Influence of cyanide on diauxic oscillations in yeast



B. O. Hald, Department of Biomedical Health Sciences, University of Copenhagen, Blegdamsvej 3, 2200 Copenhagen, Denmark

Fax: +45 35327418

Tel.: +45 35327401



Coherent glycolytic oscillations in Saccharomyces cerevisiae are a multicellular property induced by addition of glucose to a starved cell population of sufficient density. However, initiation of oscillations requires an additional perturbation, usually addition of cyanide. The fate of cyanide during glycolytic oscillations has not previously been studied, and is the subject of the present paper. Using a cyanide electrode, a substantial decrease in cyanide concentration was observed. In the pH range 6–7, we found experimentally that the electrode behaves reasonably well, provided changes in pH are taken into account. To our knowledge, use of a cyanide electrode to study cyanide dynamics in living biological systems is new. Cyanide was found to enter starving yeast cells in only negligible amounts, and did not react significantly with glucose. Thus, cyanide consumption must be explained by reactions with glycolytic intermediates and evaporation. Evaporation and reaction with the signalling substance, extracellular acetaldehyde (ACAx) only explains the observed cyanide removal if [ACAx] is improbably high. Furthermore, differences in NADH traces upon cyanide addition before or after glucose addition strongly suggest that cyanide also reacts with intracellular carbonyl-containing metabolites. We show that cyanide reacts with pyruvate (Pyr) and dihydroxyacetone phosphate in addition to ACA, and estimate their rate constants. Our results strongly suggest that the major routes of cyanide removal during glycolysis are reactions with pyruvate and ACA. Cyanide removal by all carbonyl-containing intermediates led to a lower mean [ACAx], thereby increasing the amplitude of [ACAx] oscillations.


The mathematical model described here has been submitted to the JWS Online Cellular Systems Modelling Database and can be accessed at free of charge.




extracellular acetaldehyde


dihydroxyacetone phosphate




hydrogen cyanide




phosphate-buffer solution




Synchronous glycolytic oscillations in a stirred suspension of yeast cells have been studied for decades in open as well as in closed systems [1-3]. The occurrence of oscillations in a metabolic network is a system property that involves many metabolites and reactions. Quenching and bifurcation experiments on open and closed yeast cell systems [3-5] confirm that the synchronized suspension behaves as a single cell floating in the extracellular medium, and that the synchronization is mediated through oscillations in the extracellular acetaldehyde concentration ([ACAx]) [6, 7].

Addition of cyanide to the yeast suspension produces qualitative changes in the oscillatory behavior – a property that has been known since the original discovery of glycolytic oscillations. Cyanide is regularly used as a perturbation to induce oscillations. Moreover, both the amplitude and duration of oscillations are optimized in the concentration range 3 mm < [KCN] < 7 mm [5, 8]. However, cyanide is not essential for the occurrence of oscillations [9], as they have been observed in cyanide-free systems of Saccharomyces cerevisiae by bubbling with argon [10]. Presumably bubbling removes acetaldehyde (ACA) by evaporation. On the basis of modeling synchronization in a coupled Stuart–Landau system, it is known that increasing the amplitude and stiffness of oscillations also increases synchronization [11]. This is in accordance with the observation that addition of cyanide to the yeast suspension in a closed system increases the amplitude and duration of the oscillations.

Despite the effect of cyanide on glycolytic oscillations, no detailed studies of the influences of cyanide other than the reaction with ACAx [8] have so far been reported. We have previously described the influence of hydrogen cyanide (HCN) on oscillations of starved yeast cells [4]. In this paper, we investigate in more detail the influence of HCN on yeast metabolism during glycolytic oscillations. Using a cyanide electrode, we determined (a) the HCN uptake by starved cells, (b) the HCN removal during oscillations, (c) the rate of HCN evaporation, (d) rate constants of extracellular HCN reactions with ACAx and Glc under conditions as close as possible to the experimental conditions for the yeast experiments, and (e) rate constants of HCN reactions with dihydroxyacetone phosphate (DHAP) and pyruvate (Pyr) in a cytosol-like buffer. The results were used in a previously developed model for transient oscillations [4] that was able to accommodate the experimental results by minor adjustments of enzyme activities and formation of cyanohydrins by reaction with glycolytic metabolites (see Fig. 1).

Figure 1.

Network structure of the main catabolic pathways of Glc-fed yeast cells. The primary reactions include glycolysis, as well as membrane transport, fermentative and extracellular reactions [4]. Subscript ’x‘ denote species in the extracellular compartment. Black and blue reactions constitute the main intracellular and transport reactions, respectively. Red reactions are the main HCN reactions that may take place both intra- and extracellularly. Enzyme names are shown in green.


The original rationale behind cyanide addition to a yeast suspension was to prevent oxidative metabolism in the mitochondria. As described previously [4], this is indeed the case, but, for the experimental protocol used in the present investigation, the concentration of O2 decreases to zero soon after preparation of the cell suspension. The evaporation of gases from the suspension combined with a partially covered cuvette ensure that only a minor amount of O2 interferes with the system during the experiment. Moreover, a higher cyanide concentration is required to initiate oscillations compared to inhibition of respiration, and other inhibitors of respiration do not induce sustained oscillations [12]. Finally, the characteristics of the oscillations depend on the cyanide concentration. If the concentration is too high, the oscillations disappear altogether [8]. The effects of cyanide on glycolytic oscillations must therefore be explained by other cyanide reactions.

The buffer used for the yeast experiments is a 0.1 m phosphate buffer adjusted to a pH of 6.9. As the pKa of HCN is 9.2, most of the added cyanide will be present as HCN. HCN reacts with carbonyl groups, producing cyanohydrins. Another fate of HCN is evaporation. These processes must be included quantitatively in the model.

The most important of the carbonyl reactions of HCN is thought to be the formation of lactonitrile [CH3CHCNOH, (lacto)] by reaction with ACA (CH3CHO) [13]:

display math(1)

The reactions proceed intra- as well as extracellularly, and are not elementary reactions. In accordance with Yates and Heider [14], the forward and reverse reaction rates are

display math(2)
display math(3)

Introducing math formula and Kw = [H+][OH], we get

display math(4)
display math(5)

showing that both reactions are fast at high pH and slow at low pH. The equilibrium constant becomes

display math

This mechanism predicts reaction rates that depend strongly on pH and an equilibrium constant that is independent of pH. We have thus estimated kf by experimental measurements in the same buffer as used for the yeast experiments, and then calculated kr as math formula. The same mechanism was assumed in reactions of HCN with DHAP or Pyr.



Evaporation of HCN was followed using the cyanide electrode as shown in Fig. 2 and Fig. S1. We ascertained that the drift in [HCN] measurements was due to evaporation of HCN by bubbling the HCN/PBS solution with nitrogen (Fig. S2). However, over a long time scale (approximately 3 h), electrode drift contributed to signal variations (see Fig. 2).

Figure 2.

Evaporation profiles of HCN. Pulses of KCN were added to the 0.1 m PBS buffer under conditions equivalent to the usual yeast suspension experiments. The temporal evolution of total [HCN] was then followed for approximately 14 h. Within the first approximately 8000 s, evaporation of HCN was well described by an exponential decay with a decay rate of 4.8 × 10–5 s−1, as indicated by the dotted lines from 0 to 10 000 s.

In our experimental set-up, the yeast suspension is in direct contact with a small volume of air. Diffusion of gases from this volume to the surroundings is impeded by the experimental set-up but cannot be avoided entirely. Evaporation followed two slightly different exponential decays before and after approximately 2 h subsequent to HCN addition. In our experimental set-up, the initial exponential evaporation rate (see Fig. 2) was fitted as:

display math(6)

[ACA] was assumed to evaporate proportionally to HCN evaporation according to the ratio of their respective Henry constants.

[HCN] in yeast cells

The ratio of volumes, math formula, is fixed by the chosen cell density [15]. Thus, most HCN is expected to be present in the extracellular medium. However, intracellular components may adsorb or react with the added HCN. Using the cyanide electrode, we tested whether HCN was taken up by non-metabolizing cells (Table 1). At 5, 25 and 45 min after addition of cyanide, we compared the yeast suspension with a yeast supernatant obtained by centrifugation and a control without yeast cells. As no differences were observed, the data were pooled. The [HCN] in the yeast suspension and supernatant are equal, indicating that no HCN adheres to the cell wall or intracellular components. Compared with the control, the [HCN] values in the yeast suspension and the supernatant are slightly higher, probably due to the finite volume occupied by the cells. Thus, only tiny quantities of HCN enter the non-metabolizing cells.

Carbonyl compounds and HCN

Nucleophilic addition reactions between HCN and aldehydes/ketones occur spontaneously in nature (reactions fastest at pH of approximately 4–5).


The Glc + HCN reaction is well-known (e.g. Kiliani–Fischer synthesis): KCN + C6H12O6 + 2H2O → OH · CH2(CH · OH)5CO2K + NH3, forming the non-toxic potassium-salt of α-glucoheptonic acid. Using the cyanide electrode, we tested whether an instantaneous increase of [Glc] to 24 mm Glc in a 5 mm [HCN] solution in 0.1 m PBS resulted in the expected reaction. However, no significant changes were found in measured [HCN] in the pH range 6.0–7.2, indicating that α-glucoheptonate formation is negligible in our experimental set-up (Fig. S3). It is probable that fast formation of the glucose hemiacetal (in aqueous solution, <0.1% is in the open-chain form of glucose) and the low amounts of nucleophilic CN prevent a rapid reaction.


Lactonitrile formation (HCN + ACA ⇌ lacto) has previously been shown to occur during oscillations [8]. Using Eqns (2) and (3), we fitted rate constants to the HCN profiles of the lactonitrile reaction at pH 6.9. At pH 6.9, the forward reaction clearly dominates under our experimental conditions (Fig. S4A). The fitted reaction rate constants are shown in Table 2. kr is poorly determined by the forward ACA + HCN reaction. The variance depends partially on initial concentrations, suggesting that the proposed mechanism is incomplete.

Table 1. HCN entry into yeast cells. The initial [HCN] was 0.79 mm in each preparation (= 8)
 Control (0.1 m PBS)Yeast suspensionSupernatanta
  1. a

    Obtained by separation of cells by centrifugation.

Cmeasured0.80 ± 0.020.84 ± 0.030.84 ± 0.02

In the model, we used kf = 5.9 × 10−4 mm−1·s−1 and kr = 3.3 × 10–8 s−1, calculated using math formula [16]. Our experimental data are consistent with this choice.

Pyr and DHAP

DHAP, glyceraldehyde-3–phosphate, phosphoenolpyruvate, Pyr and ACA are the only intermediates during glycolysis that contain a functional group (carbonyl group or alkene) that can be attacked by the nucleophile, CN. However, only DHAP and Pyr are present in high concentrations, i.e. approximately 2.5 and 8.7 mm, respectively, whereas the glyceraldehyde-3–phosphate concentration is approximately 0.1 mm and the phosphoenolpyruvate concentration is approximately 0.04 mm [6, 15]. The rate constants of DHAP and Pyr were estimated by changing to a cytosol-like buffer [17] (Table 2). The reaction with DHAP was noisy as evidenced by the large standard deviation.

Yeast suspension measurements


The fermentation of Glc in yeast mainly produces ethanol, CO2 and H2O, as well as MgATP2− (‘free energy’). MgATP2− cannot accumulate in the cells and must be hydrolyzed for fermentation to proceed. The mitochondrial membrane has been suggested as a primary location for MgATP2− hydrolysis [18].

As mentioned, the cyanide electrode is sensitive to pH changes in the interval 6–7 [19]. The fermentation process does not produce protons intracellularly, but hydration of CO2 leads to a detectable decrease in pH (CO2 + H2O ⇌ H2CO3, where kf = 0.039 mm−1·s−1 and kr = 23 s−1). The PBS buffer is important in preventing a precipitous decrease in the pH during fermentation as shown in Fig. 3A in response to two consecutive Glc additions. Acidity increases almost linearly during fermentation, indicating a steady flux though glycolysis in the oscillatory phase. Note that protons do not accumulate instantaneously after Glc addition. A lag time of approximately 60 s in the first run and approximately 120 s in the second run (influenced by the HCN ⇌ cyanohydrin equilibrium) is apparent. This may be due to a slow CO2 hydration process and/or may be associated with a regulation of glycolysis (as suggested previously [4]).

Figure 3.

Comparative traces of (A) NADH versus pH, and (B) NADH versus [HCN]. Representative traces for experiments in which a Glc pulse (Δ[Glc] = 24 mm) followed by a KCN pulse (Δ[KCN] = 4.8 mm) were added to the yeast suspension sequentially in pairs (g1,c1) and (g2,c2). (A) Simultaneous recording of extracellular pH and NADH fluorescence (arbitrary units). Note the slight delays in pH decrease upon administration of Glc (vertical black bars) of approximately 60 s and approximately 120 s, respectively. (B) Simultaneous recording of [HCNx] and NADH fluorescence. The different amplitudes of the NADH traces are due to minor disturbances in the positioning of the optical fiber.


The primary objective of the present study was to determine the fate of HCN upon diauxic glycolytic oscillations. Addition of HCN to the yeast suspension elicited a small temporary increase in [NADH] [4]. However, the equipment used did not enable detection of a decrease in extracellular HCN ([HCNx]) compared to the control solution of 0.1 m PBS (Table 1), confirming that only a small fraction of HCN actually enters or adheres to non-metabolizing yeast cells at the given pH.

Table 2. Experimentally determined reaction rates for X + HCN math formula cyanohydrin
XnBufferkf (mm·s)−1kr (s−1)
ACA6PBS5.9 × 10−4 ± 2.7 × 10−51.3 × 10−3 ± 8.7 × 10−4
Pyr3PBS4.3 × 10−4 ± 2.2 × 10−56.5 × 10−5 ± 1.7 × 10−5
Pyr3Cytosol-like5.6 × 10−4 ± 4.7 × 10−56.2 × 10−5 ± 2.8 × 10−5
DHAP3Cytosol-like6.2 × 10−4 ± 2.9 × 10−42.0 × 10−5 ± 2.0 × 10−4

After addition of Glc, the HCN response of metabolizing yeast cells is markedly different. As shown in Fig. 3B, HCN was added approximately 200 s after the Glc pulse. After the [HCN] spike (an artefact from the electrode), [HCN] decreases almost linearly from approximately 4.5 mm to 2.75 mm. The rate of HCN removal is approximately 2.0 μm·s−1 (in both runs). The end of the linear [HCN] decrease matches exactly the cessation of NADH oscillations. Glc depletion is associated with a small increase in [HCN], which agrees well with the HCN ⇌ lactonitrile equilibrium being pushed towards an [HCN] yield. Given another pulse of Glc, the [HCN] trace shows a similar pattern of events. The small oscillations visible from the [HCN] trace are an artefact resulting from the ‘staircase’ appearance of measured pH, which is a consequence of the resolution of the equipment (see Fig. 3A). However, as expelled CO2 probably oscillates, it is possible that small non-detectable oscillations in pH are super-imposed on the seemingly linear decrease. Compared with the small amplitude of extracellular ACA oscillations, [HCN] is too large and the cyanide electrode is too insensitive to detect oscillations in [HCN].

Cyanide addition

Reversing the usual order of KCN addition influences the dynamics, as shown in Fig. 4. Addition of KCN after Glc (Fig. 4A) yields essentially the same result as in Fig. 3B. However, when KCN is added before Glc (Fig. 4B), HCN is removed fast initially upon Glc addition (0.5 mm), and NADH oscillations are not evident before the second Glc pulse. In addition, the width of the initial NADH spike is much broader compared with Fig. 4A, i.e. the re-oxidation of NADH takes longer when HCN is present. As NADH is primarily re-oxidized by reducing ACA to EtOH, any delay in the production of ACA is equivalent to a slower re-oxidation. This indicates that HCN reacts with glycolytic intermediates, most probably DHAP and Pyr (see Table 2). The fast HCN removal upon Glc addition supports this interpretation. The model (see below) reproduces this behavior in the initial NADH spike (Fig. 5C and Fig. S5). The fact that no oscillations turn up on the first run may be a consequence of the lack of a synchronizing perturbation. If the cells oscillate out of phase, no overall oscillations are observed. When KCN is added after Glc, it acts as a resetting signal that globally lowers [ACA]x, and thus synchronizes the individual oscillators. In the second run, the glycolytic intermediates have been replenished, and the Glc pulse itself may acts as the synchronizing perturbation.

Figure 4.

Representative traces of NADH fluorescence (grey, arbitrary units) and [HCN] (black, mm). KCN pulses (c1,c2) (Δ[KCN] = 4.8 mm) were added to the yeast suspension after (A) or before (B) the Glc pulses (g1,g2) (Δ[Glc] = 24 mm). Note the wider NADH spikes, the higher initial [HCN] level, and the lack of oscillations in (B) compared to (A).

Figure 5.

Simulation results. (A) Dotted and solid traces: full model simulations. Extracellular [HCN] without (dotted) and with (solid) inclusion of intracellular HCN drains. In both simulations, 5 mm HCN and 24 mm Glc were added at = 0 s. Dashed trace: simulation of model with only the extracellular lactonitrile reaction and evaporation. These reactions only explain the observed HCN consumption if [ACAx] is set to 0.5 mm for the first 1400 s. (B) Traces of HCN-consuming fluxes: Pyr reaction (dotted), evaporation (dashed/dotted), ACA reaction (dashed) and the DHAP reaction (solid). Note that the flux is relative to the extracellular volume. (C) [NADH] traces after addition of 24 mm Glc with (solid) or without (dotted) simultaneous addition of 5 mm HCN. Note that the initial NADH spike is slightly wider when HCN is added with Glc.


Incorporating kACA,f and kACA,r of the extracellular lactonitrile reaction (Eqn (1)) and HCN evaporation (Eqn (6)) into the model did not explain the measured [HCN] reduction of approximately 2 mm (see Fig. 5A, dotted trace). In fact, for these extracellular reactions to explain an HCN removal of approximately 2 mm, [ACAx] constantly has to be as high as 0.5 mm (Fig. 5A, dashed trace), further suggesting simultaneous intracellular cyanide reactions.

The estimated rate constants for the HCN reactions with Pyr or DHAP in a cytosol-like buffer were next introduced into the model, followed by adjustments of a few Vmax values to obtain oscillations in NADH (ATPase −17%, pyruvate decarboxylase −30%, Pyr drain +15%, alcohol dehydrogenase +200%), as shown in Fig. 5C. The simulations reproduce the finding that HCN addition increases the amplitude of oscillations. However, although a significant reduction in total HCN consumption was obtained, the model did not fully explain the observed consumption of approximately 2 mm (Fig. 5A, solid trace). However, this may be due to the low [ACAx] in simulations of the transient model (approximately 0.03 mm). A higher [ACAx] values would obviously result in increased cyanide consumption. Fitting a large model to transient experimental data is beyond the scope of this paper.

Nevertheless, due to the similar rate constants of the HCN reactions with Pyr, ACA and DHAP (Table 2), the model predicts that the intracellular Pyr + HCN reaction accounts for the largest removal of HCN during glycolysis (see Fig. 5B).


In the present paper, we have studied the fate of [HCN] during glycolytic oscillations in a suspension of yeast. [HCN] was measured using a cyanide electrode, and the influence of pH on the potential recordings of the electrode was taken into account using a calibration curve in the pH range 6–7. Changes in extracellular pH are induced by (a) instantaneous addition of KCN, and (b) the continuous fermentation process, through hydration of expelled CO2 (Fig. 3A). Nernstian behavior of the cyanide electrode was observed at individual pH values (Fig. S6). To our knowledge, use of a cyanide electrode to follow cyanide dynamics in living biological systems at the relevant pH is new. We found that starved yeast cells remove approximately 2 mm of [HCN] after addition of 24 mm Glc and 4.8 mm KCN.

As HCN has a low vapor pressure, evaporation must be taken into account. We found that, for relevant levels of [HCN], the evaporation could be fitted piecewise by two exponential decays before and after the first 2 h subsequent to HCN addition (Fig. 2). As the longest time scale of the glycolytic oscillations is about 30 min, we used Eqn (6) to account for evaporation in our experimental set-up.

Table 1 shows that almost no HCN was adsorbed to the cells or crossed the plasma membrane to bind to various intracellular components under glucose-free conditions. Furthermore, Glc did not react with HCN at the pH range of our experimental set-up (Fig. S3). Thus, the surprisingly large consumption of HCN during glycolysis is most likely due to reaction with glycolytic aldehydes or ketones, i.e. with ACAx extracellularly and with DHAP, glyceraldehyde-3–phosphate, Pyr and ACA intracellularly. The forward rate constant of Eqn (1), kACA,f, has been reported to be 1.5 × 10–3 mm−1·s−1 [13], i.e. in the same range as measured for our experimental set-up (Table 2 and Fig. S4A). For extracellular evaporation and lactonitrile formation alone to explain a 2 mm reduction of HCN in the time frame of glycolytic oscillations (approximately 1400 s), [ACAx] must be set to 0.5 mm constantly (Fig. 5A, dashed curve). However, [ACAx] has been reported to be below 0.1 mm [13], and our model estimated [ACAx] as low as approximately 0.03 mm.

Thus, intracellular consumption of HCN during glycolysis is strongly suggested. This was experimentally corroborated by our finding that the NADH spike is wider when KCN is added before Glc (Fig. 4). Intracellularly, only Pyr and DHAP are known to be present in the millimolar concentration range during oscillations [6, 15]. Glyceraldehyde is a well-known starting material for the Kiliani–Fischer synthesis, but the glyceraldehyde-3–phosphate concentration is too low to be an important HCN drain. This also applies for phosphoenolpyruvate. DHAP and Pyr reactions with HCN in a cytosol-like buffer at pH 6.9 were shown experimentally, and reaction rates were estimated (Table 2 and Fig. S4B,C).

Introducing evaporation (Eqn (6)) and all the estimated rate constants into a previously established model of transient glucose additions did not fully explain the measured HCN consumption during glycolytic oscillations (Fig. 5A, solid trace). However, model predictions of [ACA] and [ACAx] are on the low side, which may partially explain the insufficient HCN removal. Optimization of the full-blown transient model is a difficult problem that will be addressed in a future study. However, any carbon flux into the pentose phosphate pathway would produce aldehydes and ketones that may bind to HCN, although the flux would be expected to be small.

Intracellular cyanohydrin formation from glycolytic intermediates will cause a slight reduction in the ACA level as carbon is diverted from production of ACA (Fig. S7). Thus, these reactions will have the same overall effect as the known lactonitrile reaction, namely a reduction of [ACAx] and a simultaneous relative increase in oscillation amplitude. Although the amplitude of [ACAx] oscillations is small, ACA is thought to be the major synchronizer of oscillations, and very high [HCNx] abolish oscillations [8, 15]. These results suggest that HCN reduces intra- and extracellular [ACA] by the known extracellular lactonitrile reaction but also by intracellular cyanohydrin formation.

Finally, the model is consistent with the observation that oscillations in S. cerevisiae are not autonomous without KCN addition, suggesting that the drain of ACA in itself increases the feedback strength needed for oscillations. The reaction of HCN with intracellular metabolites may also explain the inability of our previous models and a recent model to capture the Hopf bifurcation points using KCN as the bifurcation parameter [4, 20, 21].

In conclusion, we have shown that [HCN] may be followed using a cyanide electrode in a biologically relevant pH range. Furthermore, we have shown that HCN reacts with both extra- and intracellular carbonyl groups, particularly DHAP, Pyr and ACA, during glycolytic oscillations. Elucidating the precise cause of the intracellular cyanide reactions requires further investigations.

Experimental procedures

In all experiments, the solutions were stirred using a propeller driven by a synchronous motor.

Cell preparation

Yeast cells (S. cerevisiae X2180) were grown aerobically at 30 °C in a closed-batch culture in a rotary shaker to the point of Glc depletion as described previously [22]. Harvested cells were washed twice for a few seconds in 0.1 m PBS before resuspending in 0.1 m PBS to the desired cell density. Next, the cells were starved at 30 °C for > 1 h and kept at approximately 0 °C until the start of the experiments (all experiments started within 24 h from cell preparation).

NAD(P)H measurement

NAD(P)H fluorescence was achieved by excitation with the 365 nm line from a mercury lamp, and measured as emission of light in the interval 420–660 nm using a photo-multiplier (arbitrary units) as described previously [1, 3]. In closed-system experiments, oscillations were induced by an instantaneous increase of [Glc] to 24 mm followed by an instantaneous increase of KCN to 5 mm after the initial spike in NADH.

[HCN] measurements

[HCN] in the extracellular medium was measured using a cyanide ion electrode (Cole-Parmer Instrument Co., Vernon Hills, IL, USA) at 25 °C. Electromotive force (EMF) was measured using a standard pH meter (Radiometer, Copenhagen, Denmark). The procedure recommended by the electrode manufacturer for measurements of [HCN] is to measure the potential after increasing the pH to > 11. At this pH, HCN is converted to CN and cyanohydrin is rapidly converted to CN and the corresponding carbonyl compound. We thus decided to use two protocols.

To estimate the amount of HCN taken up by yeast cells, the recommended approach was used. A calibration curve covering the range cCN ∊ [0.2 mm:1.6 mm] in 0.1 m PBS at pH 11 was constructed. Samples were prepared by adding 300 μL 0.1 m KCN to 6 mL 0.1 m PBS or 6 mL yeast suspension in 0.1 m PBS. At 5, 25 and 45 min after KCN addition, 1 mL aliquots of the PBS sample (control), the yeast suspension, and the supernatant obtained by centrifugation (18 000 g for 1 min at room temperature) of the yeast sample were obtained, and diluted sixfold using 0.1 m PBS. Then 60 μL 10 m NaOH was added to a pH of 11. Total cyanide was then measured.

For measurements of intact HCN during glycolytic oscillations without interfering with the lactonitrile reaction (Eqn (1)), we calibrated the electrode in 0.1 m PBS as described previously [19, 23]. Calibrations were performed using combined potentiometric (EMF) and pH measurements in PBS containing a range of [H+] and [HCN] concentrations. From a series of (EMF, pH, [HCN]) points, a bivariate cubic spline curve was fitted, yielding a calibration curve as shown in Fig. 6.

Figure 6.

2D calibration curve for the cyanide electrode. Simultaneous measurements of pH and electromotive force (EMF) upon successive additions of 0.1 m KCN pH was adjusted by addition of either 2 m H2SO4 or 2 m KOH) led to a series of points that could be fitted to bivariate cubic spline curves.

For absolute determinations of [HCN], the first protocol performs best, whereas the second protocol is suitable for following [HCN] changes. It should be noted that a particular calibration curve is only valid for one working day, and should be checked every 2 h (both protocols). We confirmed that the electrode behaved in an apparently Nernstian manner under these conditions (Fig. S6).

Rate of lactonitrile reaction

The rate constants in Eqn (1) were estimated by using the cyanide electrode to follow [HCN] in mixtures of ACA and HCN in 0.1 m PBS. Initial values of [ACA] and [HCN] were either 3 or 5 mm.

Rate of HCN reactions with Pyr or DHAP

Estimation of intracellular rate constants for the Pyr or DHAP reactions with HCN were performed as for the ACA reaction, except that the 0.1 m PBS buffer was exchanged with a cytosol-like buffer containing 245 mm potassium glutamate, 2 mm MgSO4, 0.5 mm CaSO4, 10 mm NaH2PO4, 40 mm KH2PO4 and 15 mm KOH [17]. Li2DHAP and Pyr were purchased from Sigma (St. Louis, MO, USA) and Boehringer Mannheim GmbH (Mannheim, Germany), respectively.

Glycolytic oscillations

Experiments were performed in a termperature-controlled (25 °C) cuvette using 8 mL yeast cell suspension with a dry weight of 12.8 mg·mL−1. A large cuvette was necessary to measure NADH fluorescence and [HCN] simultaneously.

pH measurements

The pH was measured using a glass electrode (Hamilton Polilyte Lab 120, Reno, NV, USA) and a standard pH meter (Radiometer, Copenhagen, Denmark).


Both HCN and ACA have low vapor pressures and evaporate during experiments. The Henry constants KH,HCN = 7.5 m·atm−1 and KH,ACA = 14 m·atm−1 were obtained from Determining the evaporation of HCN empirically using the cyanide electrode allowed correction of the [HCN] measurements and circumvented the complexities associated with a theoretical estimate of the evaporation.

Electrode limitations and solutions

Although the results obtained using the cyanide electrode should be reproducible to ± 2% if calibrated every hour, we experienced accuracy problems of a few mV by repetition of measurements. This has a significant influence when measuring small changes in concentrated solutions of [HCN] due to the logarithmic electrode response. However, the electrode behaved well when it remained in the solution. In this case, slow changes of [HCN] (time scale of seconds) were reproduced well. To ameliorate the variation in absolute potential readings, we recommend calibrating the starting potential against a known [HCN], and scaling the readings afterwards (as opposed to preparing fresh calibration curves prior to each experiment). This adjustment has been performed in Fig. 3.


This work was supported by the European Union through the Network of Excellence “Bio-Sim” (contract number LSHB-CT-2004-005137), by the European Science Foundation FuncDyn Program, and by the Danish Agency for Science Technology and Innovation (grant number 272-07-0487). The authors thank Dorthe Boelskifte for technical assistance. B.O.H. was the recipient of a PhD fellowship from the Faculty of Health Sciences, University of Copenhagen, Denmark.