H. V. Westerhoff, Department of Molecular Cell Physiology, CRbCS, BioCentrum Amsterdam, Faculty of Earth and Life Sciences, Vrije Universiteit, Amsterdam, the Netherlands Fax: +31 20 444 7229 Tel: +31 20 444 7230 E-mail: firstname.lastname@example.org
In both industrial fermenters and in their natural habitats, microorganisms often experience an inhomogeneous and fluctuating environment. In this paper we mimicked one aspect of this nonideal behaviour by imposing a low and oscillating extracellular glucose concentration on nonoscillating suspensions of yeast cells. The extracellular dynamics changed the intracellular dynamics – which was monitored through NADH fluorescence – from steady to equally dynamic; the latter followed the extracellular dynamics at the frequency of glucose pulsing. Interestingly, the amplitude of the oscillation of the NADH fluorescence increased with time. This increase in amplitude was sensitive to inhibition of protein synthesis, and was due to a change in the cells rather than in the medium; the cell population was ‘trained’ to respond to the extracellular dynamics. To examine the mechanism behind this ‘training’, we subjected the cells to a low and constant extracellular glucose concentration. Seventy-five minutes of adaptation to a low and constant glucose concentration induced the same increase of the amplitude of the forced NADH oscillations as did the train of glucose pulses. Furthermore, 75 min of adaptation to a low (oscillating or continuous) glucose concentration decreased the KM of the glucose transporter from 26 mm to 3.5 mm. When subsequently the apparent KM was increased by addition of maltose, the amplitude of the forced oscillations dropped to its original value. This demonstrated that the increased affinity of glucose transport was essential for the training of the cells' dynamics.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Physiological and biochemical experiments are usually designed in a way that conditions are ideal and homogeneous. For example, cell densities are low, substrate concentrations are well defined and constant, and product concentrations are kept low to reduce product inhibition. However, in their natural environment cells often encounter less ideal conditions. For example, in industrial fermenters the discontinuous feeding of substrate and the high cell density can cause mixing problems and inhomogeneous cultures . The organism in the fermenter experiences a fluctuating extracellular environment, e.g. in terms of concentrations of substrates  and products as well as with respect to pH and oxygen tension. Glucose fluctuations encountered by microorganisms in an industrial bioreactor were found to be in the subminute to minute timescale . In this paper we set out to mimic one aspect of these nonideal conditions and studied whether the temporal dynamics of the extracellular glucose concentration could induce a dynamic response inside the yeast cells.
This issue of whether extracellular dynamics can cause intracellular dynamics should not be confused with the issue of whether static extracellular or intracellular conditions can influence autonomous intracellular dynamics. The latter issue was addressed by Reijenga et al., for so-called ‘autonomous’ glycolytic oscillations arising under static extracellular conditions, i.e. limit-cycle oscillations. The glucose transporter and therewith the effective concentrations of extracellular glucose and of inhibitors of glucose transport had substantial control on the frequency of the oscillations. Subsequently, it was shown that the glucose transporter and several of the glycolytic enzymes exerted substantial control over the eigenvalues of the glycolytic pathway, thereby codetermining whether yeast glycolysis oscillates or is steady . These findings confirmed that static intra- and extracellular conditions can affect intracellular dynamics. Here we shall address how dynamic extracellular conditions affect otherwise steady intracellular processes.
In previous studies on glycolytic oscillations there was an aspect of adaptation. Alterations in the make-up of the yeast cells made them prone to limit cycle oscillations [6,7]. For the cells to engage in sustained rather than transient oscillations, they had to be harvested at the diauxic shift (i.e. the shift from using glucose to using ethanol as a carbon source), and subsequently starved for 2 h. The essential nature of the pretreatment has remained unclear, but it is known that the growth conditions affect the affinity of the glucose transporters over a wide range . Moreover, starvation induces the (partial) degradation of glucose transporters [9–11] possibly bringing yeast glycolysis into the ‘oscillatory window’. Indeed, glucose transport offers a wealth of possibilities for Saccharomyces cerevisiae to exhibit adaptation. The organism has a large family of hexose transporters, six of which confer growth on glucose. These are HXT1–4 and HXT6–7 . An hxt1–7 deletion mutant does not grow on glucose anymore . HXT1 and HXT3 are transporters with low affinity kinetics, whereas HXT4 has a moderate affinity for extracellular glucose and HXT6 and HXT7 display a high affinity . HXT2 exhibits different affinities depending on the growth conditions [13,14].
The effects of external periodic events on oscillations, and their possible advantages, have been studied previously, however, mostly by means of mathematical models [15–17,18]. Here we perform in vivo experiments to study these effects. In this study we address cells that should themselves settle for a steady-state, were it not for an entrainment by extracellular dynamics. We harvested the cells during exponential growth on glucose and immediately put them on ice, without starvation . Subsequently the cells were subjected to the repetitive addition of aliquots (‘pulses’) of glucose. As the cells consumed (part of) the glucose during the intervals, this led to an oscillating extracellular glucose concentration. Previously , we combined modelling and experiments to demonstrate resonance of the intracellular dynamics with an extracellular glucose oscillation under these conditions. It was shown that the amplitude of the intracellular oscillation was considerably higher when the frequency of the extracellular oscillations was close to the eigenfrequency of the cells, than at other frequencies (the eigenfrequency of the cells was taken to be the frequency of the autonomous oscillations they engaged in under slightly different conditions). Here, the same experimental system is used, but the focus is not on resonance, but on the question of whether the cellular make-up can change such that their dynamic response increases. We report that the cells' response to the dynamic extracellular glucose increased with prolonged exposure to the extracellular glucose oscillations, and identify a possible mechanism that may be responsible for this increase.
Response of intracellular dynamics to the pulsing of glucose
S. cerevisiae X2180 was grown on glucose and harvested during exponential phase. After washing, the cells were resuspended to a protein concentration of approximately 13 g·l−1. The cell suspension was incubated in a thermostated cuvette (25 °C), and aliquots of 0.8 mm glucose were administered at a frequency of 1.5 min−1. This frequency is in the range that is observed in industrial fermenters due to incomplete mixing . As the cells consumed the glucose, this should lead to an oscillating extracellular glucose concentration. It was observed, using glucose indicators that the extracellular glucose concentration at the end of the experiment was below 2.8 mm (results not shown). This indicated that glucose did not accumulate during the experiment. Intracellular NADH fluorescence was measured continuously in the suspension. The cell suspension responded to the glucose pulses by an increase in NADH fluorescence, followed by a decrease (Fig. 1). The insert to Fig. 1 focuses on a part of the experiment and indicates the time points where the glucose pulses were given. The cells were forced into an oscillation with a frequency that was equal to the frequency of pulsing. Glucose was given in series of eight pulses, with 3 min and 20 s without glucose addition between the series. For the later pulses the NADH fluorescence increased before the glucose was added (c.f. the insert to Fig. 1). When the pulsing was stopped, the oscillations continued for a few periods. These trailing oscillations were damped and had a slightly higher frequency than the forced oscillations (c.f. the insert to Fig. 1). Clearly, extracellular dynamics induced intracellular dynamics.
Interestingly, the number of trailing oscillations increased with the time of exposure to the glucose pulsing regime. At first, no damped oscillations were observed, whereas later on, two or three periods were observed (Fig. 1). Moreover, the amplitude of the enforced oscillations increased strongly in time, ultimately by a factor of four (Figs 1 and 2, Table 1). During the first two series of pulses the NADH oscillations had a low amplitude. From the third series onward its amplitude started to increase until it reached a plateau after eight series at t = 60 min (Figs 1 and 2). We will refer to this increase in amplitude as ‘training’, and to the cells that have reached the maximum amplitude as ‘trained’ cells. When a trained cell suspension was mixed with a fresh cell suspension, the amplitude was close to the average amplitude of the two independent cultures (minimum 0.04, maximum 0.14, average 0.09, after mixing 0.10), indicating that the fresh cells were not entrained by the trained cells (Table 1). To distinguish between the increase of the amplitude of the NADH oscillation being due to a change of the cells or to a change in their environment, the cell suspension was centrifuged and the cells were separated from the supernatant. First, trained cells were resuspended in fresh medium and pulses of glucose were given to the suspension as described before. In this case the amplitude was similar to the maximum amplitude of trained cells (0.145). Subsequently, fresh cells were resuspended in the supernatant of trained cells. In the latter case the amplitude was similar to the minimum amplitude at t = 6 min (0.021) (Table 1). These results indicated that the increase in amplitude as seen in Figs 1 and 2 was due to a change in the make-up of the cells rather than to some extracellular product in the supernatant; the cells appeared to have been trained.
Table 1. Amplitude training of yeast cells. Summary of the results of the different experiments described. The second and third column reflect the average amplitude at t = 10 min and t = 90 min, respectively. The fourth column reflects the average amplitude after different treatments of the cell suspension. The first experiment (no additions) corresponds to Figs 1 and 2.
Average amplitude after different treatments
t = 10 min
t = 90 min
Amplitude after mixing or resuspending of cells.
Amplitude after 75 min of continuous feeding of glucose to the yeast cells. a.u., Arbitrary units of NADH fluorescence.
To test whether the synthesis of new proteins was required for the increase in amplitude, protein synthesis was blocked by cycloheximide. Inhibiting eukaryotic peptidyl transferase, this compound prevents the formation of new peptide bonds. No increase in amplitude was observed when cycloheximide had been added to the cell suspension before glucose (average amplitude 0.032) (Fig. 3). Two different concentrations were used (5 µg·mL−1, 50 µg·mL−1) and the results were similar. This indicated that protein synthesis was essential for the training of the cells and that the proteins involved have a net positive control on the amplitude.
Continuous low glucose concentration
The increase in amplitude could be caused by two different properties of the extracellular glucose signal: either (a) its pulsatile character, or (b) its low concentration. The former hypothesis should require that the cells have a memory for dynamics. The second hypothesis would involve regulation of protein synthesis and breakdown, after the shift of the cells from a high glucose concentration during exponential growth to a low concentration in the cuvette. To distinguish between these possibilities, cells were subjected to a continuous glucose feed for 75 min. On average, the rate of glucose addition was the same as in the pulse experiments described above. After the incubation, cells were subjected to pulses of glucose in the cuvette, as described before, and NADH fluorescence was measured. The amplitude of the NADH oscillations after a continuous and low glucose feed (0.17 of our arbitrary units) was similar to the maximum amplitude cells that were trained by a pulsatile and low glucose feed (0.15) (Fig. 4). This indicated that the low glucose concentration, rather than its pulsatile addition, led to the expression of proteins that have a pronounced effect on the amplitude of the intracellular NADH oscillation.
The above-described results raised the question of which proteins were expressed that caused the increase of the amplitude of the NADH oscillations during incubation at low glucose concentration. High affinity glucose transporters were obvious candidates for two reasons. First, it is known that after a shift from high to low glucose concentration, as was the case in these experiments, high affinity glucose transporters are induced . It had to be investigated, however, whether this occurred even over the short time scale of these experiments. Secondly, glucose transport exerts a substantial control on the frequency of autonomous oscillations . We wondered if it also controlled the amplitude of these forced oscillations. To test the hypothesis that high affinity transporters were responsible for the training of the cells, we first investigated whether the affinity of glucose transport did change during the pulsing of glucose. Glucose transport kinetics were measured both in cells harvested during exponential phase, and in cells that had been subjected to a pulsatile glucose concentration for 75 min. From Fig. 5 it becomes clear that during the experiment the kinetics of the glucose transporter changed substantially. For cells harvested during exponential phase, the Vmax and the KM of the transporter were 496 ± 23 nmol·min−1 per mg protein and 26.2 ± 0.3 mm, respectively. For the cells after pulsatile addition of glucose, the Vmax and the KM of the transporter were 518 ± 11 nmol·min−1 per mg protein and 3.5 ± 0.1 mm, respectively. We concluded that the Vmax of the transport system hardly changed, whereas the affinity of transport for glucose changed from low to high.
Subsequently, we asked the question whether the observed decrease of the KM of glucose transport was required for the increase of the amplitude of the NADH oscillation. To this end we increased the KM of glucose transport in trained cells artificially to approximately its value in fresh cells, by addition of maltose, a competitive inhibitor . As the Ki of glucose transport for maltose varies with strains and conditions  (presumably depending on which transporters are expressed), we determined its value in the trained cells. The Ki was then 32 mm (results not shown). It was calculated that a final concentration of 210 mm of maltose was needed to increase the apparent KM for glucose transport from 3.5 mm (trained cells) back to 26 mm (exponentially grown cells). At t = 100 min maltose was added to various final concentrations. Fig. 6 shows a substantial decrease in the amplitude of the forced oscillations, after the addition of maltose. The drop in amplitude increased with the final concentration of maltose as can be seen in Fig. 7. The minor increase in amplitude upon addition of phosphate buffer may reflect an effect on the fluorescence signal due to dilution. The values for the amplitude after addition of maltose were corrected for this dilution. From interpolation, it was concluded that a final concentration of 210 mm of maltose, which should increase the KM of the transporter back to its original value, was sufficient to abolish the increase of the amplitude due to training by 90% (Fig. 7). These results indicated that, within experimental error, the difference in KM of glucose transport was sufficient to explain the difference in amplitudes of the forced oscillation between trained and untrained cells.
In this paper we studied the dynamic response of otherwise steady yeast glycolysis to an oscillating extracellular substrate concentration. Cells were harvested during exponential growth phase, as these cells do not engage in limit cycle oscillations, when subjected to glucose . Indeed, when given steady extracellular glucose (data not shown), or after the first single addition of 3.2 mm glucose (Fig. 1) the cell populations in our experiments exhibited a fairly steady, nonoscillatory level of NADH fluorescence, but when glucose was added in a pulsatile fashion, the NADH fluorescence followed dynamically. This showed either that extracellular dynamics can induce dynamics of intracellular NADH, hence presumably of intracellular glycolysis, or that the pulsatile addition of glucose synchronized pre-existing oscillations of the individual cells.
The latter explanation assumes that the individual cells oscillated anyway but out of phase. We consider this explanation unlikely for the following reasons: (a) it has been demonstrated that the exponentially grown cells used here could not be synchronized by partly trapping acetaldehyde, the synchronizing agent ; (b) mixing trained with nontrained cells led to oscillations of average amplitude, whereas synchronization should have led rapidly to the full amplitude (compare with ); (c) cycloheximide prevented the training, suggesting that at least the amplitude increase due to training was not due to synchronization; (d) added maltose reduced the amplitude immediately of cells that should already have been synchronized, suggesting again that at least the amplitude increase due to training is not an effect of synchronization.
The dynamics of the intracellular response was quasi-sinusoidal at the same frequency as the extracellular perturbation. It was not chaotic, as it might have been if the cells engaged in limit cycle oscillations themselves (compare with [22,23]). There was an active aspect to the cellular response, i.e. after a few glucose pulses, the response seemed to run slightly ahead of the extracellular pulsing and the intracellular oscillations persisted for some time after the extracellular pulsing had been stopped, and at this somewhat higher frequency. This behaviour suggests that the cells were not in a stable node but in a stable focus, and had a dominant eigenfrequency that was slightly higher than the frequency of extracellular pulsing. This was the fastest response of the cells.
There were two slower responses. The first of these was the increase in average NADH fluorescence observed during the first three series of pulses (Fig. 1). From the data on the presumed extracellular glucose concentration and the measured KM and Vmax values of glucose transport, we estimated the glucose transport rate at the beginning and at the end of the experiment. At the beginning of the experiment the glucose transport activity should have been too low to completely consume the extracellular glucose within 40 s, and therefore extracellular glucose may have accumulated. Two effects will cause the glucose transport rate to increase during the experiment: (a) the presumed accumulation of extracellular glucose; and (b) the decrease of the KM of the transporter. These two effects should have led to an increased glucose transport activity and therefore they should eventually have stabilized the glucose concentration averaged over a pulse and therewith the NADH fluorescence averaged over a pulse. This may explain the initial increase in average NADH fluorescence.
The cellular response at the slowest time scale was the gradual and more than fourfold increase of the amplitude of intracellular NADH oscillations, a phenomenon we attributed to ‘training’. We observed a decrease of the KM of glucose transport from 26 to 3.5 mm during the training period and proved that increasing the KM back to its original value abolished the effect of the training on the amplitude. Strictly speaking we cannot conclude from this result that the change in KM of glucose transport is sufficient to achieve the training effect and that no other change in the cellular make-up is involved. This conclusion would require the opposite experiment of decreasing the KM of untrained cells, without changing anything else in the cells, which is a more complicated if not impossible task. Considering, however, that increasing the KM of untrained cells decreased the amplitude back to its original value, we may conclude that the KM change is necessary. And although we have not proven it definitively, we find it at least likely that the KM change is the major if not the only change required to achieve the training effect. Which of the Hxt proteins are involved, remains open for further investigation.
Previously, we have determined the control of glucose transport on the frequency of autonomous oscillations . Glucose transport had substantial control over the frequency of these oscillations, i.e. on this dynamic behaviour. Additionally, we have studied forced oscillations in terms of resonance phenomena in the same, exponentially grown yeast cells that were used in this study . As resonance occurs only in systems around a stable focus, we can conclude that these yeast cell populations are stable but do have a dynamic component. We might hypothesize that through the change in the kinetics of the glucose transporter the training has an effect on the dynamic component of the stable cells, i.e. on their eigenfrequency. The required change of eigenfrequency would be very small, since a sharp resonance peak was observed in the untrained cells at approximately 1.7 min−1, indicating that the untrained cells have an eigenfrequency of 1.7 min−1 (as compared to a glucose pulsing frequency of 1.5 min−1 in this study). However, the decrease of the KM of the transporter in time should lead to an increase of the rate of the transporter and according to the positive control of the transporter on the frequency  this should increase the eigenfrequency of the system and thereby move the eigenfrequency even further away from the extracellular frequency. Moreover, the previously observed resonance increased the amplitude by less than a factor of two, while we observe an increase of a factor of four in this study. Therefore we consider it more likely that the mechanism of training merely reflects the positive control exerted by glucose transport on the amplitude of enforced oscillations. After all, enhanced influx of glucose should be expected to enhance the reduction of NAD(P)H.
This study effectively also explored consequences of fluctuating extracellular conditions in industrial fermenters. What do the results mean then for yeast cells in an industrial environment? During industrial production of yeast, the average sugar concentration in the fermenter is low, the cell density is high and high affinity glucose transporters are expected to be expressed, as is the case in glucose-limited chemostat experiments . This situation resembles the conditions described in this paper, particularly as the time scale of sugar fluctuations in an industrial reactor may be similar to the frequency of pulsing applied here . Consequently, we might speculate that the intracellular metabolism of yeast cells in industrial fermenters also reacts to extracellular dynamics by engaging itself in those dynamics. And, it is conceivable that the response itself is subject to training. Of course the regular glucose pulses that were applied in the present study are a only a first approximation of the more complex glucose profile that the cells encounter in a bioreactor when they pass the medium inlet repeatedly. Due to the complex time dependence of extracellular glucose concentration for individual cells in such a bioreactor, the time scales involved might have subtle or drastic implications for the metabolic performance of those cells and for their gene expression pattern.
This paper illustrates that metabolism can only be understood by including gene expression and protein synthesis, as these processes partly determine the activity of the enzymes involved. As described earlier , the different levels of regulation and control cannot be treated separately and therefore a hierarchical approach should be taken to explain the overall cellular behaviour. Furthermore, the ability of cells to adapt to their environment and to anticipate to certain changes therein may enhance their chances of survival, by tightly regulating the use of their available free energy.
Yeast nitrogen base without amino acids was from BD (Franklin Lakes, NJ, USA). Glucose was from Boom (Meppel, the Netherlands) (when used as carbon and energy source in the medium) or from Sigma (St. Louis, MO, USA) (when used in the glucose transport assay). d-[U-14C]glucose was from GE Healthcare (St. Giles, UK), glass microfibre filters (GF/C) from Whatman (Brentford, Middlesex, UK) and liquid scintillation fluid from Perkin-Elmer (Boston, MA, USA). All other reagents were obtained from Merck (Whitehouse Station, NJ, USA), Sigma or Fluka (St. Louis, MO, USA), and were of analytical grade or higher.
Strain, growth conditions and preparation of the cell suspensions
The yeast strain S. cerevisiae X2180 was used in all experiments (grown in-house). Cells were grown semiaerobically on yeast nitrogen base, containing 1% glucose and 100 mm phthalic acid (pH 5.0, KOH) at 30 °C and harvested at an D600 of ≈ 1.0. Cells were washed twice with 100 mm phosphate buffer (pH 6.8, KOH) and resuspended in the same phosphate buffer to an D600 of 80, corresponding to a protein concentration of 13 g·L−1. Protein concentrations were determined according to Lowry . When large amounts were needed, cells were grown in a 2-L batch fermenter at a working volume of 1.5 L, a stirrer speed of 800 r.p.m., an air flow of 45 L·h−1, and at 30 °C and pH 5.0. The medium was yeast nitrogen base containing 1% glucose and 100 mm phthalic acid, set to an initial pH of 5.0 by addition of KOH, but not pH controlled.
Cells were incubated in a thermostated cuvette (25 °C, 300–600 µL) and NADH fluorescence was measured on-line (excitation 338 nm, emission 456 nm). Fluorescence intensity is given in arbitrary units (a.u.), as the value depends on instrument settings and cell density. The latter parameters were standardized, in order to be able to compare experiments between each other. The D600 of the suspension was 80 (13 g·L−1 protein) and conditions in the cuvette were semianaerobic, i.e. the dense suspension was stirred in the absence of an additional air supply. Glucose was added to the suspension either manually or by means of an automated pump. This computer controlled pump (KD Scientific 200 two-syringe pump, Holliston, MA, USA) was operated with a 100-µL Hamilton syringe (#1710 with Teflon luer lock; Ø = 1.46 mm). The syringe was connected to the cuvette through Teflon tubing (≈ 40 cm, Ø = ≈1 mm) using two needles (one with a luer lock for the Hamilton syringe). It was checked that both methods gave similar results (results not shown). At t = 1.0 min, glucose was added to a final concentration of 3.2 mm. Starting at t = 6.0 min, eight aliquots (‘pulses’) of glucose were given, each corresponding to an increase in concentration of 0.80 mm, at a frequency of 1.5 min−1 (T = 40 s). Pulsing was stopped for 200 s and at t = 14 min the pulsing was repeated. During a typical experiment, a total of 12 series of eight pulses were given (Fig. 1). The average amplitude of NADH fluorescence during one series was determined by calculating the amplitudes of all the individual periods, and averaging them over that series. The first two periods of each series were not taken into account, as, especially at the end of the experiment, the frequency and amplitude of these periods were substantially different from the six following periods. The damped oscillations that were observed after the pulsing was stopped were not taken into account either. Average amplitudes were determined for all 12 subsequent series of pulses.
Continuous glucose feed
For a continuous feed of glucose, a thermostated (25 °C) and stirred 15-mL vessel was used. Ten millilitres of cells with an D600 of 95 were incubated, and glucose was added continuously using a masterflex pump (microprocessor pump drive; Cole-Palmer Instrument Company (Vernon Hills, IL, USA); flow 43.5 µL·min−1; glucose stock 0.22 m). In terms of its time average, the rate of glucose addition was the same as in the pulse experiments. After 75 min, cells were taken out of the vessel and incubated in a cuvette, under the same conditions. Subsequently, glucose pulses were given, as described above, and the cellular response was again read in terms of NADH fluorescence.
Before and after 75 min of pulsing of glucose, as described above, the kinetic characteristics of glucose transport were measured according to Walsh et al. . Cells that had been subjected to a pulsatile glucose concentration were washed and resuspended in phosphate buffer (100 mm, pH 6.8). The transport assay was performed at 25 °C.
This work was supported financially by the Netherlands Organization of Scientific Research (NWO) and the Technology Foundation (STW).