We have studied oscillating glycolysis in the strain BY4743 and isogenic strains with deletions of genes encoding enzymes in glycolysis, mitochondrial electron transport and ATP synthesis. We found that deletion of the gene encoding the hexokinase 1 isoform does not affect the oscillations while deletion of the gene encoding the hexokinase 2 isoform results in oscillations with smaller amplitude. The latter is associated with an almost 50% decrease in hexokinase activity. Deletions in the genes encoding the α- and β-subunits of phosphofructokinase abolish the oscillations entirely. This loss in oscillatory activity is associated with a fourfold decrease in phosphofructokinase activity. Deletions of genes encoding subunits of the F1F0 ATPase also inhibit the oscillations in accordance with earlier studies using for example inhibitors. Finally, we identified an apparently new control point involving the mitochondrial cytochrome c oxidase. The latter is difficult to explain as oscillatory activity entails 100% inhibition of this enzyme. The mitochondria of this strain seem to have normal F1F0 ATPase activity. Overall these results support earlier experimental and model studies suggesting that in addition to processes within glycolysis also processes outside this pathway contribute to the control of the oscillatory behaviour.
Glycolysis in yeast has been studied extensively for decades and most of the enzymes of this pathway have been characterized in terms of their structure and kinetic properties. It has been known for more than 50 years that glycolysis in the yeast Saccharomyces cerevisiae can exhibit oscillatory behaviour, i.e. the occurrence of temporal oscillations in the concentrations of metabolites in cells where respiration has been blocked [1, 2]. The oscillations have been observed both in intact yeast cells  and in cell-free extracts . The function of these oscillations is still not clear, but it was recently suggested that they may appear as an inevitable side effect in the tradeoff between hard robustness and efficiency of metabolic pathways . In suspensions of yeast cells of high density the glycolytic oscillations in individual cells are synchronized . The synchronizing agent is most probably acetaldehyde. In less dense suspensions the individual cells may also oscillate, but the oscillations are not synchronized [7, 8].
While glycolytic oscillations in yeast extracts are now fairly well understood to the point that they can be simulated by mathematical models [9, 10], similar behaviour in intact cells is less well understood. Even though models have been developed that fairly well reproduce many of the features of oscillations in cells [11-14] we are still lacking fundamental knowledge about some of the biochemistry behind the oscillations, e.g. which factors control the behaviour. In intact cells the ability to exhibit glycolytic oscillation has its optimum near the diauxic shift where the cells' energy metabolism shifts from being predominantly fermentative to being predominantly respiratory . This could indicate that the development of mitochondria plays a central role in the emergence of glycolytic oscillations . Indeed, recently mitochondria were found to be important for the glycolytic oscillations in yeast cells. First, it was observed  that the mitochondrial membrane potential (Δψm) oscillates with the same frequency as NADH. Next, it was found that mitochondrial F1F0 ATPase activity seems to be important for the oscillations to take place [18, 19]. This experimental evidence for the participation of ATPase activity in the regulation of glycolytic oscillations is supported by studies of models of the pathway [20-22].
In the current study we extend the work on oscillations obtained with isogenic deletion mutants of the laboratory strain BY4743 initiated by Williamson et al. . In addition to strains with deletions of glycolytic enzymes we have studied glycolytic oscillations in strains with deletions of genes encoding mitochondrial enzymes and translocators. Furthermore, using an ATP nanosensor we have measured the activity of hexokinase (EC 18.104.22.168) in cell-free extracts of the wild-type BY4743 and in two mutant strains with deletions of genes encoding the two hexokinase isoforms Hxk1p and Hxk2p, and the activity of phosphofructokinase (PFK) in the wild-type BY4743 and in one mutant strain with deletion of the gene encoding the PFK α-subunit. The current experiments were carried out in order to provide data for studies leading to better models of the glycolytic pathway in intact yeast cells. Our study supports the results of previous experimental and some modelling studies [18, 20, 23, 24] that the control of glycolytic oscillations is distributed throughout the pathway with strong control points in the phosphorylation of fructose 6-phosphate catalysed by PFK (EC 22.214.171.124) and ATP utilizing ion pumps, e.g. F1F0 ATPase (EC 126.96.36.199), and minor control points in the phosphorylation of glucose catalysed by hexokinase. The data obtained are valuable in improving current detailed models of glycolysis, which so far are based on data obtained with a single strain X2180. Furthermore, we present quantitative data that relate enzyme activities to the ability to exhibit oscillatory dynamics. Finally, we have identified an apparently new and unexpected regulation of oscillations involving cytochrome c oxidase (EC 188.8.131.52).
It was recently shown that glycolytic oscillations much like those obtained previously with the strain X2180 could be observed in the laboratory strain BY4743  and several of its isogenic deletion mutants . This enables a more biochemically oriented study of the enzymes controlling the oscillatory activity. In the study by Williamson et al.  the following glycolytic enzymes were found to be important for the occurrence of oscillations: the hexokinase isoform Hxk2p, the β-subunit Pfk2p of PFK and the glyceraldehyde-3-phosphate dehydrogenase (EC 184.108.40.206) isoform Tdh3p. In addition it was found that an intact cAMP-mediated protein kinase A (PKA) signalling pathway was important for the oscillations. Here we extend these studies by further investigating some of the mutants already studied by Williamson et al.  and some new mutants with deletions of mitochondrial enzymes.
First, we tested the ability of the three sugars glucose, fructose and mannose to induce oscillations in glycolysis. The oscillations were induced by addition of sugar and KCN to a dense suspension of yeast cells. In addition to glucose other sugars such as fructose, mannose and sucrose may induce oscillations . An example of adding fructose to the wild-type BY4743 and two mutants with deletions of hexokinase activity is shown in Fig. 1. In the mutants hxk1Δ and hxk2Δ the hexokinase 1 and 2 isoforms respectively have been deleted. Similar experiments where glucose and mannose were added to induce oscillations in these three strains are shown in Figs S3 and S4. We note that all three sugars can induce oscillations in the wild-type and in the hxk1Δ mutant, while in the hxk2Δ mutant glucose and fructose, but not mannose, may induce oscillations in glycolysis. Furthermore, in the latter strain, only oscillations in NADH of low amplitude are observed with glucose and fructose, but, as noted by Williamson et al. , the oscillations last longer. In our study we did not observe significant differences in oscillation period or amplitude between the wild-type BY4743 strain and the hxk1Δ mutant. The frequencies determined for the wild-type strain and the hxk1Δ strain were 0.026 s−1 for the substrates glucose and fructose and 0.024–0.025 s−1 for mannose. The frequency obtained for the strain hxk2Δ was 0.023 s−1 for glucose and fructose. What is the reason for the difference in oscillatory behaviour in the three strains and what is the reason for the different oscillatory patterns obtained with the different glycolytic substrates?
In order to answer the first question concerning the different oscillatory behaviours obtained with the three different strains we tested the total hexokinase activity in cell-free extracts of the three strains using a recently developed ATP nanobiosensor [27, 28]. Figure 2 shows the decline in concentration of ATP after adding 20 mm glucose to a solution containing ATP, ATP sensor and yeast extract of the wild-type strain BY4743. The initial rate of ATP consumption was measured as 0.75 U·mg−1 protein in good agreement with the Vmax value of 0.84 U·mg−1 protein reported by Teusink et al. . However, in the following we have calculated the enzyme activities in per cent of the corresponding activity in the wild-type strain BY4743. The values are listed in Table 1. We note that in the strain where hexokinase 1 has been deleted the hexokinase activity is essentially the same as in the wild-type B4743 strain, while in the strain where hexokinase 2 has been deleted the activity is only about half of that in the wild-type. This suggests that hexokinase activity is important for the oscillations to take place. If hexokinase activity is important in controlling the oscillations this might also explain the differences in the oscillatory patterns obtained with the different glycolytic substrates. To this end we tested the rate of phosphorylation of glucose, fructose and mannose by isolated hexokinase. All three sugars are readily funnelled through hexokinase-catalysed reactions into glycolysis, where they are further broken down to pyruvate and ethanol with a concomitant production of energy in the form of ATP. The results are shown in Fig. 3. From the changes in ATP concentration we can estimate the initial rate of phosphorylation and we find that, compared with the initial rate of glucose phosphorylation, the rates of fructose and mannose phosphorylations are 150% and 64%, respectively. A similar result was obtained by measuring the phosphorylation of the three sugars catalysed by a cell-free extract of the wild-type BY4743 strain (data not shown).
Table 1. Relative kinase activities in strains with deletions of genes encoding hexokinase (HK) and phosphofructokinase (PFK) (n =3–9)
HK activity (%)
PFK activity (%)
Likewise, we tested the ability of strains with deletions of genes encoding the α- and β-subunits of heterooctameric PFK to exhibit oscillations in glycolysis. The specific functions of the two subunits are currently not known, but both carry regulatory and catalytic functions . A comparison of NADH measurements in the wild-type BY4743 strain and the strain with deletion of the PFK1 gene is shown in Fig. 4. We were unable to observe oscillations in glycolysis with either of the two deletion strains. This is at variance with the results obtained by Williamson et al.  who obtained oscillations with the pfk1Δ strain but not with the pfk2Δ strain. For the pfk1Δ strain we tested the activity of hexokinase and PFK in a cell extract and the results are listed in Table 1. For hexokinase the activity is essentially the same as in the wild-type BY4743 strain, but for PFK the activity was reduced to 25% of the activity in the wild-type. Thus, PFK may also be a strong regulator of the oscillatory activity of glycolysis.
In previous publications [18, 19, 22, 27] we have provided evidence that various ATP-dependent proton pumps also control the glycolytic oscillations. The pumps hydrolyse large amounts of ATP produced in glycolysis. This can be observed as a slow decline in intracellular ATP concentration because ATP hydrolysis apparently exceeds ATP production [19, 27]. In particular the plasma membrane proton pump Pma1p and the F1F0 ATPase, operating in reverse, contribute to this ATP hydrolysis. This evidence was obtained using inhibitors and yeast strains lacking functional mitochondria and F1F0 proton pumping activity as well as the S. cerevisiae strain RS-72, where the natural constitutive promoter of yeast PMA1 has been replaced by the galactose-dependent GAL1 promoter . When these cells are transferred from galactose to glucose medium the cells stop growing when the preformed ATPase is diluted to 20% of normal, and the oscillations in glycolysis vanish due to a decreased ATP hydrolysis rate [19, 22]. Mutants with deletions of the PMA1 gene are not viable. However, the many isogenic deletion mutants available for BY4743 enabled us to perform more detailed studies of the effect on glycolytic oscillations of deletions of subunits of F1F0 ATPase. An example is shown in Fig. 5 for a strain with a deletion of the gene encoding Atp1p, the α-subunit of the F1 part of the F1F0 ATPase. In addition to NADH fluorescence we have measured the fluorescence of the dye 3,3′-diethyloxacarbocyanine iodide (DiOC2(3)), which previously was shown to monitor the Δψm [17, 18]. For comparison we have also included the corresponding graphs for the wild-type BY4743 strain. We note that the glycolytic oscillations are absent in the atp1Δ strain. However, the observation of a membrane potential, although much reduced compared with the potential in the wild-type, could indicate that some ATPase activity is retained in the atp1Δ strain, even though this strain is unable to synthesize ATP through F1F0. Note also that the increase in Δψm only occurs after addition of glucose and KCN. If only glucose is added we still observe an increase in Δψm. Measurements of the intracellular ATP concentration using the ATP nanosensor revealed an initial ATP concentration of 2–2.5 mm before addition of glucose and KCN (data not shown). Similar results to those shown in Fig. 5 for atp1Δ were obtained with strains with deletions of genes encoding Atp2p, the β-subunit of the F1 part of the F1F0 ATPase, and Atp5p, which is subunit 5 of the stator stalk of F1F0 ATPase. We also tested the importance of an intact ADP/ATP exchange across the mitochondrial membrane by studying a strain with deletion of the gene encoding the Aac1p transporter. Here we observed glycolytic oscillations very similar to the wild-type BY4743 strain (see Fig. S5).
Finally, we tested if deletions of genes encoding components of the respiratory chain, and in particular cytochrome c oxidase, would have an effect on the glycolytic oscillations. Here we studied a strain with a deletion of the gene encoding Cox6p, subunit VI in mitochondrial cytochrome c oxidase. Our expectation was that deletions of genes encoding cytochrome c oxidase or other respiratory enzymes would not affect the oscillating activity of glycolysis. Nevertheless, as shown in Fig. 6 deletion of the COX6 gene results in a loss of oscillatory behaviour. Unlike the mutants with reduced or vanishing F1F0 ATPase activity the cox6Δ mutant seems to have the same F1F0 ATPase activity as the wild-type BY4743 strain, since the signal due to Δψm is the same as in the wild-type BY4743 strain. To test if any cytochrome c oxidase activity was remaining in the cox6Δ strain we measured the respiratory activity of that strain and the wild-type BY4743 strain. A typical result is shown in Fig. S6. We note from the figure that the cox6Δ strain has no respiratory activity, evidencing that cytochrome c oxidase activity is indeed completely inhibited in this strain.
In the present work we have biochemically characterized three control points that are believed to be important for an oscillating glycolysis. Two of these control points are the hexokinase and PFK catalysed reactions while the third control point is cellular ATPase activity represented by Pma1p and F1F0 ATPase. An additional control point involving the enzyme glyceraldehyde 3-phospate dehydrogenase was demonstrated by Williamson et al. . Furthermore, here we have found an apparently new control of oscillations involving cytochrome c oxidase. To explain the effect of deletions of the enzymes hexokinase and PFK we tested the activities of these enzymes in cell extracts. We found that there are strong correlations between the enzyme activity and the ability to show oscillations. The most affected enzyme seems to be PFK, which drops by 75% in activity after deletion of the α-subunit (pfk1Δ) resulting in a complete stop of the oscillatory activity. Williamson et al.  obtained a different result after deletion of PFK1. Here the oscillatory activity was intact after deletion of PFK1 while deletion of PFK2 resulted in a complete stop of the oscillations. We are currently unable to explain this difference in results. It is possible that it could be due to a difference in growth conditions. In fact, both strains grew only slowly in our growth medium. Also, we found it difficult to grow the strain with deletion of PFK2 in an overnight culture such that glucose became depleted. Therefore, tests for oscillatory activity could not be made with this strain. As opposed to previous studies, where no remaining in vitro PFK activity was found after deleting either PFK1 or PFK2 [32, 33], we found a small but detectable PFK activity in the pfk1Δ strain. However, we could not measure the generation of a Δψm in the pfk1Δ strain (data not shown). Thus, the F1F0 ATPase seems to have reduced activity in this mutant. Because F1F0 ATPase competes with other ATPases, e.g. Pma1p, for ATP, a reduced activity of this enzyme suggests that Pma1p is first in line for ATP produced in glycolysis . In summary, our results demonstrate that a substantial decrease in PFK activity has dramatic effects on the dynamics of glycolysis. Even though a decrease in hexokinase activity also has an effect on the oscillations it is difficult to assess the relative contributions of hexokinase and PFK to the control of the oscillations, since the drop in hexokinase activity in the hxk2Δ strain is smaller than the drop in activity of PFK activity in the pfk1Δ strain. Together these data suggest that hexokinase and PFK play important roles in controlling the oscillations in glycolysis. It is interesting though that our results suggest that it is the maximum enzyme activities of the two enzymes that determine whether oscillations will occur or not. In order to study further the relative contributions of hexokinase and PFK to the control of the oscillations we performed simulations of a detailed model of glycolysis originally proposed by Hald and Sørensen . This model has the advantage that it has been adapted to batch experiments like those presented here, where glucose and KCN are added as a single large injection to a suspension of starved yeast cells. In our simulations we varied the maximum rate of the two enzymes hexokinase and PFK and compared the results with results obtained with the BY4743 wild-type and the hxk1Δ, hxk2Δ and the pfk1Δ mutants. The results are obtained as frequencies and amplitudes of NADH oscillations and are shown in Fig. S7. The simulations predict that the activity of hexokinase can be reduced to about 60% of maximum rate before the oscillations disappear. Therefore, there is a slight discrepancy between the model and the experiments since in the mutant hxk2Δ small oscillations are still present at a hexokinase activity of 53% of that in the wild-type. However, if we increase the Vmax of hexokinase from the original value of 51.7547 mm·min−1 reported by Hald and Sørensen  to a value of 65 mm·min−1 oscillations disappear at around 50% of that value in perfect agreement with the experimental findings. In both experiments and in the model a reduction in hexokinase activity has only a small effect on the frequency of the oscillations. This robustness of frequency distinguishes glycolytic oscillations in the intact cell from the oscillations in cell extracts where the frequency is highly variable [1, 9]. As for the activity of PFK the model predicts that its activity can only be reduced slightly before the oscillations disappear. Also here the frequency of the oscillations changes only slightly. It needs to be determined whether the disappearance of the oscillations is caused by a reduction in the glycolytic flux. At least for the hxk1Δ and hxk2Δ mutants the data of Williamson et al.  and our data indicate that the rate of glucose consumption as evidenced from the time from addition of glucose to the final increase in NADH (where glucose is used up) is also reduced. The same is true for our results with the pfk1Δ mutant where the glycolytic flux (measured as in ) is markedly reduced (data not shown).
Our results also explain the difference in oscillatory behaviour obtained with the sugars glucose, fructose and mannose. In yeast all three sugars are funnelled into glycolysis by phosphorylation catalysed by hexokinase . We have demonstrated here that the rate of phosphorylation increases from mannose to glucose to fructose, which essentially correlates with the ability of these sugars to induce oscillations in glycolysis. Thus, the higher rates of phosphorylation of fructose and glucose could explain why these sugars are better for inducing oscillations in glycolysis compared with mannose.
The results with the deletions of genes encoding F1F0 ATPase subunits, resulting in loss of oscillatory activity, confirm earlier work [18, 19, 27] indicating that this enzyme is important for the oscillations. In yeast cells where respiration is inhibited by cyanide F1F0 ATPase operates in reverse, i.e. it hydrolyses ATP while pumping protons out of the mitochondria, hereby establishing a proton gradient and a membrane potential over the inner mitochondrial membrane [17-19, 27, 36]. Other pumps such as Pma1p [18, 19] and the vacuolar V-ATPase  also contribute to the ATPase activity in the cell and to the regulation of the glycolytic oscillations. However, while our previous studies on the role of F1F0 ATPase activity in regulating oscillatory activity [18, 19, 22] were based on studies of inhibitors of the ATPase and mutants devoid of functional mitochondria, the present study has enabled us to study mutations in individual subunits of the enzyme. The results of these studies indicate that an intact F1F0 ATPase seems to be necessary for glycolytic oscillations to occur as previously suggested [18, 19, 22]. On the other hand, deletions of genes encoding these subunits do not seem to completely inhibit the ability to generate a Δψm. Whether this is because some F1F0 ATPase activity remains or because other processes contribute to the generation of the membrane potential needs to be resolved. Our data suggest that F1F0 ATPase activity is controlled by glucose as Δψm is only generated when glucose is present. ATP alone cannot drive F1F0 ATPase activity as our measurements of intracellular ATP reveal concentrations of 2–2.5 mm in the starved cells before addition of glucose.
Deletion of the AAC1, the gene encoding the mitochondrial ADP/ATP antiporter Aac1p, did not have any effect on the glycolytic oscillations. Aac1p is a mitochondrial inner membrane ADP/ATP translocator, which exchanges cytosolic ADP for mitochondrially synthesized ATP. The result is not surprising as Aac1p is a minor isoform while Aac2p is the major ADP/ATP translocator . Unfortunately an isogenic homozygous diploid Aac2Δ mutant could not be obtained from EUROSCARF (Frankfurt am Main, Germany). Nevertheless, these results demonstrate that an intact ADP/ATP exchange over the inner mitochondrial membrane is not a prerequisite for glycolytic oscillations.
It comes as a surprise that a strain with deletion of COX6, the gene coding for subunit VI of cytochrome c oxidase, also displays no oscillations. It is generally assumed that cytochrome c oxidase does not contribute to the oscillations since this enzyme is 100% inhibited by the presence of the high concentration of cyanide. The role of cyanide is twofold: (a) it serves to arrest respiration through inhibition of cytochrome c oxidase and (b) it reacts with acetaldehyde released by the cells, maintaining a low concentration of acetaldehyde in the external solution [38, 39]. Thus, a deletion of cytochrome c oxidase should not affect the oscillations. One possible explanation for the effect of lack of cytochrome c oxidase could be that the COX6 deletion results in the cells producing fewer mitochondria. However, this seems not to be the case since the cox6Δ strain generates a Δψm of similar magnitude to the wild-type BY4743 strain. We have currently no explanation for the effect of lack of cytochrome c oxidase on the glycolytic oscillations.
In summary, our results confirm earlier work [18-20, 22-24] suggesting that, in addition to processes distributed over the entire glycolytic pathway, glycolytic oscillations are also regulated by several processes outside this pathway. The experimental data may serve as critical tests of new and existing detailed models of glycolysis. This is demonstrated by for example the comparison of our experimental measurements of hexokinase activity and those assumed in the detailed model by Hald and Sørensen . Quantitative data like those reported here are important as the current detailed models of glycolysis in intact yeast cells [11-14] have only been constructed and validated on the basis of a few limited data sets.
The aptamer switch probe S10 was synthesized by VBC Biotech (Vienna, Austria). Texas Red Dextran was obtained from Life Technologies Europe (Nærum, Denmark). DiOC2(3) was purchased from Molecular Probes (Eugene, OR, USA). Zymolyase from Arthrobacter luteus (20T) was obtained from AMSBIO (Abingdon, UK). Hexokinase from yeast was purchased from USB Products (Cleveland, OH, USA) while pyruvate kinase (PK) from rabbit muscle was obtained from Sigma (Munich, Germany). All other chemicals were purchased from Sigma.
Aptamer switch probes and nanosensors
The ATP-binding aptamer sequence S10 from a previous paper  was converted into an aptamer switch probe, which can signal the presence of ATP, by adding a polyethylene glycol spacer (36 ethylene glycol molecules) at the 3′-end with an extension of seven nucleotides which are complementary to the 5′-end of the aptamer sequence using the principle proposed by Tang et al. . A fluorophore (Alexa Fluor 488) was attached at the 5′-end and a quencher (Black Hole 1) was attached at the 3′-end. The sequence of the aptamer switch probe is as follows: Alexa Fluor 488-GTAGTAAGAACTAAAGTAAAAAAAAAATTAAAGTAGCCACGCTT-[CH2-CH2-O]36-TTACTAC-Black Hole 1.
Polyacrylamide nanoparticles were prepared by an inverse microemulsion polymerization reaction modified from  in the presence of the ATP aptamer switch probe. Briefly, 3.08 g of dioctyl sulfosuccinate (AOT) and 1.08 g of Brij 30 were dissolved in 43 mL of hexane and deoxygenated by sonication for 1 h with a flow of argon into the gas head space above the liquid. Meanwhile 1.35 g of acrylamide and 0.4 g of N,N-methylenebisacrylamide were dissolved in 4.5 mL of 10 mm sodium phosphate buffer, pH 7.25. Next, 150 μL of a 100 μm aptamer switch probe solution and 15 μL of 50 mg·mL−1 Texas Red Dextran were added to 2 mL of this monomer solution. Finally, 2.0 mL of the solution containing monomers and dyes was added dropwise to the hexane solution and left for 20 min under argon to form the microemulsion. Polymerization was initiated by addition of 50 μL of a 10% (w/v) solution of sodium bisulfite and allowed to proceed for 3 h. The hexane was removed in vacuo and the nanosensors were precipitated by the addition of 100 mL ethanol. The suspension was transferred to an Amicon ultrafiltration cell model 2800 (Millipore Corp., Bedford, MA, USA), filtered through a 100 kDa filter and washed with 4 × 100 mL of ethanol to remove unreacted monomers and surfactants. The particles were resuspended in 50 mL ethanol, filtered (0.025 μm nitrocellulose filter membrane) and dried in vacuo. Vacuum dried particles were kept at −20 °C until they were used in experiments. The sensors are stable under these conditions for at least 12 months.
Dynamic light scattering measurements
The sensor size was determined by dynamic light scattering. The measurements were done in a Brookhaven Instruments goniometer setup (BI-200SM) equipped with a Melles Griot He-Ne laser, λ = 632.8 nm. The measurements were performed at a single scattering angle, 90°, in self beating detection mode. Cylindrical glass cuvettes, with an outer diameter of 10 mm, were embedded within an index matching and thermostatted bath. The glass cuvettes were autoclaved prior to use. The samples were allowed to temperature stabilize before and sample temperature was maintained at 25 °C during the measurements. The autocorrelation function generated from the intensity fluctuations was analysed to obtain the scattered intensity versus the distribution of particle sizes using multiple pass non-negative least squares . For the dynamic light scattering measurements the sensors were suspended to a concentration of 1 mg·mL−1 in the same buffer as the kinetic experiments (10 mm phosphate buffer, pH 7.0; 50 mm Na2SO4; 5 mm MgCl2). Figure S1 shows the size distribution of the sensor. The average diameter of the sensor was determined as 34 ± 9 nm (n =6).
Sensor response time and ATP binding curves
Previously, we determined the response time to changes in ATP concentration (10–90%) to be less than a second . Figure S2A shows the change in fluorescence upon addition of first ADP and then an equimolar concentration of phosphoenolpyruvate (PEP) followed by PK to a solution containing the ATP sensor and glucose. The increase in fluorescence following addition of PK is due to the transfer of a phosphate from PEP to ADP and this process is essentially irreversible. Thus, the resulting increase in fluorescence is determined by the concentration of ADP and PEP in the solution. When the fluorescence has stabilized hexokinase is added to the solution and the fluorescence decreases again to the original level before addition of PEP and PK due to the transfer of phosphate from ATP to glucose. This implies that the response of the ATP sensor is reversible, and therefore curves like those in Fig. S2A, where different equimolar concentrations of ADP and PEP are added, can be used to construct binding curves of the sensor for ADP and ATP. The fluorescence signals due to Alexa Fluor 488 (F520) were normalized to the fluorescence of Texas Red (F610) using the equation
Figure S2B shows such binding curves for ATP and ADP constructed by plotting ϕA against the concentration of ADP or ATP for the sensor containing the S10 aptamer switch probe. Note that the nanosensor is not absolutely specific for ATP, but the affinity of the sensor is much lower for ADP compared with ATP. Furthermore the change in fluorescence ratio essentially depends linearly on the concentrations of ADP and ATP in the range 1–6 mm. This also implies that if the total concentration of ADP plus ATP is constant we can determine the concentrations of ADP and ATP from any fluorescence ratio ϕA by solving a set of two linear equations. The nanosensor does not bind AMP or any other purine or pyrimidine nucleotide as shown previously .
The binding curves were fitted to linear equations using linear regression (SigmaPlot, Version 11, San Jose, CA, USA). It should be emphasized that the relative amounts of aptamer switch probe and the reference dye (Texas Red Dextran) may vary from one batch of sensors to another, i.e. ϕA may vary from batch to batch, and therefore a new calibration curve must be constructed for each new batch of sensors. The calibration curves were used to calculate the ATP concentrations shown in Figs 2 and 3 from corresponding fluorescence measurements.
Yeast strains and growth
The yeast strains used in this study are listed in Table 2. The strains, all obtained from EUROSCARF (Frankfurt am Main, Germany), were grown essentially as described by Poulsen et al. , i.e. under semiaerobic conditions at 30 °C on a rotary shaker, 180 rpm, in a medium containing 10 g·L−1 glucose, 6.7 g·L−1 yeast nitrogen base (Bacto) and 100 mm potassium phthalate (Sigma-Aldrich, Steinhem, Germany) at pH 5.0. The medium was further supplemented with 60 mg·L−1 histidine, 60 mg·L−1 methionine, 80 mg·L−1 leucine, 80 mg·L−1 lysine and 80 mg·mL−1 uracil. The yeast was harvested at the point when glucose was depleted as measured with a glucose test strip (Macherey-Nagel, Düren, Germany). The cells were washed twice with 100 mm potassium phosphate buffer (Merck, Darmstadt, Germany), pH 6.8 (centrifugation, 5 min at 4066 g, GSA, Sorvall, Newtown, CT, USA), resuspended in the same buffer to a cell density of 10% by weight and starved for 3 h on a rotary shaker at 30 °C.
Table 2. Yeast strains used in this study
MAT a/α; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; lys2Δ0/LYS2; MET15/met15Δ0; ura3Δ0/ura3Δ0
BY4743; MAT a/α; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; lys2Δ0/LYS2; MET15/met15Δ0; ura3Δ0/ura3Δ0; BY4743; YFR053C::kanMX4/YFR053C::kanMX4
BY4743; MAT a/α; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; lys2Δ0/LYS2; MET15/met15Δ0; ura3Δ0/ura3Δ0; BY4743;YGL253w::kanMX4/YGL253w::kanMX4
BY4743; MAT a/α; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; lys2Δ0/LYS2; MET15/met15Δ0; ura3Δ0/ura3Δ0; BY4743;YGR240c::kanMX4/YGR240c::kanMX4
BY4743; MAT a/α; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; lys2Δ0/LYS2; MET15/met15Δ0; ura3Δ0/ura3Δ0; BY4743;YMR205c::kanMX4/YMR205c::kanMX4
BY4743; MAT a/α; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; lys2Δ0/LYS2; MET15/met15Δ0; ura3Δ0/ura3Δ0; BY4743;YYBL099w::kanMX4/YBL099w::kanMX4
BY4743; MAT a/α; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; lys2Δ0/LYS2; MET15/met15Δ0; ura3Δ0/ura3Δ0; BY4743;YJR121w::kanMX4/YJR121w::kanMX4
BY4743; MAT a/α; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; lys2Δ0/LYS2; MET15/met15Δ0; ura3Δ0/ura3Δ0; BY4743;YDR298c::kanMX4/YDR298c::kanMX4
BY4743; MAT a/α; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; lys2Δ0/LYS2; MET15/met15Δ0; ura3Δ0/ura3Δ0; BY4743;YMR056c::kanMX4/YMR056c::kanMX4
BY4743; MAT a/α; his3Δ1/his3Δ1; leu2Δ0/leu2Δ0; lys2Δ0/LYS2; MET15/met15Δ0; ura3Δ0/ura3Δ0; BY4743;YHR051w::kanMX4/YHR051w::kanMX4
Preparation of yeast extracts and measurements of hexokinase and phosphofructokinase activities
Yeast cells were first diluted to 1% by weight in NaCl/Pi buffer + 2 mm MgCl2 and incubated with 0.1 mg·mL−1 of Zymolyase at 35 °C for 60 min. The cells were then washed twice with buffer and centrifuged for 3 min at 5000 r.p.m. in an Eppendorf centrifuge and resuspended to a concentration corresponding to 20% by weight in a solution containing 10 mm potassium phosphate, pH 7.0, 50 mm Na2SO4, 5 mm MgCl2. The cells were then treated with ultrasound (Branson 5510) for 30 min and centrifuged in an Eppendorf centrifuge at 7000 r.p.m. for 5 min. For measurements of enzyme activities an aliquot of the supernatant was added to a solution containing 10 mm potassium phosphate, pH 7.0, 50 mm Na2SO4, 5 mm MgCl2, 0.5–5 mm ATP and ATP sensor corresponding to 1 mg·mL−1. For measurements of hexokinase activity 20 mm glucose was added to this solution while for measurements of PFK activity we added instead 10 mm fructose 6-phosphate. The cells' protein content was determined according to Lowry et al. .
Measurements of intracellular ATP concentration
The intracellular ATP concentration was measured using the ATP nanosensor as described previously [27, 28].
Measurements of NADH and mitochondrial membrane potential (Δψm)
Experiments were performed in an Edinburgh FL 920 spectrofluorometer using a 2 mL sample cell with constant stirring and temperature control (Quantum Northwest, Liberty Lake, WA, USA). The samples contained yeast cells with or without inserted DiOC2(3). The fluorescence signal due to NADH was measured at 450 nm (slit 10 nm) with excitation at 366 nm (slit 3 nm). Measurements of fluorescence due to DiOC2(3) (Δψm) were done with excitation at 480 nm (slit 3 nm) and emission at 600 nm (slit 3 nm) using a 5 μm probe as described in . Yeast cells at a density of 10% by weight were suspended in 100 mm potassium phosphate buffer, pH 6.8. The temperature around the sample cell was maintained at 25.00 ± 0.01 °C. All measurements were done in triplicate.
Measurements of respiration rates
Respiration rates of a suspension of intact yeast cells following addition of sugars or other substrates were measured in a Hansatech Oxygraph (King's Lynn, UK).
This research was supported by a grant from the Danish Research Council for Technology and Innovation (project NaKIM, grant no. 09-052140) and the Lundbeck Foundation (NanoCAN).