Abstract
Nutrient sensors allow cells to adapt their metabolisms to match nutrient availability by regulating metabolic pathway expression. Many such sensors are cytosolic receptors that measure intracellular nutrient concentrations. One might expect that inducing the metabolic pathway that degrades a nutrient would reduce intracellular nutrient levels, destabilizing induction. However, in the galactose-responsive (GAL) pathway of Saccharomyces cerevisiae, we find that induction is stabilized by flux sensing. Previously proposed mechanisms for flux sensing postulate the existence of metabolites whose concentrations correlate with flux. The GAL pathway flux sensor uses a different principle: the galactokinase Gal1p both performs the first step in GAL metabolism and reports on flux by signalling to the GAL repressor, Gal80p. Both Gal1p catalysis and Gal1p signalling depend on the concentration of the Gal1p–GAL complex and are therefore directly correlated. Given the simplicity of this mechanism, flux sensing is probably a general feature throughout metabolic regulation.
Similar content being viewed by others
Main
Nutrient sensing systems allow cells to adjust metabolism to their environment by sensing changes in extracellular conditions. There are two main known types of sensors: plasma membrane receptors1 and cytosolic receptors2,3. Plasma membrane receptors measure extracellular conditions directly by sensing extracellular nutrient concentrations. Cytosolic receptors measure intracellular nutrient concentrations, which are affected by both extracellular conditions (supply) and intracellular metabolism (demand). Thus, nutrient sensing systems that rely on the intracellular nutrient concentration cannot distinguish whether receptor sensing changes because of changes in supply or demand: the intracellular nutrient concentration could decrease because of a decrease in supply, for example, when a nutrient is no longer available in the environment, or because of an increase in demand, for example, when enzymes of a metabolic pathway are produced. This ambiguity could be problematic because changes in supply and changes in demand probably require different cellular responses. Nevertheless, many nutrient sensing systems rely on intracellular nutrient concentrations, raising the question of whether there are additional mechanisms that disambiguate these measurements.
Different regulatory mechanisms have been proposed that would solve this problem by unambiguously correlating intracellular nutrient concentrations with physiologically relevant quantities. One proposed mechanism results in the correlation of the intracellular nutrient concentration with extracellular nutrient concentrations4. This was achieved in mathematical models by matching any change in the expression of catabolic enzyme genes (demand) with corresponding changes in the expression of transport genes (supply). Coordinating the expression of these genes prevents changes in intracellular nutrient concentrations. Therefore, the intracellular nutrient concentration is only responsive to changes in supply, for example, in the extracellular nutrient concentration. However, the implementation of this mechanism in cells is complicated by the need to finely balance the production and activity of the proteins responsible for supply and demand across physiological and evolutionary timescales.
An alternative proposed mechanism results in the correlation of the intracellular nutrient concentration with demand, that is, metabolic flux. Flux sensing would be a useful biological property, and there has been some phenomenological evidence that flux sensing occurs in multiple metabolic pathways5,6 and transporters7,8,9. While no mechanism has been found by which flux can be directly measured, flux can be inferred by sensing the concentration of metabolites that are correlated with flux10,11,12. The initial model that achieved a correlation between nutrient concentration and flux required a complex interplay of allosteric and transcriptional regulation of the metabolic pathway12. A recent model proposed a simple mechanism for achieving the correlation but required thermodynamic constraints in the corresponding biochemical reaction13. Currently, few biological instantiations where such correlations are used to regulate nutrient sensing have been described.
The mechanisms that correlate intracellular nutrient concentrations with metabolic flux or extracellular conditions have mostly been described theoretically; the contribution of these mechanisms to nutrient sensing in cells is unclear. In this study, we combined theory and experiments to investigate the effect of changing supply and demand on the galactose-responsive (GAL) pathway of Saccharomyces cerevisiae. We find that the GAL pathway contains a flux sensor. This flux sensing is not achieved by correlating intracellular nutrient concentrations with metabolic flux, but by a new mechanism that measures metabolic flux at the level of a metabolic enzyme.
Results
Signalling in the GAL pathway increases with increasing flux
To illustrate how substrate metabolism affects sensing of intracellular substrate concentrations, we first constructed a toy model in which substrate molecules were transported into a cell through a transporter, and then either activated a concentration sensor or were metabolized by an enzyme (Fig. 1a, Supplementary Note 1 and Extended Data Fig. 1). As expected, in this simple model increasing metabolic enzyme levels led to a decrease in signalling (Fig. 1b).
a, Schematic of a canonical nutrient sensing pathway. Nutrient molecules diffuse into the cell through a transporter protein where they bind a nutrient sensor to activate signalling and are introduced into a metabolic pathway by a metabolic enzyme. b, Model prediction for the relationship between the level of a metabolic enzyme and nutrient sensing signalling (see Supplementary Note 1 for the model equations). c, Schematic of nutrient sensing in the GAL pathway. Galactose molecules diffuse into the cell through a transporter protein where they bind the nutrient sensor Gal3p to inhibit the GAL inhibitor Gal80p and are metabolized by the galactokinase enzyme Gal1p. d, Relationship between the level of Gal1p and GAL signalling. GAL1-mScarlet-I was expressed under the control of the TetO7 promoter in a gal1Δ background with a fluorescent reporter for GAL signalling (GAL1pr-mVenus). Fluorescent protein expression was quantified using flow cytometry. For each mScarlet-I bin, the mean and s.d. of the mVenus fluorescence measurements of at least three independent replicates were calculated and are represented by points and background shading, respectively. Cells were grown in the presence of 4 mM galactose and 4 mM glucose at low cell densities so that depletion did not substantially change the extracellular sugar concentrations during the experiment. e, Schematic of nutrient sensing in the GAL pathway with signalling by the galactokinase enzyme Gal1p. f, Model prediction for the relationship between the level of a metabolic enzyme and nutrient sensing signalling with signalling by the metabolic enzyme. g, Schematic of nutrient sensing in the GAL pathway when S. cerevisiae Gal1p is replaced with S. pombe Gal1p so that Gal1p signalling is removed while Gal1p catalysis is retained. h, Relationship between the level of S. pombe Gal1p and GAL signalling. The experiment was performed and analysed as described in d. a.u., arbitrary unit.
We compared this prediction to the behaviour of the GAL pathway of S. cerevisiae (Fig. 1c), a well-studied model system for nutrient sensing14,15. In this pathway, the concentration sensor Gal3p signals after binding to intracellular galactose, leading to the expression of enzymes involved in galactose metabolism16,17,18. One of these enzymes is Gal1p, a galactokinase that catalyses the first step of GAL metabolism. Throughout the article, we refer to this signalling process that results in GAL1 expression as ‘GAL signalling’.
To investigate how galactose metabolism affects GAL signalling, we varied the level of fluorescently tagged Gal1p by placing the expression of a GAL1-mScarlet-I fusion gene under the control of an inducible promoter, and measured GAL signalling in this strain by placing mVenus expression under the control of the GAL1 promoter. Surprisingly, increasing the Gal1p level increased GAL signalling (Fig. 1d, P < 0.0001).
We hypothesized that this unexpected result could be due to signalling by Gal1p itself. Like Gal3p, Gal1p can regulate GAL signalling: both the Gal3–galactose complex17,19,20,21 and the Gal1p–galactose complex19,22,23,24 bind the GAL repressor Gal80p, thus upregulating GAL gene expression. The role of Gal1p in GAL signalling has often been ignored as the basal expression of GAL1 is too low to activate GAL signalling on its own15,25,26 and most studies of the GAL pathway have focused on the activation of GAL signalling. However, once the GAL pathway is upregulated, the amount of Gal1p that is bound to Gal80p increases, while the amount of Gal3p that is bound to Gal80p decreases27, suggesting that Gal1p signalling could be important in this circumstance. Introducing this enzymatic signalling to the toy model (Fig. 1e), we found that increasing the enzyme level now increased signalling (Fig. 1f), which is consistent with the behaviour in the Gal1p titration experiment (Fig. 1d).
To determine the contribution of Gal1p signalling in the Gal1p titration experiment, we instead titrated the galactokinase gene from the evolutionarily distant yeast Saccharomyces pombe (SpGAL1). SpGal1p is a functional galactokinase in S. cerevisiae (Extended Data Fig. 2a), but given that the last common ancestor for these two yeasts lived more than 300 million years ago, we hypothesized that it does not retain Gal80p binding in S. cerevisiae and would therefore be unable to signal (Fig. 1g). Indeed, in a strain where GAL1 and GAL3 are deleted, we found that SpGal1p expression did not lead to GAL signalling (Extended Data Fig. 2b). To test whether SpGal1p catalysis affects Gal3p sensing, we titrated SpGAL1 in the presence of the wild-type (WT) GAL3 gene and found that increasing the SpGal1p level decreases GAL signalling (Fig. 1h, P < 0.0001), which is consistent with the initial model prediction (Fig. 1b). Therefore, the increase in GAL signalling in the Gal1p titration experiment is not achieved by sensing the concentrations of metabolites downstream of Gal1p: SpGal1p and endogenous Gal1p have the same catalytic activity, but increasing SpGal1p and Gal1p levels have opposite effects on GAL signalling (Fig. 1d,h). Therefore, our data indicate that native Gal1p signalling has an important role in preventing a decrease in GAL signalling that would otherwise result from increasing the expression of metabolic enzymes.
Gal1p is an enzymatic flux sensor
Increasing the Gal1p level increases GAL signalling, but as Gal1p catalyses the first step of galactose metabolism, this could also increase flux through Gal1p, raising the question of how changes in flux could be kept from interfering with signalling. We hypothesized that an intriguing solution could result from the mechanisms that underlie signalling and flux. Galactose binds to Gal1p to form the Gal1p–galactose complex; this complex is responsible for both the signalling and flux activities of Gal1p: (1) the Gal1p–galactose complex causes a conformational change in Gal1p that allows it to bind to and inhibit the GAL repressor protein Gal80p24, resulting in signalling; (2) the Gal1p–galactose complex is the molecular species that leads to the enzymatic reaction, resulting in flux. When the enzymatic reaction is largely irreversible, both signalling and flux through the enzyme are linearly dependent on this enzyme–substrate complex, effectively coupling the rates of the two activities (see Supplementary Note 2 for a discussion of reversible enzymes). For Gal1p, the enzymatic reaction is galactose phosphorylation, which consumes ATP and is close to irreversible; thus, this simple mechanism could effectively allow Gal1p to sense the flux through its reaction (Fig. 2a).
a, Schematic of a flux sensor. Signalling and flux depend on the same molecular species, the Gal1p–galactose complex, leading to a coupling of the two activities. We refer to this system as coupled because increasing the sensor level leads to a corresponding increase in signalling and flux. b, Schematic of a concentration sensor. Signalling and flux depend on different molecular species (the Gal3p–galactose complex and the SpGal1p–complex, respectively), leading to an uncoupling of the two activities. We refer to this system as uncoupled because increasing the sensor levels leads to an increase in signalling without a corresponding increase in flux (and vice versa). c,d, Prediction for the relationship between signalling and flux at different galactose and sensor levels for a flux (c) and concentration (d) sensor, respectively. e,f, Galactose and Gal1p level titration in GAL feedback-removed strains. e, We used a strain designed to report on Gal1p signalling only (gal1,7,10Δ::TDH3pr-GAL7-TDH3pr-GAL10 gal3Δ TetO7pr-GAL1 gal2Δ:TDH3pr-GAL2 gal80Δ::ACT1pr-GAL80 GAL1pr-mVenus). f, We used a strain designed to report on Gal3p sensing, with SpGAL1 overexpression to provide galactokinase activity (gal1,7,10Δ::TDH3pr-SpGAL1-TDH3pr-GAL7-TDH3pr-GAL10 gal3Δ TetO7pr-GAL3 gal2Δ::TDH3pr-GAL2 gal80Δ::ACT1pr-GAL80 GAL1pr-mVenus). In both cases, GAL signalling is defined as the mean mVenus fluorescence at a given combination of external galactose and doxycycline (DOX) inducer concentrations. The lines connect galactose titration series with the same DOX concentration (that is, similar sensor levels). The dashed red line indicates the mean GAL1pr-mVenus signal of a gal80Δ strain in saturating galactose (128 mM) and 4 mM glucose, an estimate for the upper limit of pathway signalling (see Extended Data Fig. 3 for replicate experiments).
To understand why sensing the flux through Gal1p could be useful, it is important to understand how this flux relates to the overall flux of galactose through metabolism. Because galactose metabolism is a single, unbranched pathway (Extended Data Fig. 1a), the steady-state flux through every reaction is equal. Therefore, sensing the flux through Gal1p is sufficient to measure the metabolic flux through the pathway. In addition, Gal1p flux is equal to the net transport (or uptake) flux of extracellular galactose. This is true because any galactose molecule that is imported and not lost by being exported will be metabolized by Gal1p. As all these fluxes are equivalent, we refer to them as ‘galactose flux’ throughout the article.
Based on the mechanism we propose, flux sensing should be a unique behaviour of Gal1p; the canonical galactose sensor Gal3p should not be a flux sensor. In contrast to Gal1p, the Gal3p–galactose complex has no catalytic activity19. Thus, there is no Gal3p flux that Gal3p signalling could be proportional to. Therefore, we propose that Gal3p acts as a classical galactose concentration sensor (Fig. 2b) and only Gal1p acts as a galactose flux sensor.
To identify potential experimental perturbations that clearly distinguish if Gal1p or Gal3p sense the galactose flux or the galactose concentration, we first analysed the two types of sensors in the toy model. A simple perturbation could be to vary the extracellular substrate concentration, and thus change flux. As flux and flux sensor signalling must be correlated, this perturbation should equally change flux and flux sensor signalling. However, varying the extracellular substrate concentration could also change the intracellular substrate concentration, and therefore concentration sensor signalling. Therefore, this perturbation cannot distinguish between flux and concentration sensors. Another simple perturbation would be to vary the sensor protein level. Varying the level of the sensor protein can affect both signalling and flux for a flux sensor (Fig. 2a), but only signalling for a concentration sensor (Fig. 2b). The experimental difference between the two behaviours can be magnified by simultaneously changing both the extracellular substrate concentration and the level of the sensor protein. For a flux sensor, varying the sensor level and the extracellular substrate concentration leads to all measurements of flux and signalling to lie on the same line (Fig. 2c). In contrast, for a concentration sensor, varying the sensor level and the extracellular substrate concentration leads to measurements of flux and signalling to lie on different lines for each sensor level (Fig. 2d). In summary, the relationship between flux and signalling does not depend on the sensor level for a flux sensor but depends on the sensor level for a concentration sensor.
To test this prediction experimentally, we constructed strains in which we could vary Gal1p or Gal3p levels while varying the extracellular galactose concentration. We deleted both endogenous galactose sensor genes (GAL1 and GAL3) and introduced either GAL1 or GAL3 under the control of an inducible promoter. As the inducible GAL3 strain would not be able to metabolize galactose because of the absence of GAL1, we constitutively expressed a signalling-deficient galactokinase gene (SpGAL1). To ensure that flux measurements would not be influenced by signalling feedback, we also expressed the genes for GAL80 (GAL regulation), GAL2 (galactose import), and GAL7 and GAL10 (galactose metabolism) from strong, constitutive promoters (Methods).
First, we measured the relationship between galactose flux and GAL signalling at different Gal1p levels. Consistent with the flux sensor prediction (Fig. 2c), varying the Gal1p level did not change the relationship between galactose flux and GAL signalling (Fig. 2e). Next, we measured the relationship between galactose flux and GAL signalling at different Gal3p levels. Unlike Gal1p, but consistent with the concentration sensor prediction (Fig. 2d), the relationship between galactose flux and GAL signalling depends on the Gal3p level (Fig. 2f).
It is worth noting that galactose flux and GAL signalling are only correlated until a plateau in signalling is reached for both Gal1p and Gal3p (Fig. 2e). The signalling level of this plateau is equivalent to maximal GAL signalling, that is, GAL signalling in a strain in which the GAL repressor gene GAL80 is deleted (Fig. 2e, dashed horizontal line). Therefore, we infer that this is a signalling limit where a signalling reaction downstream of Gal1p or Gal3p is saturated so that GAL signalling can no longer increase and refer to it as the ‘signalling limit’.
Gal1p flux sensing has a unique signalling regime
Next, we analysed the modelling data from Fig. 2 to determine how increasing sensor levels affects signalling instead of how it affects the relationship between flux and signalling. We found that flux and concentration sensors behaved identically at low sensor levels: increasing the sensor level increases signalling (Fig. 3a,b, regime I). Surprisingly, flux and concentration sensors behaved differently at high sensor levels (Fig. 3a,b). Concentration sensor signalling saturated at the signalling limit (Fig. 3b, regime III), while flux sensor signalling saturated below the signalling limit (Fig. 3a, regime II). In this regime, the flux sensor displayed a unique behaviour where signalling is independent of the sensor level, but still responsive to changes in the extracellular substrate concentration (Fig. 3a, regime II).
a,b, Top, model prediction of the effect of a flux sensor (a) and concentration sensor (b) titration on signalling. Bottom, schematic overview of the responses to sensor level increases. c,d, Relationship between Gal1p (c) and Gal3p (d) levels and GAL signalling at different external galactose concentrations in the gal1Δ gal3Δ TetO7pr-GAL1-mScarlet-I (c) or gal1Δ gal3Δ TetO7pr-GAL3-mScarlet-I (d) strains. GAL signalling and sensor levels were determined and analysed as described in Fig. 1d. See Supplementary Note 3 for the quantitative definition of the regime-separating lines in a–d. See Extended Data Fig. 7 a comparison of replicate experiments.
The independence of signalling and the sensor level is caused by the catalytic activity of the flux sensor. For a concentration sensor, that is, a sensor protein with no catalytic activity, increasing the sensor level increases the concentration of sensor proteins that can bind the substrate and thereby the concentration of the sensor–substrate complex. Therefore, increasing the sensor level increases signalling (Fig. 3b, regime I). For a flux sensor, increasing the sensor levels has a second, opposing effect. Because of the catalytic activity of the flux sensor, increasing the flux sensor level also decreases the intracellular concentration of substrate molecules that can bind flux sensor proteins. When the two opposing effects of increasing the flux sensor level are perfectly balanced, the concentration of the flux sensor–substrate complex remains constant. In this situation, increasing the flux sensor level would not increase signalling.
To understand when increases in the flux sensor level will be perfectly balanced by decreases in the intracellular substrate concentration, we need to consider the flux sensor reaction as a part of a metabolic pathway. This is necessary because the intracellular substrate concentration depends on all the reactions in its metabolic pathway. In our mathematical model, we simplified the pathway to two core reactions: substrate transport through a plasma membrane transporter and substrate catalysis by the flux sensor. The extent to which these two reactions control the intracellular substrate concentration and pathway flux is described using metabolic control analysis28,29 and depends in part on the levels of the proteins that carry out the reactions. When the flux sensor level is low, the pathway flux is controlled by the flux sensor (Extended Data Fig. 4a). In this regime, reverse transport through the plasma membrane is much greater than the rate of metabolism by the flux sensor, so that the intracellular substrate concentration is nearly equal to the extracellular substrate concentration. Therefore, increasing the flux sensor level only negligibly decreases the intracellular substrate concentration (Extended Data Fig. 4b), and thus the concentration of the flux sensor–substrate complex, signalling and flux all increase (Fig. 3a, regime I, and Extended Data Fig. 4c). When the flux sensor level is high, control is reversed: pathway flux is controlled by the transporter (Extended Data Fig. 4a). In this regime, doubling the flux sensor level halves the intracellular substrate concentration (Extended Data Fig. 4b), that is, the system is perfectly balanced and the concentration of the flux sensor–substrate complex, signalling and flux all remain constant (Fig. 3a, regime II, and Extended Data Fig. 4c).
Regime II, where signalling is independent of the flux sensor, corresponds to a limit in the transport flux. We refer to this limit to as the ‘flux limit’. As described above, regime II occurs when the pathway flux is controlled by substrate transport (Extended Data Fig. 4a). This also means that the substrate transport reaction is maximal for the given transporter level and extracellular substrate concentration (Extended Data Fig. 4c). Because this maximal transport flux determines pathway flux in the flux limit, it also determines flux sensor signalling (Extended Data Fig. 4d and Supplementary Note 4). Notably, this design allows flux sensor signalling to be independent of the sensor level but responsive to changes in the extracellular galactose concentration (Fig. 3a, regime II). Thus, flux sensor signalling can measure the extracellular environment in a way that is robust to any changes in flux sensor level arising from gene regulation or gene expression noise.
The model creates experimentally testable predictions for the galactose sensors Gal1p and Gal3p. Specifically, the model predicts that increasing the level of the flux sensor Gal1p or the concentration sensor Gal3p should increase GAL signalling until reaching different limits, the flux limit or the signalling limit. These limits can be distinguished because only the flux limit depends on the maximal transport flux and therefore on the extracellular substrate concentration (Fig. 3a,b). In addition, as described in Fig. 2, the signalling limit can be determined genetically by deleting the GAL80 gene (Fig. 3c,d, regime III); this orthogonal measurement can be compared to the limits that are observed when increasing sensor levels.
To test which limits are reached by Gal1p and Gal3p, we experimentally varied their levels by increasing their expression from an inducible promoter. We found excellent agreement between the flux sensor predictions and the Gal1p data (Fig. 3a,c), and the concentration sensor predictions and the Gal3p data (Fig. 3b,d). Increasing the Gal1p level increases GAL signalling when the Gal1p level is low (Fig. 3c, regime I), but reaches the flux limit when the Gal1p level is high (Fig. 3c, regime II). As predicted, signalling in the flux limit is responsive to changes in the maximal transport flux: changing the extracellular galactose concentration (Fig. 3c, regime II) or the level of the transporter Gal2p (Extended Data Fig. 5) changes GAL signalling in the flux limit. As signalling already saturates in the flux limit, Gal1p titration does not reach the signalling limit (Fig. 3c). In contrast to Gal1p, increasing the Gal3p level continuously increases GAL signalling (Fig. 3d, regime I) until reaching the signalling limit where signalling is no longer responsive to the extracellular galactose concentration (Fig. 3d, regime III). Together, these experiments confirm that flux sensing by Gal1p enables a unique regime, the flux limit, where GAL signalling is independent of Gal1p level but responsive to the extracellular galactose concentration.
The flux limit is reached by endogenous GAL regulation. In our experiments, we controlled the Gal1p level by expressing GAL1 from an inducible promoter; this resulted in Gal1p levels that were high enough to reach the flux limit (Fig. 3c, regime II). In WT cells, GAL signalling controls the Gal1p level, raising the question whether this endogenous regulation allows Gal1p levels that reach the flux limit. Thus, we measured the Gal1p level when expressed from the WT GAL1 promoter. We found that in the absence of galactose, the Gal1p level is too low to reach the flux regime (Extended Data Fig. 6). On the other hand, in the presence of galactose, the Gal1p level is high enough to reach the flux limit (Extended Data Fig. 6). Thus, both the dependence of GAL signalling on Gal1p levels in regime I and the independence of GAL signalling on Gal1p levels in regime II (the flux limit) are encountered in physiological settings.
Flux sensing stabilizes GAL signalling
While most studies have focused on the initial activation of GAL signalling, continued growth on galactose also requires maintenance of GAL signalling. The initial activation is achieved by Gal3p concentration sensing30,31 and leads to the expression of metabolic GAL genes such as GAL1. As we describe above, increasing the level of a metabolic enzyme like Gal1p can decrease the intracellular concentration of its substrate. Initially, this might not be a problem because the enzyme level is low and only leads to a negligible decrease in the intracellular substrate concentration (Extended Data Fig. 4b, regime I) as the enzyme has a high flux control coefficient (Extended Data Fig. 4a). However, eventually increasing the enzyme level will transfer flux control to a different reaction (Extended Data Fig. 4a); then, any increase in the enzyme level will lead to a corresponding decrease in the intracellular substrate concentration (Extended Data Fig. 4b, regime II). Consistent with this, when GAL signalling is only regulated by Gal3p concentration sensing, we observed that increasing the SpGal1p level only had small effects on GAL signalling when the SpGal1p level was low, but strongly reduced GAL signalling once the SpGal1p was high (Fig. 1h). In contrast, when GAL signalling is regulated by Gal1p flux sensing, increasing the Gal1p level increases GAL signalling at low Gal1p levels (Figs. 1d and 3a, regime I). Moreover, at high Gal1p levels, Gal1p flux sensing has the useful property that GAL signalling is constant regardless of fluctuations in Gal1p level, but responsive to changes in the extracellular galactose concentration (Fig. 3a, regime II). Therefore, we hypothesized that the physiological role of Gal1p flux sensing is to compensate for the decrease in Gal3p concentration sensing that occurs when galactose flux increases, thereby ensuring the stability of GAL signalling, that is, the ability to maintain GAL signalling at a given signalling level as long as the extracellular galactose concentration is constant.
If the physiological role of Gal1p flux sensing is to compensate for decreasing Gal3p concentration sensing, Gal1p flux sensing should not be needed when there is no galactose flux. We removed galactose flux and Gal1p flux sensing by deleting the GAL1 gene. To determine whether GAL signalling is stable after deleting GAL1, we used time-lapse microscopy to identify cells with high levels of GAL signalling after 10 h of growth in a medium with galactose (see Extended Data Fig. 8 for the population distributions of GAL signalling). We then monitored GAL signalling in these cells for the following 4 h. The reporter for GAL signalling we used in previous experiments, the expression of mVenus from the GAL1 promoter, is not suitable for these experiments because the fluorescent protein is stable over this time span. Therefore, we replaced mVenus with an actively degraded fluorescent protein fusion, that is, mNeonGreen-Ubi-Y. Surprisingly, we found that while GAL signalling is stable in the WT strain, it is unstable when GAL1 is deleted (Extended Data Figs. 8a and 9a). This indicates that Gal1p flux sensing is required for stable GAL signalling even in the absence of galactose flux.
We reasoned that GAL signalling is unstable when GAL1 is deleted because of transcriptional feedback by the GAL repressor GAL80. Gal3p signalling leads to the expression of GAL80. Gal80p in turn inhibits the GAL transcriptional activator Gal4p, constituting a negative feedback loop that has the potential to destabilize GAL signalling. When GAL1 is present, this negative feedback could be compensated by positive feedback through increased GAL1 expression. When GAL1 is deleted, this compensation would not occur, resulting in unstable GAL signalling. To test this hypothesis, we removed this negative feedback by replacing the GAL80 promoter with a constitutive promoter. Indeed, in this constitutive GAL80 expression background, GAL signalling is stable even when GAL1 is deleted (Extended Data Figs. 8b and 9b), that is, in the absence of galactose flux and Gal1p flux sensing.
Using strains with constitutive GAL80 expression removes the confounding effect of negative GAL80 feedback and thus allowed us to test whether Gal1p flux sensing is required to stabilize GAL signalling as galactose flux increases. We first tested the stability of GAL signalling in a strain with otherwise WT regulation. We found that GAL signalling in this strain is stable (Fig. 4a). To test whether this stability requires Gal1p flux sensing, we removed Gal1p flux sensing, but not galactose flux, by replacing the GAL1 coding sequence (CDS) with the SpGAL1 CDS, which has no signalling activity in S. cerevisiae (Extended Data Fig. 2b). Replacing only the CDS maintains the ability of GAL signalling to increase galactose flux by increasing the galactokinase level. In this strain, GAL signalling is unstable (Fig. 4b, P < 0.0001), indicating that Gal1p flux sensing is necessary for stabilizing GAL signalling. We note that this instability occurs in the presence of positive feedback through increased GAL2 expression and that removing this positive feedback in a strain with WT GAL1 and GAL80 expression does not destabilize signalling (Extended Data Figs. 8c and 9c). These results indicate that positive feedback through GAL2 is neither necessary nor sufficient for stable GAL signalling. Similarly, we found that abolishing Gal3p concentration sensing by deleting GAL3 does not destabilize GAL signalling (Fig. 4c, P = 0.28). These results indicate that the stabilization of GAL signalling is a physiological role that is unique to Gal1p flux sensing.
a–c, Single-cell time courses of GAL1pr-Ubi-Y-mNeonGreen strains with WT regulation (a), gal1Δ::GAL1pr-SpGAL1 (b) or gal3Δ (c) in a gal80Δ::CYC1pr-GAL80 background. For each genotype, 50 single-cell time courses were randomly selected from the population of cells whose GAL signalling was in the top 20% after 10 h in a medium with 4 mM glucose and 1 mM galactose (a and b) or 4 mM galactose (c). See Extended Data Fig. 9d for the replicate experiments.
Discussion
In this study, we showed that Gal1p is a flux sensor. We argue that this flux sensing arises from the bifunctional nature of Gal1p. Since both Gal1p activities, catalysis and signalling, depend on the formation of the Gal1p–galactose complex, they are inherently coupled: the signalling activity of Gal1p is always proportional to the catalytic activity of Gal1p. Because of this coupling, Gal1p effectively senses galactose flux.
Gal1p flux sensing acts in parallel with the canonical galactose sensor Gal3p, raising the question of why two different galactose sensors regulate GAL signalling. A possible explanation is that having two sensor genes allows for two distinct transcriptional profiles15,25,26,32: in the absence of galactose, GAL3 is expressed while GAL1 is not. In turn, in the presence of galactose, GAL1 is expressed at substantially higher levels than GAL3. These distinct transcriptional profiles define two distinct physiological tasks for the two sensors. As described before, basal GAL3 expression enables Gal3p concentration sensing to activate GAL signalling when galactose is first encountered30,31. As we described in this study, the increase in GAL1 expression in the presence of galactose enables Gal1p flux sensing to stabilize GAL signalling when galactose is metabolized (Figs. 1 and 4). The transcriptional profiles thus divide the general task of regulating GAL signalling into the two specialized tasks of activating and stabilizing GAL signalling.
In addition to having distinct tasks, Gal1p and Gal3p also sense distinct properties; specifically, Gal3p activates GAL signalling based on the intracellular galactose concentration and Gal1p stabilizes GAL signalling based on the galactose flux (Figs. 2 and 3). We suspect that these properties help Gal1p and Gal3p accomplish their respective specialized tasks. Consistent with this idea, the activation task of Gal3p concentration sensing can only partially be accomplished by Gal1p flux sensing: when the GAL3 CDS is replaced by the GAL1 CDS, growth in galactose is delayed26,33. We do not have direct data that Gal3p concentration sensing cannot replace Gal1p flux sensing solely for the sensing function as we never created a strain where both GAL3 and a signalling-deficient galactokinase gene have the transcriptional profile of GAL1. However, our data provide a hint as to why galactose sensing in this strain could be problematic. Expression of GAL3 from the more inducible GAL1 promoter would lead to high Gal3p levels in the presence of extracellular galactose. This could mean that Gal3p signalling reaches the signalling limit where Gal3p signalling is no longer responsive to the extracellular galactose concentration (Fig. 3d, regime III). This constitutive signalling would be a problem when extracellular galactose decreases because it would lead to the superfluous expression of GAL genes. Elucidating this interplay between the physiological roles of the two sensors and the properties they sense will be the subject of future work.
Even though the mechanism of flux sensing we report in this study has not, to our knowledge, been described before, we believe that it is ubiquitous. As we outline below, we reason that this is true because the mechanism is simple to evolve, leveraging features that are already common for metabolic enzymes but difficult to detect with traditional experimental approaches.
The biophysical feature that underlies Gal1p flux sensing can be described as ‘reverse allostery’. When galactose binds to the active site of Gal1p, the surface conformation of Gal1p changes18,24, allowing binding to Gal80p. This constitutes a flow of information from the active site to distal surface sites. Allostery, a common and widely appreciated form of metabolic regulation, is the inverse of this process: the flow of information from distal surface sites to the active site. As energy is not used in this process, from a thermodynamic perspective, information must simultaneously flow in both directions. Indeed, recent work showed that sites of allosteric regulation can be predicted by measuring reverse allostery, that is, identifying how substrate binding at the active site leads to conformational changes at distal surface sites34,35,36. If allostery is a common feature of metabolic enzymes and the flow of information is bidirectional, reverse allostery must be equally common. This indicates that many metabolic enzymes meet the biophysical requirements for flux sensing.
To achieve flux sensing, reverse allostery must regulate a secondary activity of the metabolic enzyme. Minimally, this requires that the metabolic enzyme carries out a secondary activity in addition to its canonical catalytic activity. Such secondary activities, often described as ‘moonlighting’ activities, are now being reported for many metabolic enzymes37,38,39,40. Thus, metabolic enzymes are increasingly appreciated as multifunctional proteins, suggesting that the presence of a secondary activity—a prerequisite for flux sensing through reverse allostery—is also a common feature of metabolic enzymes.
If flux sensing through reverse allostery is a common mechanism, why has it not been found previously? In the context of metabolism, such mechanisms are easy to miss. Typically, metabolic regulation has been studied by deleting individual genes. This approach is often not possible for metabolic enzymes because of essentiality: ~75% of the essential genes of S. cerevisiae are involved in metabolic processes41. Even for metabolic genes where deletions are viable, the effects of the deletion are difficult to disentangle. Gene deletions perturb both metabolic flux through the enzyme and any potential signalling activity of the enzyme. This makes it difficult to distinguish whether the deletion changes metabolic regulation because of indirect effects (caused by the perturbed metabolic flux) or direct effects (caused by the perturbed signalling activity).
If regulation by reverse allostery is indeed common, this mechanism could couple metabolic flux to a wide range of cellular processes. Reverse allostery is flexible and not limited to regulating binding to other proteins. Existing examples of moonlighting activities include regulation at other control points by binding to DNA and RNA42,43. In addition, reverse allostery is not limited to regulating binding. For proteins with multiple catalytic activities, reverse allostery could lead to regulation of one catalytic activity based on another catalytic activity. Such a design would allow one metabolic flux to directly regulate another metabolic flux. More generally, reverse allostery or ‘catalysis-induced enzyme movement’ has been proposed to perform work in cells, that is, ‘stirring’ the cytosol44,45. Beyond the naturally occurring regulation through reverse allostery, we envision that advances in de novo protein design will allow synthetic biologists to design proteins that specifically recognize the substrate-bound state of the enzyme and thereby couple metabolic flux to any arbitrary cellular process.
In conclusion, this study provides a framework for the identification, characterization and design of flux-sensing systems in metabolic pathways. We anticipate that applying this framework more broadly will establish flux sensing as a general theme of metabolic regulation. The mechanism of regulation by reverse allostery has implications beyond flux sensing: in principle, this mechanism should be able to report the activity of any arbitrary enzyme that displays conformational changes during its catalytic cycle.
Methods
No statistical methods were used to predetermine sample sizes but our sample sizes are similar to those reported in previous publications46,47. Data collection and analysis were performed without randomization and not blind to the conditions of the experiments
Strain construction
All strains were derived from FY4. Yeast transformations were performed using a standard lithium acetate protocol48. For chromosome integration, a selectable Cas9 expression plasmid was cotransformed with a single-guide RNA expression cassette targeting the chromosomal site, and homologous donor DNA to repair the double-strand break49. Plasmids for yeast transformations were cloned using Gibson assembly and the yeast MoClo toolkit50 (see Supplementary Table 1 for a list of strains, Supplementary Table 2 for a list of plasmids and Supplementary Table 3 for a list of oligonucleotides).
Fluorescent reporters for GAL1 expression
The GAL1 promoter sequence was amplified from genomic DNA using PCR and cloned in front of a codon-optimized mVenus CDS followed by the ADH1 terminator. For a reporter with decreased half-life, the GAL1 promoter was cloned in front of the mNeonGreen CDS fused to ubiquitin-N-degron51 with a glycine-serine linker followed by the GAL1 terminator.
Gene deletion
For gene deletion, the region from the start codon up to the stop codon was replaced with unique 20-bp sequences. Annealed oligonucleotides with 40–60 bp homology sequences to both the gene promoter and the stop codon-terminator were cotransformed for homologous recombination.
Gene titration
PCR-amplified CDS sequences were cloned as fusion proteins with a C-terminal yeast codon-optimized mScarlet-I tag, linked by a glycine-aspartate-glycine-alanine-glycine-serine and followed by the TDH1 terminator. Expression of the cloned CDS was controlled by a synthetic promoter consisting of seven tet operator sites followed by a CYC1 minimal promoter. The plasmid also contained the transcription factor rtTA2s-M2 CDS flanked by the MYO2 promoter and the ADH1 terminator.
Culture conditions
Strains were grown in synthetic minimal medium consisting of 1.7 g l−1 yeast nitrogen base, 5 g l−1 ammonium sulphate and variable concentrations of galactose, glucose and raffinose. For the titration experiments, the medium was supplemented with seven DOX concentrations, ranging from 2 μg ml−1 to 0.01 μg ml−1 in 2.44× dilution steps or no DOX. For the microscopy experiments, a low-fluorescence yeast nitrogen base lacking folic acid and riboflavin was used. Cultures were grown in 1.3-ml 96-well plates in a humified incubator at 30 °C with orbital shaking at 999 rpm.
Steady-state GAL signalling measurements
Strains were streaked out from −80 °C glycerol stocks on yeast extract peptone dextrose plates. Colonies were picked into 1 ml yeast extract peptone dextrose and incubated for 24 h. Cultures were then used to inoculate a cell dilution series in 500 μl 2% raffinose medium (supplemented with DOX when appropriate). After 16 h, the optical density at 600 nm (OD600) of cultures was measured using a Synergy H1 plate reader (BioTek). Cultures with OD600 values between 0.025 and 0.075 were washed twice using centrifugation (3,000g, 3′) and resuspension in synthetic minimal medium without any carbon sources. A 5-μl washed culture was used to inoculate 500 μl of medium containing galactose, glucose and, when appropriate, DOX to an inoculation OD600 of 0.0005. After 8 h, cultures were washed three times using centrifugation (3,000g, 3′) and resuspension in Tris-EDTA buffer (10 mM, 1 mM EDTA). Washed cells were analysed using an S1000EX flow cytometer with an A700 automated plate handling system (Stratedigm). For typical flow cytometry experiments, at least 3,000 cells were analysed in each condition.
Flow cytometry data were analysed using custom R scripts. Fluorescence channel values were normalized to cell size by dividing by the cell side scatter. To analyse the protein titration experiments, mScarlet-I fluorescence was used as a proxy for the level of the mScarlet-I fusion protein. Cells grown in different DOX concentrations were pooled computationally and binned based on their mScarlet-I fluorescence. For every combination of genotype and galactose concentration, 11 breaks between the first and 99th percentile mCherry levels were defined in logarithmic space, resulting in ten equally spaced bins. GAL signalling in each bin was summarized by averaging the fluorescein isothiocyanate (FITC) channel fluorescence of all cells of a given replicate in logarithmic space. To account for bleed-through of the mScarlet-I fluorescence into the FITC channel, the mean FITC fluorescence in 0 mM GAL was subtracted in linear space. Replicate measurements were taken from separately grown cultures. For plots with multiple summarized replicates, the mean and s.d. of background-subtracted bin means were calculated in logarithmic space.
Galactose uptake rate measurement
Strains were pre-grown as described above, but the outgrowth culture was performed in a tube shaker (230 rpm, 30 °C) with increased volume (5 ml). After three washes in synthetic minimal medium without any carbon sources (see above), 1 ml of medium was inoculated to an inoculation OD600 of 0.02. After 0.5, 4 and 8 h, samples were taken by spinning down 150 μl of culture medium (3,000g, 3′) and freezing the supernatant. Galactose concentrations in the culture supernatant were determined in duplicate using an enzymatic kit (l-Arabinose/d-Galactose Assay Kit, Megazyme). Conditions where one of the time points had an s.d. above 0.5 or where less than 95% of the initial galactose was taken up were excluded from further analysis. At every time point, 150 μl of culture was mixed with 50 μl of a 100-counting bead μl−1 suspension (CountBright, Thermo Fisher Scientific) and analysed using flow cytometry to obtain the cell concentrations and doubling time. To calculate the galactose depletion rate (r), data were fitted to the following formula:
where the external galactose concentrations (galext) at the given time points (t), the initial cell count (N0) and the doubling time (td) were the fixed parameters, while the initial galactose concentration (galinit) and the galactose depletion rate (r) were the free parameters. Mean fluorescence values in the flow cytometry measurement at the final time point were used to quantify protein levels and GAL signalling in given DOX concentrations as described above.
Time-lapse microscopy
Strains were pre-grown and washed as described above. Cells were resuspended to an OD600 = 0.1 and 7.5 μl were used to inoculate a 1.5-ml culture of synthetic minimal medium with the appropriate carbon concentrations. After shaking for 7 h in a tube shaker (230 rpm, 30 °C), cells were pelleted (3,000g, 3′) and resuspended in 20 μl medium. Concentrated cultures were loaded into concanavalin A-coated channels in a flow cell (µ-Slide VI 0.1, Ibidi) and attached using centrifugation (100g, 3′). To ensure constant carbon concentrations and to remove daughter cells, the medium was perfused through the channels at a flow rate of 10 μl min−1. Cells were imaged every 15 min using fluorescence microscopy with a charge-coupled device camera (Orca-R2, Hamamutsu), a ×60/1.40 objective (Plan Apo VC ×60/1.40 Oil, Nikon), a motorized XY-stage (ProScan II, Prior), a metal-halide lamp system (Lumen 200 Pro, Prior) and an emission filter wheel (Prior) on an Eclipse Ti inverted microscope (Nikon). To quantify Ubi-Y-mNeonGreen fluorescence, cells were excited using a 500/24-nm filter; fluorescence was collected with a 542/27-nm filter. To quantify mTagBFP2 fluorescence, cells were excited using a 390/40-nm filter; fluorescence was collected with a 452/45-nm filter. Image acquisition was controlled using custom Python scripts. At least 250 cells were analysed in each condition.
Images were segmented based on mTagBFP2 fluorescence using software available at https://github.com/alexxijielu/yeast_segmentation. Custom Python scripts were used to track individual cells. Briefly, segmented cells were assigned to a cell trace when they were the closest cell within 30 pixels (estimated by the distance between the centroids of the segmented objects), which was between 0.75× and 1.5× of the traced cell size (estimated by the area of the segmented objects) in the previous image. For each cell and image, the average pixel intensity of the Ubi-Y-mNeonGreen fluorescence measurement was subtracted with a constant value (1,675) to correct for background fluorescence.
Statistical analysis
All statistical tests were two-sided t-tests (assuming normal distribution of the underlying data). To test whether galactokinase levels changed signalling (Fig. 1), the mVenus distributions of cells of all replicates in the lowest and highest mScarlet-I bin were compared. For Fig. 1d, the test statistic was 325, the 95% confidence intervals (CIs) were 100.80 and 100.81 with 67,690 d.f. For Fig. 1h, the mVenus distributions of cells of all replicates in the lowest and highest mScarlet-I bin were compared. The test statistic was −169; the 95% CIs were 10−0.72 and 10−0.70 with 25,342 d.f. These calculations were made without subtracting the mean FITC fluorescence in 0 mM galactose. To test whether the flux and concentration sensing strains had unstable GAL signalling (Fig. 4), the relative GAL signalling values at the final time point were calculated by dividing the final time point by the 10-h time point. The resulting distributions were compared with the null hypothesis that the mean of the distribution was 0 in logarithmic space (that is the GAL signalling values at the two time points were identical and their ratio was 1 in linear space). For Fig. 4b, the test statistic was −19, the 95% CIs were 2−2.05 and 2−1.66 with 96 d.f. For Fig. 4c, the test statistic was 1, the 95% CIs were 2−0.043 and 2−0.15 with 227 d.f.
Mathematical modelling
The mathematical model is described in Supplementary Note 1. Model parameters are shown in Supplementary Table 4. The robustness of the modelling results is described in Extended Data Fig. 10.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
All raw data have been deposited in the Dryad repository (https://doi.org/10.5061/dryad.t4b8gtj7b) (ref. 52).
Code availability
All code used in the analysis has been deposited in a Dryad repository (https://doi.org/10.5061/dryad.t4b8gtj7b) (ref. 52).
References
Holsbeeks, I., Lagatie, O., Van Nuland, A., Van de Velde, S. & Thevelein, J. M. The eukaryotic plasma membrane as a nutrient-sensing device. Trends Biochem. Sci. 29, 556–564 (2004).
Conrad, M. et al. Nutrient sensing and signaling in the yeast Saccharomyces cerevisiae. FEMS Microbiol. Rev. 38, 254–299 (2014).
Zaman, S., Lippman, S. I., Zhao, X. & Broach, J. R. How Saccharomyces responds to nutrients. Annu. Rev. Genet. 42, 27–81 (2008).
Berkhout, J., Teusink, B. & Bruggeman, F. J. Gene network requirements for regulation of metabolic gene expression to a desired state. Sci. Rep. 3, 1417 (2013).
You, C. et al. Coordination of bacterial proteome with metabolism by cyclic AMP signalling. Nature 500, 301–306 (2013).
Okano, H., Hermsen, R., Kochanowski, K. & Hwa, T. Regulation underlying hierarchical and simultaneous utilization of carbon substrates by flux sensors in Escherichia coli. Nat. Microbiol. 5, 206–215 (2020).
Fritz, G. et al. A new way of sensing: need-based activation of antibiotic resistance by a flux-sensing mechanism. mBio 6, e00975 (2015).
Piepenbreier, H., Fritz, G. & Gebhard, S. Transporters as information processors in bacterial signalling pathways. Mol. Microbiol. 104, 1–15 (2017).
Somavanshi, R., Ghosh, B. & Sourjik, V. Sugar influx sensing by the phosphotransferase system of Escherichia coli. PLoS Biol. 14, e2000074 (2016).
Litsios, A., Ortega, Á. D., Wit, E. C. & Heinemann, M. Metabolic-flux dependent regulation of microbial physiology. Curr. Opin. Microbiol. 42, 71–78 (2018).
Kotte, O., Zaugg, J. B. & Heinemann, M. Bacterial adaptation through distributed sensing of metabolic fluxes. Mol. Syst. Biol. 6, 355 (2010).
Kochanowski, K. et al. Functioning of a metabolic flux sensor in Escherichia coli. Proc. Natl Acad. Sci. USA 110, 1130–1135 (2013).
Euler, C. & Mahadevan, R. On the design principles of metabolic flux sensing. Biophys. J. 121, 237–247 (2022).
Bhat, P. J. Galactose Regulon of Yeast: From Genetics to Systems Biology (Springer, 2008).
Johnston, M. & Davis, R. W. Sequences that regulate the divergent GAL1-GAL10 promoter in Saccharomyces cerevisiae. Mol. Cell. Biol. 4, 1440–1448 (1984).
Yano, K. & Fukasawa, T. Galactose-dependent reversible interaction of Gal3p with Gal80p in the induction pathway of Gal4p-activated genes of Saccharomyces cerevisiae. Proc. Natl Acad. Sci. USA 94, 1721–1726 (1997).
Platt, A. & Reece, R. J. The yeast galactose genetic switch is mediated by the formation of a Gal4p–Gal80p–Gal3p complex. EMBO J. 17, 4086–4091 (1998).
Lavy, T., Kumar, P. R., He, H. & Joshua-Tor, L. The Gal3p transducer of the GAL regulon interacts with the Gal80p repressor in its ligand-induced closed conformation. Genes Dev. 26, 294–303 (2012).
Bhat, P. J., Oh, D. & Hopper, J. E. Analysis of the GAL3 signal transduction pathway activating GAL4 protein-dependent transcription in Saccharomyces cerevisiae. Genetics 125, 281–291 (1990).
Peng, G. & Hopper, J. E. Gene activation by interaction of an inhibitor with a cytoplasmic signaling protein. Proc. Natl Acad. Sci. USA 99, 8548–8553 (2002).
Suzuki-Fujimoto, T. et al. Analysis of the galactose signal transduction pathway in Saccharomyces cerevisiae: interaction between Gal3p and Gal80p. Mol. Cell. Biol. 16, 2504–2508 (1996).
Meyer, J., Walker-Jonah, A. & Hollenberg, C. P. Galactokinase encoded by GAL1 is a bifunctional protein required for induction of the GAL genes in Kluyveromyces lactis and is able to suppress the gal3 phenotype in Saccharomyces cerevisiae. Mol. Cell. Biol. 11, 5454–5461 (1991).
Zenke, F. T. et al. Activation of Gal4p by galactose-dependent interaction of galactokinase and Gal80p. Science 272, 1662–1665 (1996).
Thoden, J. B., Sellick, C. A., Timson, D. J., Reece, R. J. & Holden, H. M. Molecular structure of Saccharomyces cerevisiae Gal1p, a bifunctional galactokinase and transcriptional inducer. J. Biol. Chem. 280, 36905–36911 (2005).
West, R. W. Jr, Yocum, R. R. & Ptashne, M. Saccharomyces cerevisiae GAL1-GAL10 divergent promoter region: location and function of the upstream activating sequence UASG. Mol. Cell. Biol. 4, 2467–2478 (1984).
Hittinger, C. T. & Carroll, S. B. Gene duplication and the adaptive evolution of a classic genetic switch. Nature 449, 677–681 (2007).
Abramczyk, D., Holden, S., Page, C. J. & Reece, R. J. Interplay of a ligand sensor and an enzyme in controlling expression of the Saccharomyces cerevisiae GAL genes. Eukaryot. Cell 11, 334–342 (2012).
Kacser, H. & Burns, J. A. The control of flux. Symp. Soc. Exp. Biol. 27, 65–104 (1973).
Heinrich, R. & Rapoport, T. A. A linear steady-state treatment of enzymatic chains. General properties, control and effector strength. Eur. J. Biochem. 42, 89–95 (1974).
Spiegelman, S., Sussman, R. R. & Pinska, E. On the cytoplasmic nature of ‘long-term adaptation’ in yeast. Proc. Natl Acad. Sci. USA 36, 591–606 (1950).
Broach, J. R. Galactose regulation in Saccharomyces cerevisiae. The enzymes encoded by the GAL7, 10, 1 cluster are co-ordinately controlled and separately translated. J. Mol. Biol. 131, 41–53 (1979).
Bajwa, W., Torchia, T. E. & Hopper, J. E. Yeast regulatory gene GAL3: carbon regulation; UASGal elements in common with GAL1, GAL2, GAL7, GAL10, GAL80, and MEL1; encoded protein strikingly similar to yeast and Escherichia coli galactokinases. Mol. Cell. Biol. 8, 3439–3447 (1988).
Lavy, T., Yanagida, H. & Tawfik, D. S. Gal3 binds Gal80 tighter than Gal1 indicating adaptive protein changes following duplication. Mol. Biol. Evol. 33, 472–477 (2016).
Ni, D. et al. Discovery of cryptic allosteric sites using reversed allosteric communication by a combined computational and experimental strategy. Chem. Sci. 12, 464–476 (2020).
Fan, J. et al. Harnessing reversed allosteric communication: a novel strategy for allosteric drug discovery. J. Med. Chem. 64, 17728–17743 (2021).
Zhang, Q. et al. Targeting a cryptic allosteric site of SIRT6 with small-molecule inhibitors that inhibit the migration of pancreatic cancer cells. Acta Pharm. Sin. B 12, 876–889 (2022).
Gupta, M. N., Kapoor, M., Majumder, A. B. & Singh, V. Isozymes, moonlighting proteins and promiscous enzymes. Curr. Sci. 100, 1152–1162 (2011).
Jeffery, C. J. Enzymes, pseudoenzymes, and moonlighting proteins: diversity of function in protein superfamilies. FEBS J. 287, 4141–4149 (2020).
Pan, C., Li, B. & Simon, M. C. Moonlighting functions of metabolic enzymes and metabolites in cancer. Mol. Cell 81, 3760–3774 (2021).
Gupta, M. N. & Uversky, V. N. Moonlighting enzymes: when cellular context defines specificity. Cell. Mol. Life Sci. 80, 130 (2023).
Zhang, Z. & Ren, Q. Why are essential genes essential?—the essentiality of Saccharomyces genes. Microb. Cell 2, 280–287 (2015).
Varghese, D. M., Nussinov, R. & Ahmad, S. Predictive modeling of moonlighting DNA-binding proteins. NAR Genom. Bioinform. 4, lqac091 (2022).
Curtis, N. J. & Jeffery, C. J. The expanding world of metabolic enzymes moonlighting as RNA binding proteins. Biochem. Soc. Trans. 49, 1099–1108 (2021).
Niebel, B., Leupold, S. & Heinemann, M. An upper limit on Gibbs energy dissipation governs cellular metabolism. Nat. Metab. 1, 125–132 (2019).
Losa, J. et al. Perspective: a stirring role for metabolism in cells. Mol. Syst. Biol. 18, e10822 (2022).
Palme, J., Wang, J. & Springer, M. Variation in the modality of a yeast signaling pathway is mediated by a single regulator. eLife 10, e69974 (2021).
Ricci-Tam, C. et al. Decoupling transcription factor expression and activity enables dimmer switch gene regulation. Science 372, 292–295 (2021).
Gietz, R. D. & Woods, R. A. Transformation of yeast by lithium acetate/single-stranded carrier DNA/polyethylene glycol method. Methods Enzymol. 350, 87–96 (2002).
Shaw, W. M. et al. Engineering a model cell for rational tuning of GPCR signaling. Cell 177, 782–796.e27 (2019).
Lee, M. E., DeLoache, W. C., Cervantes, B. & Dueber, J. E. A highly characterized yeast toolkit for modular, multipart assembly. ACS Synth. Biol. 4, 975–986 (2015).
Khmelinskii, A. et al. Incomplete proteasomal degradation of green fluorescent proteins in the context of tandem fluorescent protein timers. Mol. Biol. Cell 27, 360–370 (2016).
Palme, J., Li, A. & Springer, M. Data for: the galactokinase enzyme of yeast senses metabolic flux to stabilize GAL pathway regulation. Dryad https://doi.org/10.5061/dryad.t4b8gtj7b (2024).
Acknowledgements
We thank B. Ward, O. Weiner and Q. Justman for valuable comments on the paper and all members of the Springer laboratory for critical discussions. This work was funded by the National Institutes of Health (grant nos. R01-GM120122 and 5 R01 GM148497-02) and the National Science Foundation (grant no. MCB-1349248).
Author information
Authors and Affiliations
Contributions
Conceptualization: J.P., A.L. and M.S. Formal analysis: J.P., A.L. and M.S. Funding acquisition: M.S. Investigation: J.P. and A.L. Methodology: J.P., A.L. and M.S. Writing—original draft: J.P. and M.S. Writing—review and editing: J.P. and M.S.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Metabolism thanks Radhakrishnan Mahadevan, Bas Teusink and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Alfredo Giménez-Cassina, in collaboration with the Nature Metabolism team.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data
Extended Data Fig. 1 Schematic of GAL signaling.
a, Signaling reactions (green area) and metabolic reactions (blue area) in the GAL pathway. α-D-galactose (α-D-gal) and its anomer β-D-galactose (β-D-gal) interconvert (spontaneously or catalyzed by Gal10p). Either anomer can be transported over the plasma membrane by through passive, bidirectional hexose transporters such as Gal2p or other Hxts. α-D-galactose binds the galactose sensor Gal3p, leading to the inhibition of Gal80p. This relieves the repression of Gal4p and results in the transcriptional upregulation of GAL genes. α-D-galactose can also bind to Gal1p, leading to galactose phosphorylation, producing galactose-1-phosphate (gal-1-P). Galactose-1-phosphate is further metabolized to glucose-1-phosphate (glc-1-P) by the galactose-1-phosphate uridyl transferase Gal7p. This reaction also converts UDP-glucose (UDP-glc) to UDP-galactose (UDP-gal). Glucose-1-phosphate converted to the glycolytic intermediate glucose-6-phosphate (glc-6-P) by the phosphoglucomutase Pgm1p. UDP-glucose is regenerated by the UDP-glucose-4-epimerase Gal10p. α-D-galactose binding to Gal1p can also lead to the inhibition of Gal80p, similar to the activity of the galactose sensor Gal3p. b, Signaling reactions (green area) and metabolic reactions (blue area) in a simple mathematical model of nutrient sensing. Substrate molecules are transported over the plasma membrane by a passive, bidirectional hexose transporters (analogous to Gal2p). The substrate molecules bind a concentration sensor protein (analogous to Gal3p), leading to the activation of downstream signaling, or an enzyme (analogous to Gal1p), leading to substrate degradation. Additionally, activation of signaling by the substrate–bound enzyme (dashed line) was included when appropriate. See Supplementary Note 1 for model equations.
Extended Data Fig. 2 SpGal1p activity in S. cerevisiae.
a, Growth curves of strains with the wild-type GAL1 CDS (ho∆::GAL1pr-mVenus, blue lines) or the SpGAL1 CDS (ho∆::GAL1pr-mVenus gal1∆::GAL1pr-spGAL1, purple lines) at the GAL1 locus. Single colonies were picked into rich medium (YEPD) and grown for 24 hours, then diluted 1:100 into minimal medium with 128 mM galactose and no glucose. After 16 hours, cultures were back-diluted to OD600 = 0.05 and OD600 was measured for 10 hours. Lines connect data points from independent replicates. b, Effect of GAL3, GAL1, and SpGAL1 expression on GAL signaling at different extracellular galactose concentrations and expression levels (inducer concentrations). Genes were expressed under control of the TetO7 promoter in a gal1∆ gal3∆ background with a fluorescent reporter for GAL signaling (GAL1pr-mVenus). Fluorescent protein expression was quantified by flow cytometry. Cells were grown in a constant environment of the indicated galactose concentration and 4 mM glucose.
Extended Data Fig. 3 Correlation between galactose flux and Gal1p or Gal3p sensing.
a-b, Replicate experiments for Fig. 2.
Extended Data Fig. 4 Model characteristics in the flux limit regime.
a, Relationship between the flux control coefficient of the flux sensor enzyme (left) or the transporter (right) and the flux sensor level. Control coefficients were calculated for the model of the intracellular substrate concentration (Supplementary Note 1). b, Relationship between the intracellular substrate concentration (Si) and the flux sensor level. The gray line in c is a guide to the eye representing a linear relationship (that is a slope of -1). c, Relationship between the net transport (Jtransport, equal to the pathway flux and flux sensor flux) and the flux sensor level. Jtransport is given by Jimport − Jexport The dashed, horizontal line represents a limit in the net transport flux where Jexport = 0. d, Predicted relationship between import flux and the flux sensing limit. Horizontal lines describe approximated flux limit plateaus at given extracellular nutrient concentrations (indicated by the line color) when the signaling limit is not considered (left) or considered (right). Flux limit plateaus were calculated based on the import flux as described in Supplementary Note 4. Modeling predictions are reproduced from Fig. 3a. Dashed lines in a-d represent regime limits and are reproduced from Fig. 3a.
Extended Data Fig. 5 Gal1p titration at different GAL2 expression levels.
GAL signaling at different Gal2p levels in gal1∆ gal3∆ TetO7pr-GAL1-mScarlet-I gal2∆::Xpr-GAL2 strains at different Gal1p levels. GAL signaling and Gal1p levels were determined and analyzed as described in Fig. 1d. Strains were grown in a constant environment of 4 mM galactose and 4 mM glucose. For the background correction with the 0 mM galactose conditions (see Methods), data from the GAL2pr strain was used for all other strains.
Extended Data Fig. 6 Endogenous Gal1p levels at different external galactose concentrations.
Endogenous Gal1p levels in a GAL1pr-GAL1-mScarlet-I strain at different external galactose concentrations. GAL signaling and Gal1p levels were determined as described in Fig. 1d. The regime threshold was determined as described in Fig. 3c with data that was collected in the same experiment as the Gal1p level measurements. Lines connect data from independent replicates.
Extended Data Fig. 7 Effect of sensor levels on Gal1p and Gal3p sensing.
Comparison of replicate experiments for Fig. 3.
Extended Data Fig. 8 Flow cytometry data of strains.
Histograms of GAL signaling of strains with wild-type GAL80 expression (a), gal80∆::CYC1pr-GAL80 expression (b) or gal2∆::ACT1pr-GAL2 and wild-type GAL80 expression (c) measured by flow cytometry after growth in the indicated galactose concentration and 4 mM glucose for 8 hours. Lines connect data from independent replicates.
Extended Data Fig. 9 Stability of GAL signaling in microscopy experiments.
a, wild-type GAL80 strains grown in the presence of 4 mM galactose (wild-type GAL1) or 128 mM galactose (gal1∆ and gal1∆::GAL1pr-SpGAL1). b, gal1∆ in a gal80∆:CYC1pr-GAL80 background grown in the presence of 1 mM galactose. b, gal1∆ strain grown in the presence of 4 mM galactose. c, gal2∆::ACT1pr-GAL2 strain grown in the presence of 4 mM galactose. d, Replicate experiments of Fig. 4. For each genotype and replicate, 50 single-cell time courses were randomly selected from the population of cells whose GAL signaling was in the top 20% after 10 hours in a medium with 4 mM glucose and the specified concentration of galactose.
Extended Data Fig. 10 Influence of model parameter choices on the flux limit (regime II).
a, Definition of flux limit criteria. Increasing the sensor level at a given extracellular substrate concentration and transporter level are defined to have reached the flux limit when the limit level criterion is met at the maximal sensor level and the signaling saturation and sensor level criteria are met for at least one sensor level. b, One-dimensional parameter scan. A set of parameter choices is defined to lead to the flux limit when the flux limit is observed in at least one extracellular substrate concentration. Encircled numbers reference parameter choices for which titration curves are shown in the next panel. c, Examples of parameter choices where the flux limit is not observed for the flux sensor (referenced by the encircled numbers).
Supplementary information
Supplementary Information
Supplementary Note 1: Nutrient concentration, metabolism and sensing model. Supplementary Note 2: Effect of enzyme reversibility. Supplementary Note 3: Quantitative definition of signalling regimes. Supplementary Note 4: Signalling in the flux and signalling limits.
Supplementary Tables
Supplementary Table 1: Strains used in this study. Supplementary Table 2: Plasmids used in this study. Supplementary Table 3: Oligonucleotides used in this study. Supplementary Table 4: Parameters used in the mathematical model.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Palme, J., Li, A. & Springer, M. The galactokinase enzyme of yeast senses metabolic flux to stabilize galactose pathway regulation. Nat Metab 7, 137–147 (2025). https://doi.org/10.1038/s42255-024-01181-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s42255-024-01181-x
