Glucose feeds the tricarboxylic acid cycle via excreted ethanol in fermenting yeast

Ethanol and lactate are typical waste products of glucose fermentation. In mammals, glucose is catabolized by glycolysis into circulating lactate, which is broadly used throughout the body as a carbohydrate fuel. Individual cells can both uptake and excrete lactate, uncoupling glycolysis from glucose oxidation. Here we show that similar uncoupling occurs in budding yeast batch cultures of Saccharomyces cerevisiae and Issatchenkia orientalis. Even in fermenting S. cerevisiae that is net releasing ethanol, media 13C-ethanol rapidly enters and is oxidized to acetaldehyde and acetyl-CoA. This is evident in exogenous ethanol being a major source of both cytosolic and mitochondrial acetyl units. 2H-tracing reveals that ethanol is also a major source of both NADH and NADPH high-energy electrons, and this role is augmented under oxidative stress conditions. Thus, uncoupling of glycolysis from the oxidation of glucose-derived carbon via rapidly reversible reactions is a conserved feature of eukaryotic metabolism. Metabolic labeling experiments on fermenting yeast revealed that ethanol is oxidized to acetaldehyde and acetyl-CoA and is also a major source of NADH and NADPH, demonstrating that ethanol can be consumed as a TCA cycle and redox fuel.

F ermentation occurs widely across kingdoms, converting glucose into organic waste products [1][2][3][4][5] . In mammals, the main such product is lactate. Until recently, it was commonly assumed that the liver and kidney were special in their capacity to clear circulating lactate, reconverting the waste (lactate) into fuel (glucose). New evidence suggests, however, that most mammalian tissues take up circulating lactate and oxidize it via the tricarboxylic acid (TCA) cycle 6 . Indeed, it seems that most carbohydrate oxidation in mammals, rather than occurring by a tissue taking up glucose and fully oxidizing it to carbon dioxide, instead involves carbon flowing through circulating lactate as a metabolic intermediate. Thus, glycolysis is uncoupled from the TCA cycle via cellular uptake and/or excretion of lactate 6 . Biochemically, this occurs through the rapidly reversible reactions linking intracellular pyruvate, via lactate dehydrogenase and monocarboxylate transporters, to circulating lactate.
Baker's yeast (S. cerevisiae) is a prototypical fermentative unicellular organism 7 . Its rapid catabolism of glucose into ethanol + CO 2 plays a central role in human society, contributing to such diverse fields as baking, beverages and biofuels. S. cerevisiae is capable of growing aerobically on substrates including galactose, glycerol and ethanol. But when provided with ample glucose, it will ferment even in the presence of adequate oxygen, a phenomenon known as the Crabtree effect 8 . When glucose runs out, after a delay to rewire metabolism, aerobic growth will resume (the diauxic shift) 9 . Some other budding yeast, such as Issatchenkia orientalis, engage in extensive oxidative metabolism even when glucose is present (that is, are Crabtree negative) and do not show a diauxic shift but, nevertheless, secrete some ethanol. As glucose fermentation in yeast parallels aerobic glycolysis in mammals, we were curious whether it similarly involves reversible excretion and uptake of the 'waste product' (ethanol) rather than distinct phases of waste production and subsequent consumption. We further wondered whether any such ethanol uptake during net fermentative metabolism might contribute to yeast's metabolic robustness.
Understanding these questions is relevant both for basic science and for bioengineering, with ethanol uptake undesirable in yeast deployed for producing ethanol as biofuel. With these motivations in mind, we show that, even when fermenting, yeasts actively exchange environmental ethanol for intracellular acetaldehyde at a sufficiently rapid rate that intracellular acetyl units come substantially from environmental ethanol, in addition to directly from glucose. Moreover, such exchange enables ethanol to be a major source of NADH and NADPH, especially under oxidative stress conditions.

Fermenting baker's yeast assimilates environmental ethanol.
Ethanol can enter and exit cells via simple diffusion 10 . Thus, exogenous ethanol may enter yeast, even if they are simultaneously excreting ethanol made internally from glucose. To differentiate two-carbon (2 C) units from environmental ethanol versus internal glucose catabolism, we grew yeasts in typical minimal media (yeast nitrogen base, aerated, 30 °C) with unlabeled glucose until mid-exponential phase. We then pelleted the cells and resuspended them in yeast nitrogen base containing both glucose and ethanol, whose isotopic composition was under experimental control. The glucose and ethanol concentrations in the resuspension media were selected to approximate those naturally occurring during mid-exponential S. cerevisiae growth in yeast nitrogen base with glucose as the carbon source (recognizing that by mid-exponential phase yeast will have converted a substantial amount of glucose into ethanol). Specifically, we provided glucose and ethanol at either equimolar concentrations (42 mM each, 'equimolar') or a 1:1 mixture based on the number of carbon atoms (28 mM glucose and 84 mM ethanol, 'equicarbon') ( Fig. 1a).
We used 13 C NMR to measure rates of glucose uptake (f gluc_up ) and conversion to environmental ethanol via pyruvate decarboxylase (f 13C_etoh_out ) from the S. cerevisiae cultures with [U-13 C]glucose and unlabeled ethanol ( Fig. 1b and Extended Data Fig. 1). The rates measured by 13 C NMR (f gluc_up and f 13C_etoh_out ) are similar among Glucose feeds the tricarboxylic acid cycle via excreted ethanol in fermenting yeast Tianxia Xiao 1,2 , Artem Khan 1,4 , Yihui Shen 2 , Li Chen 2,5 and Joshua D. Rabinowitz 1,2,3 ✉ Ethanol and lactate are typical waste products of glucose fermentation. In mammals, glucose is catabolized by glycolysis into circulating lactate, which is broadly used throughout the body as a carbohydrate fuel. Individual cells can both uptake and excrete lactate, uncoupling glycolysis from glucose oxidation. Here we show that similar uncoupling occurs in budding yeast batch cultures of Saccharomyces cerevisiae and Issatchenkia orientalis. Even in fermenting S. cerevisiae that is net releasing ethanol, media 13 C-ethanol rapidly enters and is oxidized to acetaldehyde and acetyl-CoA. This is evident in exogenous ethanol being a major source of both cytosolic and mitochondrial acetyl units. 2 H-tracing reveals that ethanol is also a major source of both NADH and NADPH high-energy electrons, and this role is augmented under oxidative stress conditions. Thus, uncoupling of glycolysis from the oxidation of glucose-derived carbon via rapidly reversible reactions is a conserved feature of eukaryotic metabolism.
strains of S. cerevisiae with different respiratory capacity (FY4 and CEN.PK) and media substrate ratios (equimolar or equicarbon) (Extended Data Fig. 1a,b). In parallel, net ethanol flux was measured by 1 H NMR, revealing active fermentation (that is, net ethanol excretion) (Fig. 1c).
The yeast ethanol assimilation pathway involves oxidation of ethanol to acetate, which is converted into cytosolic acetyl-CoA by acetyl-CoA synthetases 11 . To trace potential ethanol uptake and use, we directly measured cellular acetyl-CoA labeling distributions by liquid chromatography-mass spectrometry (LC-MS) in S. cerevisiae   13 C-metabolic flux analysis based on model in f and data in c and e. PDC flux is substantial, whereas PDH flux is below detection, signifying that acetaldehyde is the direct contributor for acetyl-CoA. High exchange flux at ADH implicates environmental ethanol as a major contributor to cellular acetaldehyde and acetyl-CoA (mean, upper bound/lower bound, n = 3 biological replicates).
grown with unlabeled glucose and 13 C-labeled ethanol (Fig. 1d), finding substantial labeling (more than 50%) from environmental ethanol (Fig. 1e). This high [M + 2] labeled fraction of acetyl-CoA is a steady-state measurement (Extended Data Fig. 1c) and is consistent across strains and media compositions (Extended Data Fig. 1d). In these fermenting cells, ethanol carbon did not enter glycolytic intermediates, as shown by the absence of [M + 2] labeling, consistent with gluconeogenesis being inactive (Extended Data Fig. 2). We built a 13 C metabolic flux model to estimate the reversibility of the ethanol assimilation pathway (Fig. 1f). The model was constrained by the measured glucose uptake rate, the net ethanol excretion rate, PDC flux (the flux representing gross glucose conversion to ethanol) and acetyl-CoA labeling from [U-13 C]ethanol. The model confirmed low PDH and high PDC flux, as typical for fermenting S. cerevisiae (Fig. 1g). Notably, it revealed a fast exchange flux between ethanol and acetaldehyde, with ethanol a major source of acetaldehyde even though net flux is in the direction of ethanol excretion (Fig. 1g). This rapid exchange flux explains the substantial acetyl-CoA labeling from environmental ethanol (Fig. 1e).

Environmental ethanol contributes to fatty acid synthesis.
Acetyl-CoA exists as discrete cytosolic and mitochondrial pools. Fatty acid synthesis uses cytosolic acetyl-CoA (Fig. 2a); thus, fatty acid labeling selectively represents cytosolic acetyl-CoA labeling. In S. cerevisiae fed either the equimolar or equicarbon mixture of unlabeled glucose and [U-13 C]ethanol, most of the carbon in newly synthesized fatty acids (that is, those containing at least some label) was from environmental ethanol ( Fig. 2b and Extended Data Fig. 3).
To quantify the fraction of cytosolic acetyl-CoA coming from environmental ethanol, we fit the observed fatty acid mass isotope distribution to a binomial, reflecting the fact that each 2 C unit incorporated into fat is selected stochastically, with the assumption that only labeled fatty acids are newly synthesized. The simple binomial fit well, consistent with a homogeneous environmental ethanol contribution across different cells in the population of around 60% of lipogenic acetyl-CoA (Fig. 2c). Thus, rather than being derived mainly internally by glycolysis and subsequent pyruvate catabolism, when environmental ethanol is present, cytosolic acetyl-CoA in baker's yeast comes also from ethanol.

Environmental ethanol supplies mitochondrial acetyl-CoA.
Formation of cytosolic acetyl-CoA from acetate is catalyzed by acetyl-CoA synthetases 12,13 . Such synthetases are not known in yeast mitochondria. Accordingly, we were curious whether environmental ethanol could also contribute to mitochondrial acetyl-CoA. To this end, using the same tracing strategy as above, we examined whether environmental ethanol would label a metabolic product that is produced mitochondrially from acetyl-CoA, N-acetylglutamate (NAG), an intermediate in the arginine biosynthesis pathway (Fig. 2d). In S. cerevisiae fed either equimolar or equicarbon unlabeled glucose and [U- 13 Fig. 4). To quantitate the fraction of mitochondrial acetyl-CoA coming from environmental ethanol, we inferred mitochondrial acetyl-CoA labeling from the observed mass isotope distribution of NAG and glutamate. The calculated [M + 2] fraction of mitochondrial acetyl-CoA is around 60% (Fig. 2f), similar to cytosolic acetyl-CoA. Thus, environmental ethanol is a major source of both cytosolic and mitochondrial acetyl-CoA.
The enzyme succinyl-CoA acetate CoA transferase (Ach1) has been proposed as a potential means of generating mitochondrial acetyl-CoA from acetate in S. cerevisiae, but its physiological role has remained unproven 14 . Ach1 deletion completely abolished [M + 4] NAG ( Fig. 2e and Extended Data Fig. 4a,b), with the inferred mitochondrial acetyl-CoA labeling zero in this deletion strain (Fig. 2f). Notably, ∆ach1 nevertheless has similar whole cell [M + 2] acetyl-CoA labeling from ethanol (Extended Data Fig. 4c,d), implying that only a small fraction of total cellular acetyl-CoA is mitochondrial, with Ach1 the key mitochondrial acetate assimilation enzyme.

TCA intermediates are made from environmental ethanol.
Acetyl-CoA contributes to the TCA cycle via citrate synthase (Fig. 3a). From [U-13 C]ethanol tracing in fermenting S. cerevisiae, we observed [M + 2] (iso)citrate, aconitate, α-ketoglutarate and succinate. Fumarate, malate and aspartate (whose carbon skeleton comes from oxaloacetate) remained, however, largely unlabeled ( Fig. 3b and Extended Data Fig. 5). The extensive labeling of succinate with limited labeling of fumarate or oxaloacetate pinpoints succinate dehydrogenase (complex II in the electron transport chain) as being functionally blocked during fermentative growth of baker's yeast 15 . Instead of being made by TCA turning, oxaloacetate and malate are generated by pyruvate carboxylase, using pyruvate made from glucose. Nevertheless, acetate from environmental ethanol is assimilated into the TCA cycle and drives conversion of these four-carbon TCA intermediates into citrate, α-ketoglutarate and α-ketoglutarate's amino acid products.
To explore the generality of ethanol's contributing to acetyl-CoA and, thereby, TCA intermediates in glucose-fed budding yeast, we carried out analogous experiments in I. orientalis, Crabtree-negative yeast diverged from S. cerevisiae roughly 200 million years ago. In both equimolar and equicarbon conditions, [U 13 C]ethanol generated [M + 2] labeled TCA intermediates to a similar extent in both yeast species (Fig. 3b,c). In I. orientalis, we also observed more heavily labeled TCA intermediates indicative of active ethanol assimilation and full TCA turning (Extended Data Fig. 5). Thus, assimilation of environmental ethanol in the presence of glucose is a feature of both Crabtree-positive and Crabtree-negative budding yeasts.

Concentration dependence of ethanol use.
Although reflecting the levels of ethanol typically found in dense fermenting yeast cultures, the equimolar and equicarbon conditions (28 mM and 42 mM ethanol, respectively) contain more ethanol than is found in many physiological environments. For example, early log-phase growth of S. cerevisiae in glucose batch culture (initial optical density of 0.1 OD grown for 1 hour) generates ethanol concentrations around 0.5 mM 16,17 (Fig. 1c). At this concentration, ethanol accounts for less than 1% of carbon in the media (>99% being glucose). Nevertheless, the 13 C-ethanol contributes discernibly to the cellular acetyl-CoA in S. cerevisiae and is a major source in I. orientalis (Fig. 3d,e). At 5 mM concentration, ethanol becomes a major acetyl-CoA source also in S. cerevisiae. Thus, the importance of environmental ethanol as an acetyl-CoA source depends on concentration and is substantial at ethanol levels found in moderate to dense glucose-fed yeast cultures (Extended Data Fig. 6).
Acetaldehyde oxidation feeds NADPH. Ethanol re-assimilation has redox consequences. Ethanol conversion to acetaldehyde generates NADH. Further oxidation of acetaldehyde into acetate via aldehyde dehydrogenase generates NADPH. When the canonical main NAPDH production pathway, the oxidative pentose phosphate pathway, is deleted, the acetaldehyde dehydrogenase Ald6 is essential for yeast growth 18 .
To measure redox cofactor contributions from environmental ethanol, we transferred fermenting S. cerevisiae into glucose:ethanol as above, with either the glucose or the ethanol deuterated at position 1. Specifically, we compared NADH and NADPH labeling from [1-2 H]glucose (the labeled hydrogen is transferred to NADPH via G6PD, encoded by gene zwf1) and [1,1-2 H 2 ]ethanol (the labeled hydrogen is transferred to NADH by ADH and to NADPH via Ald). Direct measurement of 2 H-labeling in NADH and NADPH is technically challenging due to limited abundance and stability, but,   the resulting 13 C-labeled acetyl-CoA is incorporated into newly synthesized fatty acids. As both labeled and unlabeled cytosolic acetyl-CoA are randomly incorporated into growing fatty acid chains, the resulting fatty acid mass isotope distribution follows a binomial probability distribution. b, Fatty acid (palmitate) labeling pattern from equimolar glucose: 13 C-ethanol co-feeding experiment as in Fig. 1d (mean, s.e., n = 3 biological replicates). In brief, newly synthesized fatty acids are getting labeled by [M + 2] acetyl-CoA, which is a result of 13 C-ethanol uptake from growth media by S. cerevisiae. c, Cytosolic acetyl-CoA labeling fitted from fatty acid labeling from equimolar glucose: 13 C-ethanol co-feeding experiment as in Fig. 1d (mean, s.e., n = 6, three biological replicates with results from both C16:0 and C18:0) and whole-cell data from Fig. 1e (mean, s.e., n = 3 biological replicates). d, Synthesis of the arginine precursor N-acetylglutamate (NAG) in S. cerevisae takes place in mitochondria (created with BioRender). A linear algebra deconvolution of the labeling fractions of glutamate and NAG can compute the mitochondrial acetyl-CoA labeling. e, Glutamate (Glu) and NAG labeling from the glucose: 13 C-ethanol co-feeding experiment as in Fig. 1d, including also data for ∆ach1 yeast (thereby identifying Ach1 as an enzyme essential for mitochondrial assimilation of ethanol-derived carbon into acetyl-CoA) (mean, s.e., n = 3 biological replicates; ***P < .001 by two-sided t-test   Fig. 7a-f).
To obtain more precise and compartment-specific information, we used fatty acid labeling to read out cytosolic NADPH labeling. Fatty acid synthesis incorporates two NADPH hydrides per acetyl group (Fig. 4a) 19 . Strikingly, we observed greater deuterium labeling of fatty acids from [1,1-2 H 2 ]ethanol than from [1-2 H]glucose (Extended Data Fig. 8). This reflects a major contribution of acetaldehyde to cytosolic NADPH via Ald6, of yet greater magnitude than the contribution of glucose-6-phosphate to cytosolic NADPH via the oxPPP (Fig. 4b,c).
To convert the observed labeling into quantitative contributions to NADPH, we need to account for deuterium loss from NADPH via hydrogen-deuterium exchange with water mediated by flavin enzymes 19,20 . Experiments culturing cells in D 2 O revealed that about half of cytosolic NADPH hydrogen nuclei come from water via hydrogen-deuterium exchange. Such exchange does not account for any NADPH's high-energy electrons but merely dilutes deuterium tracer signal from the actual hydride donors such as [1-2 H]glucose or [1-2 H]acetaldehyde (Extended Data Fig. 7g-i). Correcting for such exchange (and for the extent of substrate labeling), we observed that the oxPPP and Ald6 together account for most cytosolic NADPH, with the contribution of ethanol via Ald6 roughly double that of glucose via the oxidative pentose phosphate pathway (Fig. 4d).
Consistent with ethanol oxidation and oxPPP being alternative cytosolic NADPH production pathways, in ∆ald6, oxPPP contribution to NADPH production (based on fatty acid labeling patterns) is nearly twice as high as in wild-type ( Fig. 4c and Extended Data Figs. 8a,b and 9). In ∆zwf1, Ald6 contribution to NADPH production (based on fatty acid labeling patterns) similarly doubles ( Fig. 4b and Extended Data Figs. 8c,d and 9). Thus, ethanol is an important source of both acetyl and hydride units in baker's yeast.

Ethanol becomes a greater NAD(P)H source upon H 2 O 2 stress.
We were curious whether S. cerevisiae cells might shift between glucose or ethanol as NAD(P)H sources in response to environmental conditions. To explore this possibility, we grew yeast in glucose:ethanol with one substrate 2 H-labeled as above, spiked in H 2 O 2 to a final concentration of 20 mM, and rapidly sampled metabolites and their labeling 21,22 (Fig. 5a). Upon adding H 2 O 2 , the NADH concentration and NADH/NAD + ratio fell markedly (Fig. 5b,c). Such a drop was expected, given that oxidative stress is known to oxidize the GADPH's active site cysteine and, thereby, block glycolytic flux and NADH production 23 . Consistent with GAPDH being shut off, in addition to an increase of fructose-1,6-bisphosphate (FBP) (Extended Data Fig. 10), we observed increased NADH labeling from ethanol, which became the dominant NADH hydride source (Fig. 5d). Thus, ethanol catabolism is a crucial source for NADH when glycolysis is blocked by oxidative stress.
A classical rationale for glycolytic blockade by oxidative stress is to divert flux into the oxidative pentose phosphate pathway to help maintain NADPH homeostasis. The same concentration of hydrogen peroxide that markedly suppressed NADH had no overt effect on NADPH pool size or the NADPH:NADP + ratio (Fig. 5e,f). However, rather than increasing the fractional oxidative pentose phosphate pathway contribution to NADPH as measured by [1-2 H]glucose, this contribution was decreased, with ethanol's fractional contribution to NADPH markedly increased (Fig. 5g). Thus, in contrast to the common assumption that the  , n = 6, three biological replicates with results from both C16:0 and C18:0; ***P < .001 by two-sided t-test). c, Cytosolic NADPH labeling as in b from [1-2 H]glucose in wild-type S. cerevisiae and ∆ald6 strain multiplied by 2 to account both G6PD and 6PGD fluxes in oxPPP (mean, s.e., n = 6, three biological replicates with results from both C16:0 and C18:0; ***p < .001 by two-sided t-test). d, Summary of data from b and c, and D 2 O tracing into fat shows sources of cytosolic NADPH redox-active hydrogen nucleus in wild-type S. cerevisiae FY4 (mean, s.e., n = 6, three biological replicates with results from both C16:0 and C18:0). The redox-active hydrogen nucleus, but not the associated high-energy electrons, is in a rapid H-D exchange with water, which explains the fractional contribution not accounted for by the pentose phosphate pathway and Ald6. WT, wild-type.
predominant NADPH production route during oxidative stress is the oxidative pentose phosphate pathway, we observe a substantial and augmented NADPH contribution from ethanol oxidation under H 2 O 2 stress 18,24 .

Discussion
A fundamental metabolic question is, 'Which pathways are coupled versus independent?' Here we present evidence that uncoupling of glycolysis from the TCA cycle is an evolutionarily conserved design principle in eukaryotic metabolism. Specifically, we show that fermenting budding yeast simultaneously release and uptake ethanol, much as many mammalian cells simultaneously produce and consume circulating lactate. Both lactate and ethanol are redox-balanced with glucose 25 . Thus, their release allows glycolysis to run without need for the TCA cycle or oxidative phosphorylation. Release of these electron-rich products anticorrelates with internal NADH consumption by the electron transport chain 26 and can be suppressed by inducing synthesis pathways of other electron-rich molecules such as fats 27 .
Although the net release of ethanol by fermenting yeast has been long appreciated, we are unaware of prior demonstration that fermenting yeast cultures simultaneously engage in extensive ethanol uptake. Through experiments with 13 C-ethanol, we show that, under typical mid-exponential fermentative growth conditions, environmental ethanol, rather than mitochondrial pyruvate catabolism, supplies a majority of both cytosolic and mitochondrial acetyl-CoA. This observation aligns with prior literature finding that, when [U 13 C]glucose is spiked into fermenting yeast cultures (which naturally contain ethanol), acetyl-CoA is labeled less than other central metabolites 28 . The importance of the ethanol assimilation pathway depends on the environmental ethanol concentration, with ethanol becoming a major acetyl-CoA source at millimolar level. In mitochondria, we prove that the ethanol assimilation pathway involves the CoA-transferase Ach1 [29][30][31] . The assimilated ethanol was originally produced from glucose. But, at the population level, the pathway from glycolysis to the TCA cycle (and other acetyl-CoA products such as amino acids and fatty acids) flows through pyruvate decarboxylase and environmental ethanol. Temporal changes within the first 60 seconds after H 2 O 2 shock are captured by rapid quenching of metabolism at time points of 15 seconds, 30 seconds and 60 seconds. b, NADH pool size (mean, s.e., n = 6 biological replicates, negative linear trend, P = .0016, ordinary one-way ANOVA). c, NADH:NAD + ratio (mean, s.e., n = 6 biological replicates, negative linear trend, P = .0006, ordinary one-way ANOVA). d, NADH active hydride labeling from [1,1-2 H 2 ] ethanol (orange) and [1-2 H]glucose (gray) (mean, s.e., n = 3 biological replicates, P = .011(*), two-tailed paired t-test). Increase in the NADH active hydride labeling from labeled ethanol is statistically significant (positive linear trend, P < 0.001, ordinary one-way ANOVA). e, NADPH pool size (mean, s.e., n = 6 biological replicates). f, NADPH:NADP + ratio (mean, s.e.m., n = 6 biological replicates). g, NADPH active hydride labeling from [1,1-2 H 2 ]ethanol (orange) and [1-2 H]glucose (gray) (mean, s.e., n = 3 biological replicates, P = .02, two-tailed paired t-test). Increase in the NADPH active hydride labeled by ethanol is statistically significant (positive linear trend, P < .001, ordinary one-way ANOVA).
Our methods cannot differentiate whether this flux through environmental ethanol involves intercellular ethanol exchange (that is, from ethanol-secreting fermentative cells to ethanol-consuming oxidative cells) or pathway reversibility at the single-cell level (that is, simultaneous ethanol secretion and re-uptake by the same cells). An intriguing possibility is that ethanol re-uptake coordinates with the yeast metabolic cycle 32,33 , occurring primarily in cells in the metabolic cycle's oxidative phase, with ethanol produced by cells in the reductive building and charging phases.
Simultaneous ethanol excretion and uptake simplify regulation of the fate of pyruvate, circumventing the challenge of partitioning it optimally between fermentation and acetyl-CoA. Ethanol excretion is the default, with the resulting environmental ethanol providing a reservoir to help meet cellular two-carbon unit demands. Such a reservoir is not essential, as yeast can grow rapidly in dilute cultures where ethanol is scarce or in contexts where excreted ethanol is diluted away by fluid flow 33 . But when available, environmental ethanol helps assure adequate availability of both carbon and high-energy electrons even if glycolysis is impaired.
Notably, because simultaneous ethanol excretion and re-assimilation is a default state, yeast can access both the carbon and high-energy electrons from environmental ethanol without any remodeling of their internal metabolic machinery. Such access is particularly evident during acute redox stress, which impairs glycolysis through inhibitory oxidation of the central glycolytic enzyme GAPDH. Under this circumstance, ethanol becomes the predominant source of both NADH and NADPH, the latter being critical for survival of oxidative stressors. Thus, uncoupling of glycolysis from the TCA cycle via ethanol provides yeast with metabolic flexibility, decreases regulatory complexity and enhances robustness.

online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41589-022-01091-7.

Materials.
Yeasts were grown in yeast nitrogen base without amino acids (Sigma-Aldrich, Y0626) with carbon source added separately. BD Difco YPD Broth (BD, 242820) was used as the media for reviving frozen cells or growing cells to be frozen. Glycerol (Sigma-Aldrich, G5516) was added at a 1:1 volume ratio to YPD yeast cultures in cryovials (Nalgene, 5000-1020, or Corning, 430289). Glucose (Sigma-Aldrich, D9434), [U- 13  Yeast strains. S. cerevisiae strain FY4, derived from S288c, was taken from in-house frozen stocks. S. cerevisiae strain CEN.PK was obtained from José Avalos. S. cerevisiae prototrophic mutant strains were obtained from David Botstein, which were also derived from S288c through a diploid intermediate strain 34,35 . I. orientalis SD108 was provided by Huimin Zhao.
Yeast batch cell culture growth. S. cerevisiae or I. orientalis colony was inoculated into an overnight culture containing 6.7 g L −1 of Yeast Nitrogen Base (YNB) without amino acids and 20 g/L of glucose. After 24 hours of growth at 30 °C, the overnight culture was diluted 1:100 into appropriate experimental media, containing 6.7 g L −1 of YNB without amino acids and a carbon source/isotope tracer as specified for each experiment. For the prototrophic ∆zwf1 strain, 20 mg L −1 of methionine was added to the media to accelerate growth. We observed that unlabeled ethanol present in the media also gives a detectable 13 C NMR signal (s, 1 C at 57 p.p.m.) due to natural 13 C abundance. This natural abundance signal was not corrected for because it is <5% signal of [U-13 C] ethanol. Relaxation time (D 1 ) was set to a generous length of 40 seconds to ensure that all spins reset properly and, thus, achieved quantification of 13 C nuclei.
Flux calculation from NMR measurements. NMR was used to measure the concentration of glucose and ethanol in the media at the time of switching (Conc t=0 ) and after 1-hour incubation (Conc t=1 ). The glucose concentrations of the same samples were independently verified by a biochemistry analyzer. OD was measured at the time of switching (OD t=0 ) and 1 hour after incubation (OD t=1 ). The integral of cell density over time was approximated by the trapezoid rule, which is calculated as 1 2 (ODt=1 + ODt=0) × Δt. Then, the uptake rate for glucose or excretion rate of ethanol are Conct=1−Conct=0 1 2 (ODt=1+ODt=0)×Δt with Δt = 1 hour. Metabolite extraction. Metabolite extraction was performed as previously 17,36 . Three milliliters of the cell culture at mid-exponential stage of growth (OD ~0.5) was vacuum filtered (0.45 µm, Millipore). For experiments evaluating ethanol concentration dependence, cells at an early stage of growth (OD = 0.1) were switched into YNB+ 55 mM unlabeled glucose and 50 μM, 500 μM or 5 mM [U-13 C]ethanol via centrifugation (1,600 r.p.m. at 4 °C for 1 minute) and incubated for 1 hour, and then 3 ml of cell culture was vacuum filtered (0.45 µm, nylon). The filter was immersed in 1.6 ml of a cold (−20 °C) extraction buffer (40% methanol:40% acetonitrile:20% water:0.5% formic acid by volume). Cells were washed out of the filter, and 88 µl of NH 4 (HCO 3 ) (15 % w/v, Sigma-Aldrich) solution per 1 ml of extraction buffer was added to neutralize formic acid. The resulting solutions were transferred to Eppendorf tubes and centrifuged for 10 minutes at 16,000 r.p.m. in a cold room (4 °C). The supernatant was collected and stored at −80 °C before loading onto a liquid chromatography mass spectrometer.
Water-soluble metabolite LC-MS analysis. Prepared samples were loaded onto a quadrupole-orbitrap mass spectrometer (Q Exactive Plus, Thermo Fisher Scientific) coupled to hydrophilic interaction chromatography (HILIC) for analysis. Measurements of acetyl-CoA labeling were achieved by reversed-phase ion-pairing liquid chromotography 37 coupled to a standalone orbitrap (Exactive, Thermo Fischer Scientific).
Fatty acid extraction and LC-MS analysis. Fatty extraction was performed according to Zhang et al. 19 . Cells were pelleted in a 1.5-ml Eppendorf tube, and 1 ml of 0.3 M KOH in 90:10 methanol:water solution was added. The resulting mixture with cells was transferred to a 4-ml glass vial. Saponification was performed by placing the samples into a water bath at 80 °C for 1 hour. Once the samples cooled down, 100 µl of formic acid (0.5%) was added, followed by 1 ml of hexane to extract the fatty acids. The extract was transferred into a glass HPLC vial and dried under nitrogen flow. Afterwards, it was diluted in 0.1 ml of 50:50 acetonitrile:methanol solution. The 0.1-ml solution was then added to a clean glass insert, placed inside an HPLC vial and cap sealed. All the samples were loaded onto the Exactive LC-MS employing a reversed-phase LC column (C8) coupled with negative-mode ESI high-resolution MS. NADPH or acetyl-CoA labeled fractions were inferred from observed fatty acid labeling patterns using a binomial model with unlabeled fat, which, in part reflects environmental contamination, omitted from the calculation.
Isotope tracing experiments. Cells were grown in YNB + 10 g L −1 of glucose up to OD = 0.5. Then, the cells were quickly centrifuged, the supernatant discarded and switched to equimolar or equicarbon media with either glucose or ethanol 13 C-or 2 H-labeled or with both carbon sources unlabeled in 50% 2 H 2 O. The cells were allowed to grow in the labeled media for 1 hour before harvesting. A potential concern in these experiments is dilution of labeled ethanol tracer by unlabeled ethanol made from glucose. As no more than 2 mM unlabeled ethanol is excreted into the media during the 1-hour incubation, environmental ethanol remains more than 95% fully labeled during the experiments. OD was taken before and after the incubation period to ensure that the cells were in exponential growth phase during the experiment.

Glutamate
Acetyl-CoA Acetyl-glutamate Data analysis and visualization. El-MAVEN version 11.1 (Elucidata) software was used to process the LC-MS data 38 . Metabolite identities were verified by both mass/charge (m/z) ratio and retention time match to authenticated standards. For 2 H-and 13 C-isotope-labeled data analysis, natural isotope abundance correction was made according to a binomial distribution model 39 . 13 C-MFA was computed with INCA 40 . NADPH active hydride labeling and acetyl-CoA labeling from fatty acids were calculated as previously 19 . MestReNova x64 software was used for the NMR data processing. Statistical analyses were performed with GraphPad Prism, including two-tailed t-tests (with false discovery rate correction by the two-stage step-up Benjamini, Krieger and Yekutieli method to confirm that any reported significant results involving statistical comparisons of multiple isotopic forms of the same metabolite remain significant after correction for the multiple comparisons); ordinary one-way ANOVA (when row matching is statistically significant, repeated-measures one-way ANOVA instead); and linear trend (between-column mean and by left-to-right column order, with P value from F-test). For routine data visualization and analysis, MATLAB 41 , R Studio, Python and Microsoft Excel were used. Schematics and diagrams were created with the aid of GraphPad Prism, ChemDraw and BioRender. Fig. 6 | The environmental ethanol contribution to TCA intermediates is concentration-dependent. Labeling pattern of TCA intermediates from the indicated budding yeast grown starting at OD = 0.1 in standard high glucose media (55 mM regular glucose) with the indicated concentrations of [U-13 C]ethanol for 1 h (mean, Se, n=3 biological replicates). Note that, for the lower 13 C-ethanol concentrations, the labeled ethanol is substantially diluted by unlabeled ethanol made from the unlabeled glucose during the duration of the experiment. In S. cerevisiae, this is about a 10-fold dilution for the 50 μM condition and 2-fold for the 500 μM condition. Thus, the above labeling patterns conservatively underestimate the contribution of low concentrations of environmental ethanol to TCA intermediates.