Energy-translation coupling limits anaerobic yeast growth

Abstract

Understanding how nutrient-specific limitations shape anaerobic metabolism in Saccharomyces cerevisiae is essential for defining the physiological limits of yeast growth. Integrating chemostat physiology, multi-omic profiling, and targeted metabolic engineering under strictly anaerobic conditions, we show that yeast maintains a conserved maximum glucose uptake (~14 mmol/gDW/h) under carbon (C), nitrogen (N), and phosphorus (P) limitation, while distinct regulatory bottlenecks constrain maximal growth rate: ATP insufficiency under C and P limitation, and aminoacyl-tRNA synthetase scarcity under N limitation. Under these stresses, S. cerevisiae reallocates proteomic resources toward anabolic functions, with nutrient-specific phosphorylation networks compensating for translational stress, most pronounced under N limitation. Building on these insights, a “push–pull” strategy enhancing energy supply (VMA3) and translational capacity (WRS1) increased the maximal anaerobic growth rate by 27.2%, 47.5% and 52.5% under C, N, and P limitation, respectively. These findings reveal energy–translation coupling as the central determinant of anaerobic growth limits and provide a framework for rational strain engineering.

Saccharomyces cerevisiae must balance limited protein resources between energy generation and translation capacity. Here, authors map these trade-offs in yeast under nutrient stress, revealing a coupled energy-translation bottleneck that can be engineered to synergistically boost anaerobic growth and ethanol yields.

Introduction

The ability of microorganisms to sense and adapt to environmental fluctuations defines the limits of their growth and survival, with nutrient limitation acting as a major evolutionary force shaping metabolic networks¹. As a model eukaryotic microorganism, Saccharomyces cerevisiae exhibits remarkable metabolic plasticity, enabling dynamic shifts between respiratory and fermentative carbon fluxes. This versatility has made yeast a cornerstone system for elucidating the molecular mechanisms of nutrient sensing, proteome allocation, and growth regulation^2,3. Under aerobic, carbon-limited conditions, previous studies have revealed the molecular basis of the Crabtree effect and established quantitative principles governing proteome resource allocation^4–6. These discoveries have also informed the design of high-density aerobic fermentation processes for bulk chemicals such as fumaric acid⁷, D-lactic acid⁸, and 2,3-butanediol⁹.

In contrast, anaerobic fermentation remains both industrially indispensable and mechanistically unresolved. It offers operational simplicity, eliminates oxygen supply requirements, and achieves superior sugar conversion yields. A landmark example is the industrial-scale synthesis of L-alanine using Escherichia coli, which demonstrated the potential of anaerobic bioprocessing for amino acid production¹⁰. For S. cerevisiae, fermentative metabolism forms the metabolic foundation of the global bioethanol industry¹¹. Yet, despite this industrial significance, the regulatory network topology and dynamic stress responses of yeast under anaerobic conditions remain poorly defined. Compared to respiration, fermentation yields less ATP per glucose molecule and is therefore generally associated with lower maximum specific growth rates^12,13. However, limited ATP production alone cannot explain this growth constraint; additional physiological factors such as redox imbalance, elevated maintenance energy costs¹⁴, and decreased translational fidelity¹⁵ further restrict anaerobic proliferation.

Nutrient availability imposes highly specific physiological and metabolic constraints^16,17. Phosphorus (P), as an essential constituent of ATP and nucleic acids, modulates glycolytic flux and phosphorylation-dependent signaling^18,19. Nitrogen (N) limitation suppresses amino acid biosynthesis, thereby reducing ribosome assembly and protein translation capacity^20,21. Carbon (C) limitation, by contrast, triggers global metabolic reprogramming to maximize energy conservation and precursor utilization. Recent studies in Escherichia coli have shown that cells adopt distinct translational strategies under C, N, and P limitation, revealing nutrient-specific physiological bottlenecks and adaptive logics in prokaryotes²². Whether eukaryotic microorganisms such as S. cerevisiae adhere to similar regulatory principles under anaerobic conditions, however, remains largely unknown. Previous work in yeast has primarily focused on single-nutrient perturbations under aerobic steady-state conditions, leaving unresolved how nutrient limitations collectively influence energy metabolism, translational regulation, and growth capacity in anaerobic environments.

In this study, we address this fundamental gap by systematically characterizing S. cerevisiae in continuous anaerobic cultures under C, N, and P limitations across a broad range of specific growth rates (0.05–0.25 h⁻¹). Through the integration of metabolic flux analysis, multi-omic profiling, and targeted metabolic engineering, we reveal how nutrient-specific bottlenecks in energy metabolism and translation define the maximal achievable growth rate. Furthermore, we demonstrate that a combined “push–pull” metabolic engineering strategy simultaneously enhancing energy supply and translational capacity synergistically surpasses these physiological limits. Collectively, our findings establish a unifying framework linking nutrient-specific metabolic regulation with coupled energy-translational control, providing mechanistic insight into the determinants of anaerobic growth limits and offering a rational foundation for the design of yeast strains optimized for nutrient-limited bioprocessing.

Results

Nutrient-specific reprogramming of energy and redox metabolism under anaerobic growth

To define how nutrient limitations shape yeast physiology under strictly anaerobic conditions, we first determined the maximum specific growth rates of S. cerevisiae in C-, N-, and P-limited batch cultures. These measurements established the appropriate dilution rate range for subsequent continuous chemostat cultivation. Distinct values were observed: 0.276 ± 0.009 h⁻¹ under C limitation, 0.215 ± 0.011 h⁻¹ under N limitation, and 0.211 ± 0.007 h⁻¹ under P limitation. Continuous chemostat cultivation was subsequently performed across a range of dilution rates (C limitation: 0.05–0.25 h⁻¹; N/P limitation: 0.05–0.20 h⁻¹), each in triplicate, to capture nutrient-specific metabolic adjustments across steady states (Fig. 1a).

Fig. 1. Nutrient limitation reprograms energy and redox metabolism in anaerobic S. cerevisiae cultures.

a Continuous cultures of S. cerevisiae were established under defined C, N, and P limitations. C-limited chemostats were operated at five dilution rates (0.05, 0.10, 0.15, 0.20, and 0.25 h⁻¹), whereas N- and P-limited chemostats were maintained at four dilution rates (0.05, 0.10, 0.15, and 0.20 h⁻¹). Each condition was performed in triplicate. b Specific glucose consumption, ethanol and glycerol production, carbon dioxide evolution, and ATP production rates at varying specific growth rates. c Protein abundance of glucose transporters across increasing µ. Transporter switching occurred at μ = 0.15 h⁻¹ under C limitation, but earlier (μ = 0.10 h⁻¹) under N or P limitation, reflecting nutrient-specific control of substrate affinity. d Correlation between ATP production rate and glucose consumption rate across C-, N-, and P-limited conditions. Linear fits indicate conserved stoichiometric coupling between carbon flux and energy generation. e Intracellular energy and redox status across growth rates, including ATP pool size and NADH/NAD⁺ ratio. All data represent means ± SD (n = 3 independent experiments). Source data are provided as a Source Data file.

Across all conditions, substrate consumption and product formation rates scaled almost linearly with specific growth rate (μ) (Fig. 1b). At the same μ, P-limited cultures displayed significantly higher glucose uptake (q_glucose) and ethanol production (q_ethanol) than C-limited cultures (P value < 0.05). Despite these differences, all cultures converged on a conserved upper boundary of glucose uptake (~14 mmol/gDW/h), indicating a rigid flux ceiling. The distinct rate profiles reflected nutrient-specific modulation of transporter utilization. Under C limitation, yeast cells transitioned from higher-affinity transporters HXT6/7 (K_m = 1-2 mM) to lower-/mid-affinity transporters HXT3/4 (K_m = 2.1 mM) at μ = 0.15 h⁻¹, whereas this transition occurred earlier (μ = 0.1 h⁻¹) under N and P limitation (Fig. 1c). Early engagement of lower-affinity transporters (HXT1/3; K_m = 8.5 mM) under N and P limitation contributed to premature imbalances between substrate uptake and metabolic demand, consistent with residual glucose accumulation (Supplementary Table 1).

While glucose acquisition was constrained by a fixed upper limit, carbon partitioning among biomass, ethanol, glycerol, and CO₂ remained highly flexible. Under C limitation, carbon allocation followed a biphasic pattern, with the fractions directed toward biomass, glycerol, and ethanol increasing and then decreasing as μ rose from 0.05 to 0.2 h⁻¹, whereas CO₂ exhibited the inverse trend. N limitation promoted progressive increases in biomass and glycerol yield accompanied by decreased CO₂ evolution, while P limitation primarily altered ethanol partitioning (Supplementary Table 1). These patterns reveal that yeast maintains a rigid limit for substrate uptake yet dynamically redistributes carbon fluxes downstream to preserve energetic balance.

Consistent with protein resource allocation theory²³, the specific ATP production rate (q_ATP) scaled linearly with both q_glucose and µ across all conditions (R² > 0.94; Fig. 1d). Because accurate quantification of substrate-level phosphorylation requires strictly unidirectional fermentative fluxes, this assumption was experimentally validated using complementary approaches. Global proteomic profiling and dynamic 1-¹³C-ethanol isotopic tracing both confirmed repression of assimilation pathways and the absence of detectable reverse carbon flux from ethanol (Supplementary Fig. 1). At equivalent µ, q_ATP followed the order P > N > C limitation, reaching maxima of 23.67 ± 2.54, 21.55 ± 1.49, and 19.43 ± 0.60 mmol/gDW/h, respectively (Table 1). This hierarchy reflected distinct ATP yields per glucose (Y_ATP/Glucose), highest under P limitation (1.88 mol/mol) and lowest under C limitation (1.31 mol/mol) (Table 1). Paradoxically, despite higher ATP generation efficiency under N and P limitation, their µ_max remained lower than under C limitation. Energy balance analysis revealed substantially higher maintenance energy demands ($m_s$) under P limitation (3.10 ± 0.96 mmol/gDW/h) and N limitation (2.01 ± 0.44 mmol/gDW/h) compared to C limitation (0.53 ± 0.19 mmol/gDW/h, P value < 0.05), leading to a collapse of growth-associated ATP flux (${\mathrm{q}}_{\mathrm{ATP}}$-${\mathrm{m}}_{\mathrm{s}}$) when µ exceeded 0.15 h⁻¹ (Fig. 1b). Accordingly, intracellular ATP pools under N and P limitation were 3.4–19.1-fold lower than under C limitation, accompanied by a pronounced decline in cellular energy charge (EC) (Fig. 1e).

Table 1.

Maintenance energy coefficients and ATP yields of S. cerevisiae under anaerobic nutrient-limited conditions

	C limitation	N limitation	P limitation
$m_s$ [mmol/gDW/h]	0.53 ± 0.19	2.01 ± 0.44	3.10 ± 0.96
YATP/glucose [mol/mol]	1.31 ± 0.10	1.62 ± 0.25	1.88 ± 0.10

Estimated maintenance coefficients ($m_s$) and ATP yields per glucose (Y_ATP/Glucose) were determined for C-, N-, and P-limited chemostat cultures. Values represent means ± SD (n = 3 independent experiments).

Redox homeostasis further distinguished the nutrient-specific responses. Under C limitation, intracellular NAD⁺ peaked at 48.7 μM (μ = 0.1 h⁻¹) before declining, while the NADH/NAD⁺ ratio increased from 0.03 to 0.66, accompanied by enhanced glycerol flux for NAD⁺ regeneration. In contrast, N- and P-limited cultures maintained substantially higher NAD⁺ pools, up to ~1.33 mM under N limitation (μ = 0.2 h⁻¹), and up to 1.10 mM under P limitation (μ = 0.15 h⁻¹), with the NADH/NAD⁺ ratio displaying a transient peak before decreasing (Fig. 1e). Together with the transporter switching thresholds, these trends underscore the tight coordination between carbon flux regulation and redox balance as a defining feature of nutrient-dependent metabolic adaptation.

In summary, while N and P limitations enhanced ATP generation efficiency, they also imposed elevated maintenance energy costs and earlier onset of substrate transporter bottlenecks, leading to rapid depletion of intracellular energy reserves. In contrast, C limitation, though less efficient in ATP yield, maintained superior EC and sustained higher µ_max. This combination of a fixed substrate uptake boundary and flexible carbon allocation reflects an adaptive trade-off that enables S. cerevisiae to reconcile nutrient scarcity with energetic homeostasis under anaerobic growth.

Hierarchical energy-dependent translational control identifies nutrient-specific bottlenecks

Protein synthesis represents the most energy-intensive process in the cell, and imbalances between ATP supply and biosynthetic demand are expected to constrain translational capacity under nutrient limitation^24,25. To elucidate the molecular basis of translational regulation under C, N, and P limitation, we performed integrated multi-omics profiling including transcriptomics, proteomics, phosphoproteomics, and metabolomics across a wide range of µ. This systems-level analysis uncovered both conserved and nutrient-specific mechanisms that govern the translational process.

Pearson correlation analysis between gene expression, protein abundance, and μ (Supplementary Fig. 2) identified coherent growth-associated signatures (Supplementary Fig. 2). Forty-seven genes, primarily encoding ribosomal proteins and aminoacyl-tRNA synthetases (aaRS), showed strong positive correlations with μ (R > 0.7), highlighting the universal demand for enhanced translational capacity during rapid growth (red bars in Supplementary Fig. 2). Conversely, 72 genes were negatively correlated, representing metabolic functions downregulated at higher μ (blue bars in Supplementary Fig. 2). These trends were consistent across omics layers (Supplementary Figs. 3–6), emphasizing translation machinery as a central driver of growth acceleration^5,26,27.

Although total ribosome abundance scaled positively with μ (Fig. 2a), the stoichiometric balance between small (40S) and large (60S) subunits displayed nutrient-specific divergence. Under C limitation, the 40S/60S ratio gradually declined beyond μ = 0.1 h⁻¹ (slope = -0.21/h), coinciding with ATP depletion and reversal of the amino acid pool at μ = 0.15 h⁻¹. N-limited cells maintained higher amino acid pools (1.8–2.1-fold vs. C/P at μ = 0.1 h⁻¹) but exhibited ribosomal imbalance after μ = 0.15 h⁻¹. P limitation caused the sharpest imbalance, with a steep decline in the 40S/60S ratio (slope = −0.92/h) and severe histidine depletion (>70%) beyond μ = 0.15 h⁻¹ (Supplementary Fig. 7). These results indicate that translational efficiency is limited not merely by ribosome quantity but by nutrient-specific subunit imbalance, coincident with energy constraints^28,29.

Fig. 2. Nutrient-dependent translational regulation and hierarchical control of ribosome homeostasis in S. cerevisiae.

a Abundance profiles of the 40S and 60S ribosomal subunits, their stoichiometric ratio (40S/60S), aminoacyl-tRNA synthetases (aaRS), and intracellular amino acid pools across specific growth rates (μ) under C, N, and P limitations (top to bottom panels). b Correlations between mRNA levels and corresponding protein abundances of aaRS genes visualized as heatmaps (Supplementary Figs. 7–9). c The relationship between intracellular ATP/EC and ribosomal protein abundance. d Diverse translational regulation under distinct nutrient limitations. Under C limitation, high μ leads to a decline in the 40S/60S ratio, ATP depletion, and low amino acid availability despite increased aaRS expression. Under N limitation, amino acid and ATP levels rise, but translation is constrained by limited aaRS abundance. Under P limitation, rapid decreases in the 40S/60S ratio, ATP, and amino acid levels jointly restrict translation. Source data are provided as a Source Data file.

Nutrient-specific regulation was also evident in aaRS expression. Under C and P limitation, aaRS genes were transcriptionally induced with increasing μ (e.g., more than twofold for five aaRS under C limitation), yet translation remained constrained by amino acid depletion or ATP shortage (up to 61% ATP reduction under P limitation; Supplementary Figs. 8 and 9). In contrast, N limitation induced a divergent aaRS expression pattern: arginine-, leucine-, and proline-specific aaRS were upregulated, whereas others were downregulated (Fig. 2b and Supplementary Figs. 10 and 11). Thus, despite sufficient amino acid pools, the limited availability of aaRS imposed a translational bottleneck under N limitation.

Our results demonstrate that the linear proportionality between ribosomal abundance and growth rate is strongly condition-dependent, and that its relationship with intracellular energy metabolism exhibits clear nutrient specificity (Fig. 2c). Plotting ribosomal abundance against intracellular ATP concentration and EC revealed distinct patterns across nutrient limitations. Under C limitation, ribosomal abundance increased linearly with growth rate while intracellular energy levels remained relatively stable, indicating sufficient energetic support for translational capacity; however, ATP and EC declined once μ exceeded 0.15 h⁻¹. Under N limitation, although intracellular ATP levels were lower than under C limitation, a positive association between cellular energy status and ribosomal abundance was still maintained. In contrast, this relationship became uncoupled under P limitation: increasing growth rates were accompanied by declining ATP and EC levels despite continued increases in ribosomal abundance. Together, these observations indicate that as C- and P-limited cells approach their maximal growth rates, increased maintenance energy demands lead to reduced intracellular energy availability. This uncoupling highlights a physiological trade-off in which cells sustain the translation flux required for rapid growth by allocating more proteomic resources to ribosome synthesis, even as intracellular energy reserves progressively decline.

Collectively, these observations reveal a hierarchical organization of translational control shaped by nutrient availability (Fig. 2d). Under C limitation, translation is primarily constrained by insufficient energy supply and is also subject to influences from subunit stoichiometric imbalance, delayed aaRS expression, and fluctuations in the amino acid availability. Under N limitation, aaRS scarcity is the dominant bottleneck, despite adequate precursor availability. Under P limitation, translation is limited predominantly by ATP depletion and aggravated by ribosomal subunit stoichiometric imbalance. Across all conditions, S. cerevisiae transcriptionally upregulated aaRS genes, yet only under C and P limitation was this response effectively translated into increased protein abundance. In contrast, N limitation decoupled transcriptional and translational regulation (Fig. 2b). Together, these findings support a hierarchical organization of translational control in which nutrient-specific constraints coordinately shape ribosome function, aaRS availability, and energy metabolism to optimize protein synthesis under diverse nutrient limitations.

Proteome allocation prioritizes anabolic functions under nutrient limitation

Dynamic proteomic profiling revealed that S. cerevisiae orchestrates metabolic homeostasis and growth rate under C, N, and P limitations through a hierarchical reallocation of proteomic resources. Central to this strategy is a preferential investment in anabolic functions, including translation machinery, nucleotide metabolism, and amino acid metabolism, while non-essential and maintenance-related processes are adaptively compressed (Fig. 3).

Fig. 3. Hierarchical proteomic resource allocation under different nutrient limitations.

Relative abundance of proteins within major functional modules under C, N, and P limitation, expressed as a percentage of total protein content. a Translation machinery. b Amino acid metabolism. c Lipid metabolism. d Mitochondrial proteins. e Nucleotide metabolism. f Energy production. g Cell wall and membrane biosynthesis. h Stress response proteins. i Molecular chaperones. j Transcription-associated proteins. Data represent mean ± SD (n = 3 independent experiments). Source data are provided as a Source Data file.

Anabolic pathways displayed nutrient-specific regulation. Proteins involved in nucleotide metabolism were lowest under C limitation at μ = 0.1 h⁻¹ (7.8% decrease), but increased steadily as growth accelerated. Under N limitation, their expression followed a biphasic trajectory, with inflection points at μ = 0.1 or 0.15 h⁻¹, while under P limitation, they showed a continuous increase. Amino acid metabolism proteins were downregulated beyond μ = 0.1 h⁻¹ under C and P limitations, but oscillated in synchrony with nucleotide metabolism under N limitation, consistent with a central role for carbon skeleton regulation during N limitation. Lipid metabolism proteins exhibited a conserved U-shaped profile across all limitations, reaching their lowest abundance at μ = 0.1 h⁻¹, before increasing by 7.7–9.9% at higher µ, indicating a compensatory mechanism to preserve membrane integrity during rapid growth.

By contrast, modules not directly supporting anabolism were systematically downregulated. Transcriptional regulatory proteins showed inflection points at μ = 0.15 h⁻¹ under C and N limitation but earlier at μ = 0.1 h⁻¹ under P limitation, suggesting that phosphorus scarcity promotes earlier transcriptional reprogramming. Molecular chaperones and stress-response proteins declined by 7.7–17.7% with increasing μ, underscoring the prioritization of growth over stress defense. Mitochondria and energy metabolism proteins exhibited distinct responses to different nutrient limitations. Under C limitation, their abundance progressively declined, reaching 22% of the baseline level at μ = 0.25 h⁻¹. In contrast, under N and P limitation, their  abundance increased slightly at growth rates above µ = 0.15 h⁻¹, by 5.8–6.4% and 3.1–5.5%, respectively, suggesting nutrient-specific compensatory regulation of respiratory capacity. Cell wall and membrane biosynthetic proteins declined after μ exceeded 0.1 h⁻¹ but partially recovered at higher growth rates (4.1% under C limitation, 7.4% under N limitation), consistent with cell-cycle-linked membrane repair responses (Fig. 3).

Collectively, these results reveal that under nutrient-limited anaerobic conditions, S. cerevisiae reallocates proteomic resources hierarchically to sustain anabolic capacity and growth. By compressing auxiliary processes, such as transcriptional regulation, stress response, and respiratory flexibility, yeast cells preserve translational and biosynthetic fluxes at the expense of maintenance energy. This dynamic rebalancing highlights proteome economy as a unifying mechanism enabling yeast to reconcile nutrient scarcity with sustained anabolic performance.

Energy-gated post-translational phosphorylation compensates for energy and translational constraints

To sustain metabolic flux under restricted energy supply and limited protein synthesis, S. cerevisiae extensively reprogrammed its phosphorylation landscape, providing a rapid and resource-efficient means for enzyme activation. This post-translational strategy constitutes a central adaptive mechanism under nutrient limitation^30–33, particularly in sustaining glycolytic capacity across different conditions (Fig. 4 and Supplementary Tables 2–4). Under P limitation, the phosphoregulatory response was the most restricted, with only five enzymes exhibiting phosphorylation-based regulation. In contrast, eight enzymes were phosphorylated under both C and N limitations. Notably, N limitation was characterized by more phosphorylation events occurring on the same enzyme, as illustrated by Hxk1, Tdh2, and Gpm1 (Fig. 4a).

Fig. 4. Phosphorylation-mediated compensation stabilizes metabolic fluxes under different nutrient limitations.

a Overview of phosphorylation regulation across enzymes in the glycolytic pathway. Changes in phosphorylation status are calculated between two adjacent dilution rates (D = 0.05 → 0.1, 0.1 → 0.15, 0.15 → 0.2, 0.2 → 0.25). Red indicates activation, whereas dark blue indicates inhibition. b Conceptual model illustrating the phosphorylation compensation mechanism. Normalized protein abundance (red), phosphorylation level (orange), and relative flux (green) are shown. c–e Quantitative validation of this mechanism in S. cerevisiae under C (c), N (d), and P (e) limitation, respectively. In each case, phosphorylation of Adh1 at specific sites (Ser2 under carbon limitation, Ser288 under nitrogen limitation, and Ser290 under phosphorus limitation) maintained metabolic flux despite reduced protein abundance. Correlation coefficients (r) between flux and protein (P-F) or phosphorylation (Phos-F) are indicated. All data represent means ± SD (n = 3 independent experiments). Source data are provided as a Source Data file.

Figure 4b summarizes the overall trends in metabolic flux, protein abundance, and protein phosphorylation. Under C limitation, a distinct metabolic inflection occurred at μ = 0.1 h⁻¹, coinciding with redistribution of flux between glycolysis and glycerol synthesis (Fig. 4c and Supplementary Figs. 12 and 13). Although the abundance of the key glycolytic enzyme Adh1 declined beyond μ = 0.15 h⁻¹, glycolytic flux remained stable. This apparent decoupling was compensated by activating phosphorylation at Ser2, which preserved catalytic efficiency despite reduced protein abundance. N limitation displayed a partially convergent yet distinct regulatory profile. Despite transcriptional downregulation of Adh1 at higher μ, phosphorylation preserved enzymatic activity, resembling the C-limited pattern (Fig. 4d and Supplementary Figs. 14 and 15). Under P limitation, phosphorylation became the predominant regulatory mechanism. The inverse relationship between Adh1 abundance and phosphorylation level indicated that activation at Ser2 and Ser290 compensated for impaired enzyme synthesis, thereby sustaining ethanol production (Fig. 4e and Supplementary Figs. 16 and 17). Similarly, phosphorylation of Gpm1 at Ser2 increased markedly beyond μ = 0.15 h⁻¹, stabilizing glycolytic flux when translation was curtailed. In contrast, enzymes such as Pgi1 and Pfk1 exhibited strong transcription–translation divergence at elevated μ (Supplementary Figs. 12–17), underscoring the nutrient-specific reliance on phosphorylation control.

These molecular patterns were closely associated with the cellular energy landscape. The higher μ_max under C limitation (0.25 h⁻¹) relative to N and P limitations (0.20 h⁻¹) reflected a less severe energetic constraint, permitting greater flexibility in translational resource allocation³⁰. Metabolomic profiling confirmed pronounced depletion of cellular energy and redox pools at high μ: ATP levels decreased by 28% and 61%, while NADH/NAD⁺ ratios dropped by 20% and 60% under N and P limitation, respectively. Although metabolic enzymes were transcriptionally upregulated under all conditions, the propagation of these regulatory responses to the protein and phosphorylation levels remained constrained by the available energy budget³⁴ (Supplementary Fig. 18).

Phosphoproteomic analyses further revealed nutrient-specific regulatory architectures. N limitation elicited the broadest phosphoregulatory response, with 17 phosphorylation events on enzymes such as Hxk1 and Eno1, indicating a system-wide effort to compensate for restricted protein synthesis. In contrast, C and P limitations invoked more selective, energy-efficient control, affecting 14 and 6 sites, respectively, on key enzymes such as Adh1 and Gpm1. These activating phosphorylations were specifically induced at high growth rates (μ ≥ 0.15 h⁻¹), coinciding with pronounced energy depletion, highlighting their role as energy-gated molecular switches.

These targeted phosphoregulatory events across diverse metabolic nodes can be interpreted within the systems-level framework of metabolic regulation proposed by Hackett et al.³⁵. According to this model, modulation of metabolic flux through global changes in enzyme abundance entails substantial bioenergetic and proteomic costs. Under conditions of severe energetic constraint, such as the pronounced decline in ATP and EC observed at high μ under P limitation, the capacity of the translational machinery becomes increasingly restricted, limiting the feasibility of abundance-based regulatory strategies. Under these conditions, cellular regulation appears to shift toward activity-based control. Phosphorylation provides a proteome-efficient regulatory mechanism in which minimal ATP investment enables rapid modulation of enzyme activity. This enables selective enhancement of catalytic efficiency in key metabolic enzymes, exemplified by increased phosphorylation of Adh1 at Ser2/Ser290 and sustained activation of Gpm1 at Ser116 (Supplementary Fig. 19), partially compensating for declining protein abundance. Collectively, these observations suggest that the transcription–translation divergence identified here reflects an energy-dependent transition toward activity-based regulation, supporting metabolic flux maintenance when large-scale protein synthesis becomes energetically constrained.

In summary, the phosphoregulatory responses observed here reveal a systems-level adaptive strategy in which metabolic control progressively shifts from enzyme abundance to enzyme activity as energetic constraints intensify. Under conditions where large-scale protein synthesis becomes energetically unfavorable, selective phosphorylation of key metabolic enzymes provides a resource-efficient mechanism for sustaining glycolytic flux and metabolic robustness. These findings highlight phosphorylation not merely as a downstream regulatory layer, but as a central component of cellular resource allocation under nutrient limitation.

Synergistic push–pull engineering overcomes coupled energy–translation limits

Our results indicate that the maximum specific growth rate of S. cerevisiae is constrained by capacity thresholds at several key metabolic nodes (e.g., substrate uptake, ATP supply, ribosome assembly). Guided by these mechanistic insights, we implemented a “push-pull” engineering strategy, in which the “push” module reinforced energy supply by enhancing glucose uptake, glycolytic flux, and ATP generation, while the “pull” module increased translational capacity through improved ribosome biogenesis and aminoacyl-tRNA charging. Together, these interventions aim to convert increased energy availability directly into biomass and product formation. To test this concept, we constructed two classes of engineered strains (Supplementary Figs. 20 and 21): Type I (energy-push) strains targeting glucose transport (HXT3, HXT4), glycolytic regulation (SNF1, HXK2), and vacuolar ATPase (VMA3); and Type II (translation-pull) strains targeting ribosome assembly (UTP20, RPS family), aminoacyl-tRNA synthetase (WRS1), and topoisomerase (TOP1).

Initial screening in nutrient-rich batch cultures confirmed that single-gene perturbations could elevate intrinsic µ_max. Overexpression of glucose transporters HXT3/HXT4 increased µ_max by 17.4% and 23.3%, respectively, and VMA3 by 18.9% (Fig. 5a). Translation-focused interventions yielded even stronger effects: UTP20 overexpression raised µ_max by 25.5%, TOP1 by 20.1%, and notably WRS1 by 40%. These effects align with the essential role of aaRS in sustaining translation throughput³⁶ and with the positive coupling between glucose uptake and q_ATP (Fig. 1c, d).

Fig. 5. Growth and metabolic verification of S. cerevisiae engineered strains under different nutrient limitations.

a Maximum specific growth rates of both wild-type (WT) and engineered strains overexpressing energy-related genes (HXT, SNF1, HXK2, VMA) or translation-related genes (UTP, RPS, tRNA synthetases). b–g Growth of selected single- and dual-gene overexpression strains under C, N, and P limitations in chemostats. h–k Comparison of glucose uptake (h), ethanol (i), CO₂ (j), and glycerol (k) production rates between engineered and WT strains under nutrient-limited conditions. All data represent means ± SD (n = 3 independent experiments). P value was calculated using a two-sided t test. *P value < 0.05, **P value < 0.01, ***P value < 0.005. Source data are provided as a Source Data file.

Under nutrient-replete conditions, combining push and pull modules did not further increase μ_max, suggesting that other proteome allocation or metabolic limits prevailed (Fig. 5a). To explore their effects under nutrient stress, we examined strain performance in chemostat cultures under C, N, and P limitation. Consistent with the mechanistic bottlenecks identified earlier, VMA3 overexpression (energy-push) was tested under C and P limitation, and WRS1 (translation-pull) under N limitation. Individually, VMA3 overexpression raised µ to 0.294 h⁻¹ (C limitation) and 0.280 h⁻¹ (P limitation) (Fig. 5b, c), while WRS1 reached 0.270 h⁻¹ under N limitation (Fig. 5d). Strikingly, the dual-overexpression strain (VMA3 + WRS1) exhibited synergistic improvements across all three limitations, reaching 0.318 h⁻¹ (C limitation) (Fig. 5e), 0.305 h⁻¹ (N limitation) (Fig. 5f), and 0.295 h⁻¹ (P limitation) (Fig. 5g), outperforming the best single-gene variants and demonstrating that combined enhancement of energy and translation capacity can significantly alleviate metabolic constraints.

To test whether the observed macroscopic growth acceleration was associated with relief of the translational bottleneck, we performed global proteomic profiling of steady-state WRS1-OE cultures. This analysis confirmed accumulation of the WRS1 target protein (Log2FC = 1.98, P value < 0.01) without detectable changes in other native aaRSs (Supplementary Fig. 22). Notably, the abundance of core components of the translational machinery remained largely unchanged (Supplementary Fig. 22), indicating that the growth enhancement is more consistent with improved tRNA-charging efficiency rather than with a generalized increase in ribosome biogenesis.

Metabolic profiling showed that engineered strains largely retained the parental glucose uptake ceiling (~14 mmol gDW⁻¹ h⁻¹) but exhibited altered carbon flux partitioning at high μ (Fig. 5h–k). Ethanol production (q_ethanol) followed parental trends at intermediate μ yet plateaued at elevated μ in engineered strains, indicating rerouting of carbon flux once growth limits were relaxed. Quantitatively, the push–pull design yielded nutrient-specific improvements: μ_max increased by 12–40% under C and P limitation and by 35% under N limitation, while ethanol-specific productivity rose by 26–55% depending on condition. Multi-omics integrative analysis revealed that translational enhancement primarily relieved N-limitation bottlenecks, whereas increased ATP supply predominantly benefited C and P limitation.

In conclusion, targeted reinforcement of either energy or translation capacity can partially relieve nutrient-imposed constraints, but only their combination produces synergistic gains that transcend the intrinsic energy–translation coupling limit. This push–pull strategy thus provides a generalizable framework for rationally improving anaerobic growth and product yield in S. cerevisiae and potentially other eukaryotic cell factories under nutrient-limited conditions.

Discussion

By integrating anaerobic chemostat physiology, multi-omic profiling, and targeted genetic perturbations, this study elucidates how S. cerevisiae reaches its maximal growth potential under anaerobic conditions and how this limit can be rationally relieved. We demonstrate that although glucose uptake and central metabolic fluxes converge toward a conserved upper bound, the nutrient-specific bottlenecks governing this bound act through coupled constraints in energy metabolism and translational capacity, which are uniquely manifested in the anaerobic system.

A central principle emerging from our analyses is convergent flux limitation with divergent upstream causes. Across C, N, and P limitations, the specific glucose uptake rate approaches a fixed upper bound (~14 mmol gDW⁻¹ h⁻¹), beyond which growth cannot increase. However, the growth rate at which this plateau appears differs: it arises earlier under N and P limitation due to rapid ATP depletion and high maintenance demands, and later under C limitation, where the cellular EC is relatively preserved. In all cases, the proximate determinant of maximal anaerobic growth rate is the availability of ATP and redox equivalents, but the mechanisms that precipitate this energetic shortfall are nutrient-specific^23,37.

For example, under P limitation, intracellular ATP levels remain low despite sustained high glycolytic flux. Previous studies have shown that inorganic phosphate (Pi) functions as a critical cofactor sustaining glycolytic throughput; insufficient Pi constrains the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) reaction, increasing the risk of decoupling ATP consumption in upper glycolysis from ATP generation in lower glycolysis³⁸. Because strictly anaerobic cells depend exclusively on substrate-level phosphorylation for ATP production, this Pi dependency imposes a major metabolic constraint. Under P-limited conditions, yeast cells therefore allocate increased proteomic resources to glycolysis to sustain energy production, consistent with the high glycolytic investment strategy proposed by Malina et al.³⁹.

This compensatory response differs fundamentally from the physiological state observed under C limitation. In agreement with established proteome allocation models³¹, C-limitation introduces a distinct systemic trade-off. Although absolute glycolytic flux quantified here using anaerobic q_ATP calculations is constrained by glucose availability, ATP synthesis efficiency remains high owing to reduced maintenance energy requirements (${\mathrm{m}}_{\mathrm{s}}=0.53$ mmol gDW⁻¹ h⁻¹). As intracellular EC is largely preserved, ATP supply for translation remains stable. Consequently, cells balance ATP production and consumption by scaling translational capacity linearly with the growth rate.

Under anaerobic conditions, the efficiency of ATP generation is intrinsically low, rendering S. cerevisiae particularly sensitive to any imbalance between energy production and consumption. Our proteome-resolved analysis shows that cells respond to increasing growth demand by reallocating proteomic resources toward anabolic and translational functions^5,26,27. Yet, the translational bottleneck diverges with nutrient context: ATP-dependent initiation limits C-limited cells, aminoacyl-tRNA synthetase (aaRS) scarcity dominates under N limitation, and P limitation imposes a combined defect in ribosome stoichiometry and EC. Consistent with the concept of translational reserves under N-limitation⁴⁰, restricted aaRS availability appears to function as a molecular braking mechanism on protein synthesis. By constraining global translation rates, this limitation maintains a substantial fraction of the pre-synthesized ribosomal pool in an idle, reserve state while preventing complete depletion of intracellular amino acid pools. Our WRS1 overexpression results further demonstrated that targeted alleviation of aaRS limitation increased the specific growth rate by more than 30%, indicating the presence of previously unexploited translational reserves.

Comparison with bacterial systems highlights both conserved principles and eukaryote-specific divergence. In E. coli, nutrient limitations modulate translation primarily through changes in elongation rate or ribosome idling²². In contrast, under strict anaerobiosis, S. cerevisiae experiences energy-linked constraints on translation initiation, aminoacyl-tRNA synthetase turnover, and ribosome assembly, reflecting its reliance on substrate-level phosphorylation and complex post-transcriptional regulation. The pronounced imbalance in ribosomal subunits under P limitation exemplifies that eukaryotic ribosome biogenesis is inherently more vulnerable to energy stress than its bacterial counterpart.

Post-translational regulation provides an additional compensatory layer⁴¹. Phosphorylation can maintain the catalytic capacity of key metabolic enzymes despite reduced protein abundance³⁷, exemplified by Adh1 phosphorylation at Ser2, Ser288, and Ser290 under C, N, and P limitation. Collectively, our phosphoproteomic analysis unveils two key principles. First, the magnitude of phosphoregulatory rewiring is nutrient-specific, with N limitation provoking the most extensive systemic response. Second, the locus of regulation is strategically targeted, with C and P limitations favoring precise activation of ATP-regenerating enzymes at high μ to directly combat energy bankruptcy. However, this compensatory mechanism depends on ATP availability and cannot fully prevent growth limitations under severe energetic stress, emphasizing the intertwined nature of energy and translational regulation.

Genetic interventions validate these mechanistic insights. Overexpression of energy-supply genes (e.g., VMA3) or translation-efficiency genes (e.g., WRS1, TOP1) improved growth rates under non-limiting conditions, but their effects under nutrient limitation were context-dependent. By contrast, a combined push–pull strategy, which simultaneously enhanced energy supply and translational efficiency, yielded synergistic improvements across all nutrient limitations, elevating growth and ethanol production rates beyond the effects of single-gene modifications. This demonstrates that anaerobic yeast growth rates under different nutrient limitations are not fixed but can be enhanced by alleviating coupled energy–translation limits.

Collectively, our results reveal that nutrient-specific limits in energy metabolism and translation govern anaerobic yeast growth, and that targeted interventions in these processes provide a generalizable strategy to surpass natural physiological constraints. By pinpointing causal regulatory nodes and experimentally validating their manipulation, this study establishes a solid framework for rational strain engineering, with direct implications for bioproduction processes operating under nutrient-limited conditions. In essence, the maximal growth rate of S. cerevisiae is a tunable trait, controllable through coordinated modulation of energy supply and translational efficiency.

Methods

Strain and cultivation conditions

The wild-type Saccharomyces cerevisiae strain CEN.PK113-7D (MATα, MAL2-8c, SUC2) was used throughout this study. Stock cultures were maintained at −80 °C in 40% (v/v) glycerol. Anaerobic continuous cultivations were carried out in a 5-L bioreactor (Shanghai Guoqiang Bioengineering Equipment Co., Ltd., Shanghai, China) with a working volume of 3 L. Carbon-, nitrogen-, and phosphorus-limited chemostat cultures were established to achieve steady states at specific growth rates of 0.05, 0.10, 0.15, 0.20, and 0.25 h⁻¹. The temperature was maintained at 30 °C, with agitation at 150 rpm and nitrogen gas sparging at 1.5 mL min⁻¹ to ensure strict anaerobiosis.

Culture media

Carbon-limited medium: 5 g/kg (NH₄)₂SO₄, 3 g/kg KH₂PO₄, 0.5 g/kg MgSO₄·7H₂O, 27.5 g/kg glucose·H₂O.

Nitrogen-limited medium: 1 g/kg (NH₄)₂SO₄, 3 g/kg KH₂PO₄, 5.3 g/kg K₂SO₄, 0.5 g/kg MgSO₄·7H₂O, 61.6 g/kg glucose·H₂O.

Phosphorus-limited medium: 5 g/kg (NH₄)₂SO₄, 0.12 g/kg KH₂PO₄, 1.9 g/kg K₂SO₄, 0.51 g/kg MgSO₄·7H₂O, 72.6 g/kg glucose·H₂O.

All media required the addition of 1 mL of trace metals solution, 1 mL of vitamins, 10 mg/L of ergosterol, 5 ‰ defoamer, and 0.4 g/kg of Tween-80 solution. The trace metal components used included: 3 g/kg FeSO₄·7H₂O, 4.5 g/kg ZnSO₄·7H₂O, 4.5 g/kg CaCl₂·2H₂O, 1 g/kg MnCl₂·4H₂O, 0.3 g/kg CoCl₂·6H₂O, 0.3 g/kg CuSO₄·5H₂O, 0.4 g/kg Na₂MoO₄·2H₂O, 1 g/kg H₃BO₃, 0.1 g/kg KI, and 19 g/kg EDTA. The vitamin mix contained: 0.05 g/kg D-biotin, 1 g/kg D-pantothenic acid, 1 g/kg hemicalcium salt, 1 g/kg thiamine-HCl, 1 g/kg pyridoxine-HCl, 1 g/kg nicotinic acid, 0.2 g/kg 4-aminobenzoic acid, and 25 g/kg myo-inositol.

Sampling

Samples were collected once steady state was achieved in chemostat cultures, confirmed after at least five residence times. Intracellular and extracellular metabolites were sampled between the fifth and tenth residence volumes. For omics analyses, cells were harvested at the sixth residence time for transcriptomic, proteomic, phosphoproteomic, and metabolomic profiling. Mass spectrometry–based analyses were performed under carbon-, nitrogen-, and phosphorus-limited conditions at dilution rates of 0.05, 0.10, 0.15, 0.20, and 0.25 h⁻¹ (for carbon limitation). All experiments were conducted in biological triplicate for each condition.

Determination of glucose, glycerol, and ethanol

Glucose, glycerol, and ethanol concentrations were determined by high-performance liquid chromatography (HPLC; Agilent 1100, USA) equipped with a 2410 refractive index detector. Metabolites were separated on an Aminex HPX-87H column (Bio-Rad, USA) maintained at 50 °C, using 5 mM H₂SO₄ as the mobile phase at a flow rate of 0.6 mL min⁻¹. The injection volume was 15 µL, and samples were kept at 15 °C throughout analysis.

Omics data acquisition and processing

Transcriptomics

Total RNA was extracted using the TRIzol reagent kit (Invitrogen, USA). RNA purity and concentration were assessed with a NanoDrop 2000 spectrophotometer, and integrity was verified using an Agilent 2100 Bioanalyzer. Only samples meeting the quality criteria were used for library construction, performed according to the VAHTS Universal V5 RNA-seq Library Prep Kit protocol (Vazyme, China). Sequencing was conducted on an Illumina NovaSeq 6000 platform, generating 150 bp paired-end reads. Raw reads in FASTQ format were quality-filtered using Fastp, and clean reads were aligned to the S. cerevisiae reference genome using HISAT2. Gene expression levels were quantified as FPKM values, and read counts were obtained with HTSeq-count. Principal component analysis (PCA) was conducted in R (v3.2.0) to evaluate sample reproducibility.

Proteomics

Cell pellets were ground in liquid nitrogen, and total proteins were extracted using Tris-phenol buffer supplemented with phosphatase and protease inhibitors (final concentration 1 mM). Proteins were precipitated with pre-cooled 0.1 M ammonium acetate–methanol, washed with methanol and acetone, and resolubilized in lysis buffer. Protein concentration was determined using the BCA assay, and quality was verified by SDS-PAGE.

Equal amounts (50 μg) of protein from each sample were reduced with 5 mM dithiothreitol at 55 °C for 30 min and alkylated with 20 mM iodoacetamide in the dark for 15 min. After acetone precipitation, proteins were resuspended in 300 mM triethylammonium bicarbonate (TEAB) and digested overnight at 37  °C with sequencing-grade trypsin (enzyme-to-protein ratio, 1:50, w/w). Peptides were labeled using TMT 10-plex reagents (Thermo Fisher Scientific) according to the manufacturer’s instructions. Briefly, the lyophilized peptides were resuspended in 100 μL 200 mM TEAB, and 41 μL of the TMT label reagent was added to each sample. The mixture was incubated at room temperature for 1 h, and the reaction was quenched with 5% hydroxylamine.

Labeled peptides were fractionated by high-pH reverse-phase liquid chromatography on an Agilent 1100 system using a Zorbax Extend-C18 column. The separation used a gradient of mobile phase A (2% acetonitrile in water) and B (90% acetonitrile in water) as follows: 0–8 min, 98% A; 8–48 min, 95%–75% A; 48–60 min, 75%–60% A. Fractions collected between 8 and 60 min were pooled and analyzed by LC-MS/MS on a Q Exactive HF mass spectrometer (Thermo Fisher Scientific). The LC gradient was 90 min with MS1 resolution set at 120,000 and MS2 resolution at 30,000. MS data were processed using Proteome Discoverer v2.4.1.15, and proteins were identified and quantified against the S. cerevisiae reference proteome with the following parameters: trypsin as the enzyme, carbamidomethylation of cysteine as a fixed modification, TMT 10-plex labeling on N-terminus and lysine as a quantitative method, oxidation of methionine, and acetylation of the protein N-terminus as variable modifications. The false discovery rate (FDR) for peptide and protein identification was set to 1%.

Phosphoproteomics

Phosphoproteomic analysis followed the same initial protein extraction and digestion procedures described for global proteomics, with additional desalting and phosphopeptide enrichment steps. After tryptic digestion, peptides were desalted using C18 cartridges. Phosphopeptides were enriched using titanium dioxide (TiO₂) beads. Briefly, peptides were incubated with TiO₂ beads in loading buffer, followed by sequential washing with buffers containing 0.5% TFA/50% ACN and 0.1% TFA/50% ACN. Enriched phosphopeptides were then eluted with 10% NH₃·H₂O, dried under vacuum, and reconstituted in 0.1% formic acid for LC-MS/MS analysis.

LC-MS/MS analysis was performed on a Q-Exactive HF mass spectrometer (Thermo Fisher Scientific) coupled to an EASY-nLC 1200 system (Thermo Fisher Scientific). Phosphopeptides were separated on a C18 column (25 cm × 75 μm) at a flow rate of 300 nL/min using a 240-min linear gradient of 3%-42% mobile phase B (0.1% formic acid in 80% ACN). Full MS scans were acquired at a resolution of 120,000, followed by data-dependent MS/MS scans of the top 15 most intense precursor ions with a resolution of 30,000.

Raw MS data were processed using MaxQuant (v1.6.17.0) with Andromeda for database searching against the S. cerevisiae reference proteome (UniProt). Carbamidomethylation of cysteine was set as a fixed modification, while oxidation (M), acetyl (Protein N-terminus), and phosphorylation (Ser, Thr, Tyr) were treated as variable modifications. The false discovery rate (FDR) was controlled below 1% at both peptide and site levels. Phosphosite quantification was performed based on the MS/MS spectra of enriched peptides.

Non-targeted metabolomics

Cell pellets were extracted with pre-chilled 80% methanol (v/v) and disrupted by ultrasonic treatment in an ice-water bath. The extracts were incubated at −40 °C to precipitate proteins and centrifuged at 12,000×g for 10 min at 4 °C. The resulting supernatant was divided for liquid chromatography–mass spectrometry (LC-MS) and gas chromatography–mass spectrometry (GC-MS) analyses and lyophilized.

For LC-MS analysis, dried extracts were reconstituted in 20% methanol, sonicated briefly, and centrifuged again. The supernatants were filtered through 0.22 μm membranes and analyzed using a UHPLC-Q Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific) operating in both positive and negative ion modes. Separation was performed on an ACQUITY UPLC HSS T3 column (1.8 μm, 2.1 × 100 mm) using a water/acetonitrile (both containing 0.1% formic acid) gradient.

For GC-MS analysis, samples were first derivatized with 15 mg/mL methoxyamine hydrochloride in pyridine at 37 °C for 90 min, followed by treatment with BSTFA (containing 1% TMCS) at 70 °C for 60 min. After cooling to room temperature, n-hexane was added to the mixture. Derivatized samples were analyzed on an Agilent 7890B GC system coupled to a 5977B MSD system. Helium was used as the carrier gas on a DB-5MS capillary column (30 m × 0.25 mm × 0.25 μm) with a specific temperature gradient.

LC-MS data acquisition and peak alignment were performed using Progenesis QI (Waters Corporation) for peak alignment and normalization. GC-MS data were processed using MS-DIAL for peak detection, deconvolution, and alignment. Metabolites were identified based on accurate mass, retention time, and MS/MS spectral matching against public databases (HMDB, METLIN, KEGG) and in-house libraries.

Targeted metabolomics

Intracellular metabolites were quantified using a ¹³C-labeled internal standard strategy to ensure absolute quantification accuracy. Cell pellets collected from steady-state chemostats were quenched in −40 °C 60% methanol and extracted by repeated freeze–thaw cycles. The extracts were centrifuged at 12,000×g for 10 min at 4 °C, and the supernatants were filtered through 0.22 μm membranes before analysis.

Metabolite quantification was performed on a Thermo Ultimate 3000 UHPLC system coupled to a TSQ QUANTUM ULTRA triple quadrupole mass spectrometer (Thermo Fisher Scientific) equipped with an electrospray ionization (ESI) source. Separation was achieved on an ACQUITY UPLC BEH C18 column (1.7 μm, 2.1 × 150 mm) preceded by a VanGuard pre-column (1.7 μm, 2.1 × 5 mm), with the column temperature maintained at 25 °C. The mobile phases consisted of (A) 5% (v/v) acetonitrile in water and (B) 84% (v/v) acetonitrile in water, both supplemented with 5 mM dibutylammonium acetate (DBAA) as an ion-pairing reagent, and a gradient elution program was used. The multiple reaction monitoring (MRM) mode was used for detection.

The ¹³C-labeled metabolites that were used as internal standards in this study were produced by growing yeast cells in a fed-batch cultivation on a feed medium containing exclusively uniformly (U)-¹³C-labeled glucose as the carbon source. The obtained ¹³C-labeled cell extract contains in principle the whole U-¹³C-labeled metabolome of the yeast cell. Such ¹³C-labeled internal standards were added to each sample at defined concentrations before extraction, enabling correction for matrix effects and ionization variability. Absolute metabolite concentrations were determined using calibration curves generated from authentic standards analyzed under identical conditions.

This targeted metabolomic approach enabled precise quantification of key intermediates in glycolysis, amino acid metabolism, and nucleotide synthesis, allowing the determination of absolute intracellular metabolite pools across different dilution rates under carbon-, nitrogen-, and phosphorus-limited conditions.

Strain engineering

Gene overexpression in S. cerevisiae was achieved through endogenous homologous recombination. Target gene fragments and their corresponding homology arms, designed based on non-coding regions of the yeast genome, were individually amplified by PCR using an annealing temperature of 54–55 °C. PCR products were verified by agarose gel electrophoresis, purified, and quantified. The target gene was then fused with its flanking homology arms via overlap extension PCR under optimized reaction conditions (15 amplification cycles). The resulting integration cassettes were purified for yeast transformation.

Competent cells were prepared by cultivating activated yeast in liquid YPD medium at 30 °C and 220 rpm until reaching an OD₆₀₀ of 0.8–1.0. Cells were washed with pre-chilled sterile water and resuspended in 0.1 M lithium acetate (LiAc). For transformation, 1–2 µg of purified DNA, denatured salmon sperm carrier DNA, LiAc, and PEG 3350 were added to 50 µL aliquots of competent cells. After gentle mixing, the mixture was incubated at 30 °C for 30 min, heat-shocked at 42 °C for 25 min, and briefly centrifuged. The resulting pellet was resuspended in YPD medium and recovered for 2 h before plating on selective YPD agar supplemented with G418. Plates were incubated at 30 °C for 2–3 days until colonies appeared.

Transformants were screened by colony PCR using primers targeting the integration flanks. PCR products were verified by gel electrophoresis and Sanger sequencing. Successfully integrated strains were confirmed by correct amplicon size and sequence. Verified colonies were cultured overnight in liquid YPD containing G418, streaked for isolation, and stored in 50% glycerol at −80 °C. Primer information is provided in Supplementary Data 1.

Specific energy production rate

Under steady-state chemostat conditions, the specific growth rate $\left(\mu \right)$ strictly equals the dilution rate (D). The specific macroscopic consumption rates of substrates ($q_s$, e.g., glucose) and production rates of extracellular metabolites ($q_p$, e.g., ethanol) were directly calculated using respective mass balances across the bioreactor:

$${\mathrm{q}}_{\mathrm{s}}=\frac{\mathrm{D}\times \left({\mathrm{C}}_{\mathrm{s},\mathrm{in}}-{\mathrm{C}}_{\mathrm{s},\mathrm{out}}\right)}{{\mathrm{C}}_{\mathrm{x}}}$$ 1

$${\mathrm{q}}_{\mathrm{p}}=\frac{\mathrm{D}\times {\mathrm{C}}_{\mathrm{p},\mathrm{out}}}{{\mathrm{C}}_{\mathrm{x}}}$$ 2

where ${\mathrm{C}}_{\mathrm{s},\mathrm{in}}$ and ${\mathrm{C}}_{\mathrm{s},\mathrm{out}}$ are the substrate concentrations in the feed and effluent (mmol L⁻¹), respectively, ${\mathrm{C}}_{\mathrm{p},\mathrm{out}}$ is the product concentration in the effluent (mmol L⁻¹), and ${\mathrm{C}}_{\mathrm{x}}$ is the steady-state biomass concentration (gDW L⁻¹).

The relationship between specific substrate consumption rate and specific growth rate was described using the Herbert–Pirt equation (Eq. (3))^42,43:

$${\mathrm{q}}_{\mathrm{s}}=\frac{1}{{\mathrm{Y}}_{\mathrm{X}/\mathrm{S}}}*\mathrm{\mu }+{\mathrm{m}}_{\mathrm{s}}$$ 3

where ${\mathrm{q}}_{\mathrm{s}}$ represents the specific substrate uptake rate (mmol gDW⁻¹ h⁻¹), ${\mathrm{Y}}_{\mathrm{X}/\mathrm{S}}$ is the biomass yield on substrate (gDW mmol⁻¹), μ is the specific growth rate (h⁻¹), and ${\mathrm{m}}_{\mathrm{s}}$ denotes the maintenance coefficient (mmol gDW⁻¹ h⁻¹).

Linear regression based on Eq. (3) revealed that the maintenance energy requirement under P limitation (3.10 mmol gDW⁻¹ h⁻¹) was 5.8-fold higher than under C limitation (0.53 mmol gDW⁻¹ h⁻¹; ANOVA, P value < 0.001). N limitation exhibited an intermediate value (2.01 mmol gDW⁻¹ h⁻¹; Table 1).

Under strictly anaerobic conditions, the mitochondrial respiratory chain is completely inactive, meaning that ATP is generated exclusively via substrate-level phosphorylation during glycolysis. Based on the established stoichiometry for anaerobic S. cerevisiae⁴⁴, the net ATP production does not rely on variable respiratory P/O ratios, but is rigidly tied to the fermentation fluxes: the formation of 1 mole of ethanol yields 1 mole of ATP, whereas the formation of 1 mole of glycerol consumes 1 mole of ATP (as it serves as a redox sink to reoxidize biomass-derived NADH, bypassing the ATP-yielding steps of lower glycolysis).

$${\mathrm{Y}}_{\frac{\mathrm{ATP}}{\mathrm{biomass}}}=\frac{{\mathrm{C}}_{\mathrm{biomass}}}{{\mathrm{C}}_{\mathrm{ethanol}}-{\mathrm{C}}_{\mathrm{glycerol}}}$$ 4

The ATP yield (${\mathrm{Y}}_{\frac{\mathrm{ATP}}{\mathrm{biomass}}}$) was determined according to Eq. (4)⁴⁴, where ${\mathrm{C}}_{\mathrm{biomass}}$ (g L⁻¹), ${\mathrm{C}}_{\mathrm{ethanol}}$ (mM), and ${\mathrm{C}}_{\mathrm{glycerol}}$ (mM) represent the concentrations of biomass, ethanol, and glycerol, respectively.

The specific ATP production rate (${\mathrm{q}}_{\mathrm{ATP}}$) was determined using Eq. (5)²⁴:

$${\mathrm{q}}_{\mathrm{ATP}}=\frac{\mathrm{\mu }}{{\mathrm{Y}}_{\frac{\mathrm{ATP}}{\mathrm{biomass}}}}$$ 5

By combining Eqs. (4) and (5), which mathematically simplifies to the exact stoichiometric balance ${\mathrm{q}}_{\mathrm{ATP}}={\mathrm{q}}_{\mathrm{ethanol}}-{\mathrm{q}}_{\mathrm{glycerol}}$, we determined the specific ATP production rate under each nutrient limitation condition. This absolute dependence on glycolytic stoichiometry avoids any uncertainties regarding respiratory ATP contributions, enabling a robust and quantitative assessment of energy allocation across growth regimes.

Statistics and reproducibility

All physiological cultivations and multi-omics experiments were performed with at least three independent biological replicates (n ≥ 3). No statistical method was used to predetermine sample size; rather, sample sizes were chosen based on standard practices in the field of microbial physiology to ensure sufficient statistics for determining significance. No data were excluded from the analyses. Because our study relies on defined microbial continuous cultivations and multi-omics profiling rather than clinical or animal trials, the experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment.

All statistical analyses were conducted using GraphPad Prism 9.0 or R software version 4.1.2. Data are presented as mean ± standard deviation (SD). Statistical significance between two groups was determined using the two-tailed unpaired Student’s t test. A P value of <0.05 was considered statistically significant.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Y.W. and Y.H. designed and performed the experiments, analyzed the data, and drafted the manuscript. Y.L. and L.C. performed computational analyses and generated data visualizations. X.Y. and Z.S. contributed to sample preparation. Y.C., F.L., S.Q. and Y.W. participated in manuscript revision. Y.Z. supervised the research and provided critical comments. G.W. conceived and coordinated the study, secured funding, and finalized the manuscript. All authors contributed to the manuscript.

Peer review

Peer review information

Nature Communications thanks Jens Nielsen, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

G.W. discloses financial support for the research of this work from the National Key R&D Program of China (grant number 2024YFA0920300), National Natural Science Foundation of China (grant number 22578122), Natural Science Foundation of Shanghai (grant number 25ZR1402110), the Taishan Scholars Program of Shandong Province (grant number tspn202408281), The Explorers Program of Shanghai (Basic Research Funding) (grant number 25TS1401100) and the Jiangsu Provincial Major Science and Technology Special Project (grant number BG2025044). All the other coauthors declare no relevant funding.

Data availability

The RNA-seq raw data supporting this study have been deposited in the NCBI Sequence Read Archive (SRA) under accession code PRJNA1347047. The proteome and phosphoproteome raw data are available in the ProteomeXchange under code PXD078963. The metabolome raw data are available in the Metabolights under [https://www.ebi.ac.uk/metabolights/MTBLS14573]. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74197-4.