTek Corporation, Ashland, MA) with 0.2% concanavalin A and aspirating the solution prior to adding 300 µL of OD600 0.2–0.3 cells. All imaging was performed at room temperature. Images for particle tracking were collected at a frame rate of 0.5 s for 60 s, unless otherwise indicated. Particles were tracked using custom MATLAB® (The MathWorks, Inc., Natick, MA) scripts based on the 2D feature finding and particle tracking software made available by Dr. Maria Kilfoil (; http://people.umass.edu/kilfoil/downloads.html).

36An average mean square displacement (MSD) was calculated for each condition using the well-described equation:

37Effective diffusion coefficients, K, were computed by finding the slope of the best-fit line between the first and tenth time points of the averaged MSD. The anomalous diffusion exponents of the log-log MSDs, α, were determined by fitting the first 40 time intervals of the averaged MSD to a power law.

38Five milliliters of cells grown in SCD media (OD600 = 0.4–0.7) were collected by centrifugation (3000 rpm) for 2 min. The supernatant was then removed and cells resuspended in 1 mL of SC media. Four additional wash steps followed, with two 2 min spins (6000 rpm) succeeded by two 1 min spins. The OD600 of the final cell suspension was determined and cells diluted into fresh SC or SCD media for analysis. Time-lapse epifluorescence imaging was performed 30 min after the wash.

39Logarithmic cells growing in SCD media (pH 5.0) were incubated with an appropriate concentration of a stock 20 mM K+Sorbate in SCD (pH 5.0) to give rise to the corresponding final concentration of K+Sorbate. Cells were treated for 30 min before imaging.

40All experiments involving sodium azide (NaN3) took place in SC or SCD media at pH ~5.8. Cells were treated with a final concentration of 0.02% NaN3 (in water) by two distinct protocols. In one treatment, sodium azide was directly added, or 'spiked', into a logarithmic culture. In the other, cells were first washed (via the acute starvation method) in SC media before being resuspended into SCD media containing 0.02% NaN3. This treatment is referred to as 'wash' and is the condition described in the main text.

41Cells were treated with either 15 µg/mL nocodazole (from a 1.5 mg/mL stock in DMSO; Sigma-Aldrich, St. Louis, MO), and/or 200 µM latrunculin A (from a 10 mM stock in DMSO; Sigma-Aldrich) for 20 min before imaging. This results in 3% DMSO as the control.

42Cells grown in SCD media (OD600 = 0.4–0.7) were collected by centrifugation (3000 rpm) for 2 min and resuspended in SCD media containing the appropriate concentration of NaCl (0.4, 0.6, or 0.8 M). Cells were then imaged after 10 min.

43The protocol employed was only marginally modified from. Briefly, 20 µL of sample (KWY1622 cells in culture) was added to 20 µL of 10% TCA and immediately vortexed for 1 min. Of the 40 µL solution, 10 µL was then added to 990 µL of ATP reaction buffer (25 mM HEPES, 2 mM EDTA, pH 7.75). After the time course, 50 µL of a luciferin/luciferase kit (ATP Determination Kit, Sensitive Assay; B-Bridge International, Inc., Cupertino, CA) was directly added to 50 µL of the buffered-ATP solutions and briefly mixed by pipetting the solutions in a Corning 96-well opaque white plate (Corning, Inc., Tewksbury, MA). Reactions took place in the dark for 10 min at room temperature. Luminescence was then measured using a Tecan Infinite 200 Pro microplatereader (Tecan Group Ltd., Männedorf, Switzerland) which was set for a 10 s integration time and no attenuation. Sample aliquots were taken at 0, 10, 30 and 60 min and standardized to a pre-treatment control. ATP concentrations were also corrected for OD600 at the beginning and end of each time course. ATP standards were used to generate a standard curve to insure linearity and estimate intracellular ATP concentrations.

44We utilized a CellASIC ONIX microfluidic profusion platform and accompanying microfluidic plates to track particles in quiescent cells (EMD Millipore, a division of Merck KGaA, Darmstadt, Germany). Quiescence was established by allowing cells to grow within the same SCD media for 7 days. Seven-day-old media, collected from the culture by centrifuging 1mL of cells and harvesting the supernatant, was used as perfusion media in the microfluidic plate at a rate of 2 psi.

45Cell volume measurements were performed as described in. In short, 1 mL of cells (KWY4736) were briefly sonicated (1 s, 2X) and diluted (1:100) into Isoton II Dilutent (Beckman Coulter, Inc., Brea, CA) such that saturation levels fluctuated between 1–3% during measurements on a Beckman Coulter Multisizer 3. If oversaturation triggered warnings at any point, the run was terminated and sample repeated. The aperture was also cleaned and flushed between each run. For yeast, the particle count was set to 50,000 and for bacteria, 100,000.

46To quantify filamentous actin, KWY165 cells were fixed with formaldehyde at a final concentration of 4% for 15 min. Fixed cells were washed twice with potassium phosphate buffer and resuspended into PBS-BSA. Cells were then stained with Alexa Fluor 568-phalloidin (Invitrogen, Thermo Fisher Scientific, Waltham, MA) and Hoechst stain. Samples were incubated in the dark at room temperature for 1 hr and washed again with PBS-BSA prior to imaging. Background was subtracted from z-stacks (0.15 µm steps over 3 µm) taken with a 100x/1.49NA APO TIRF oil immersion objective using a rolling ball radius of 1 pixel in ImageJ (NIH, Bathesda, MD). Maximal intensity projections were then blinded and cells classified based on the presence or absence of apparent filamentous or bundled actin.

47The phluorin calibration protocol was slightly adapted from to establish a standard curve. 50 µg/mL (in PBS) of digitonin was used to permeabilize cells for 10 min. After washing cells in PBS and placing them on ice, cells were then incubated in citric acid/Na2HPO4 buffer with pH values ranging from pH 5.0 to pH 8.0. phluorin emission intensity ratios were then calculated from calibration images (taken with a 100x/1.49NA APO TIRF oil immersion objective) using a custom script in ImageJ (NIH) and compared to the pH values of the corresponding buffer. The resulting curve was used to estimate experimental intracellular pH values generated from emission ratios produced from the same ImageJ script. The standard curve and all experiments were performed using KWY3661.

48Nuclear volume measurements were conducted on KWY4796 cells expressing TetR-GFP (and DsRed-HDEL). Cells were grown in complete synthetic medium with 2% dextrose overnight at 25ºC to OD 0.5–0.8. Cells were then treated as indicated and transferred to Nunc Lab-Tek chambers (Thermo Fisher Scientific) coated with concanavalin A at OD 0.2. To wash cells into different medium conditions, cells were pelleted at 6000 rpm in a table top centrifuge and washed four times with synthetic medium lacking dextrose and then resuspended in synthetic medium either lacking dextrose or supplemented with 2% dextrose. Three-dimensional stacks with 0.24 µm z stepping were acquired on a Zeiss Axio Observer Z1 spinning disk confocal microscope equipped with a motorized piezo stage (Applied Scientific Instrumentation, Inc.) and an EMCCD camera using a 100x 1.46 Oil, alpha Plan Apochromat objective. TetR-GFP was excited with a 488 nm laser line and scanned using a Quad Band 'RQFT' 405/488/568/647 dichroic and a 520/35 band pass emission filter. Image stacks were processed in Imaris (Bitplane AG, Zurich, Switzerland) to reconstruct nuclear volume.

49S. cerevisiae cells with GFP-Tub1 integrated at the URA3 locus were grown overnight to mid-log phase in SCD media (pH 5.0) and then treated by acute glucose starvation (as above). The cells were then adhered to Concanavalin A treated 384-well glass plates while maintaining the experimental condition. Cells were imaged by epifluorescence time-lapse microscopy for 20 min: a z-stack of five images, each 1 µm apart was acquired once every minute. The resulting images were processed by a maximum intensity projection and the number of microtubule elongation events per minute was manually scored by an experimental blinded to the sample identity.

50KWY165 yeast cells were grown in 400 mL synthetic complete + 2% dextrose (SCD) media and harvested during log-phase growth (OD600 ~ 0.8). Cells were spun down, supernatant removed, and cells re-suspended in 50 mL synthetic complete without dextrose (SC) media, and then repeated with two further wash steps. The sample was then finally re-suspended into 4 mL SC media. 2 mL were added to 48 mL SCD media, and 2 mL were added to 48 mL SC media. Both samples were incubated at 25°C for 30 min. Then the samples were spun down, supernatant removed, and cells re-suspended into 1 mL water, which was repeated two more times. After the final wash step, the supernatant was removed, and the cell pellet was lyophilized under 0.040 mbar pressure at -50°C for 24 hr using the Alpha 2–4 LDplus freeze-dryer (Martin Christ GmbH, Osterode am Harz, Germany). The dry mass of the cells were measured, and the cells were re-suspended in water and counted using a hemocytometer to determine the average mass per cell for both the non-starved and starved cells. For mass measurements with raffinose, the protocol was the same except dextrose was replaced with raffinose.

51KWY5112 cells expressing GFP and N-terminally tagged Vph1-mCherry were grown in SCD media and harvested during log-phase growth. The standard acute glucose starvation protocol was performed, and the cells were placed on concanavalin A-coated Lab-Tek chambered slides (Thermo Fisher Scientific Inc, Waltham, MA) and imaged on a spinning disk confocal microscope (Carl Zeiss AG, Jena, Germany). The entire cell volume was imaged using a z-stack (0.24 µm z-slice height) under 488 nm (GFP) and 561 nm (mCherry) laser excitation with a 100x α Plan-Apochromat objective. The z-stack images were segmented and the vacuolar and cellular volumes were computed using custom Python scripts.

52To calculate the change in accessible cytoplasmic volume after glucose starvation, the total cell shrinkage (15%), nuclear volume reduction (4.60 fL to 4.12 fL), and vacuolar volume expansion (25 to 40% of cellular volume) were taken into account. The accessible cytoplasmic volumes were computed using:

53where A, C, V and N denote the accessible cytoplasmic volume, total cell volume, vacuolar volume, and nuclear volume, respectively, the subscripts NS and S denote the non-starved and starved conditions, and NNS = 4.60 fL and NS = 4.12 fL.

54For stiffness measurements of yeast cells an AFM (Cellhesion 200, JPK Instruments) was mounted on an inverted microscope (Observer.Z1, Zeiss). DIC images were recorded using a 40x/1.2 C-Apochromat water immersion objective (Zeiss). Tip-less cantilevers (NSC12/noAl, Mikromasch, 350 µm long, nominal spring constant 150 mN·m-1) were modified with PDMS (Sylgard 184, Dow Corning) wedges to correct for tilt angle of the cantilever and to be able to confine yeast cells between two parallel plates [. KWY165 cells in log-phase growth were spheroplasted for 1 hr in 200 µg·mL-1 zymolyase 100T (Amsbio) in SCD plus 1 M sorbitol at 30°C. Spheroplasts were then either washed into SCD plus 1 M sorbitol (non-starved treatment) or SC plus 1 M sorbitol (starved treatment) for 1 hr at room temperature before plated onto concanavalin A-coated glass-bottomed Petri dishes (WPI). For each stiffness measurement, an isolated spheroplast was identified and DIC imaged. The image was used to determine the spheroplast diameter. Thereafter, the cantilever wedge was positioned ~20 µm above the spheroplast and lowered at 1 µm·s-1 until an upward force of 10 nN was recorded. The cantilever height was maintained for 10 s before the cantilever was raised 10 µm at 1 µm·s-1. The cantilever was then lowered onto the same cell, paused and raised successively for seven further cycles. The cantilever height and the force acting on the cantilever were recorded at 40 Hz. The slope (least-squares fit, nN·µm-1) of the data points between 3 and 7 nN of force of the downward phase of each cycle was determined. The mean of these slopes was used as the measure of spheroplast stiffness.

55For budding yeast cells, yYB5978 was grown overnight in YPD (1% yeast extract, 2% peptone, 2% dextrose) at 30°C and diluted to OD600 = 0.1 in 25 mL YPD. Two hours later, mating pheromone (α-factor, Sigma T-6901-1mg) was added at a concentration of 8 ng/mL in YPD containing 20 µg/mL casein as a blocking agent. After 4–5 hr, the shmooed cells were washed into either SCD or SC (for non-starved and starved conditions respectively) and incubated at room temperature for 1 hr. Cells were then spun down and resuspended into either SCD or SC containing 1 M sorbitol and 200 µg/mL zymolyase 100T (Amsbio). Cells were incubated at 30°C for 1 hr. Spheroplasting efficiency was examined upon cell lysis in water.

56For fission yeast cells, KWY6241 was grown overnight in Edinburgh minimal medium + 2% dextrose (EMMD) to log-phase, then washed into either EMMD or EMM (EMMD with no dextrose) and incubated at room temperature for 1 hr. Cells were then spun down and resuspended into either EMMD or EMM containing 1.2 M sorbitol and 5 mg/mL lysing enzmes from Trichoderma harzianum (Sigma L1412) and 0.5 mg/mL zymolyase 100T (Ambsio). Cells were incubated at 30°C for 1 hr. Spheroplasting efficiency was examined upon cell lysis in water.

57For experiments in which the cells were returned to dextrose media, the cells were spun down and resuspended into dextrose-containing medium (SCD + 1 M sorbitol for S. cerevisiae or EMMD + 1.2 M sorbitol for S. pombe).