📄 systems biology

Competing constraints on protein availability and nutrient uptake reshape yeast genetic interactions

This study demonstrates that varying protein availability and nutrient uptake constraints in enzyme-constrained metabolic models reveals that protein-nutrient co-limitation significantly reshapes yeast genetic interactions, causing frequent sign inversions and revealing protein-priced epistasis that standard models miss.

Original authors: Almaas, E., Kumelj, T.

Published 2026-09-14

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Almaas, E., Kumelj, T.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Inside every living cell, a constant, quiet negotiation takes place between what the organism needs to eat and what it can build with the materials it has. Cells are not just bags of chemicals; they are factories where proteins act as the workers, enzymes that speed up the chemical reactions necessary for life. But a cell has a limited budget of protein it can make. It cannot build an infinite number of workers, so it must decide how to spend its protein resources. If it spends too much on one task, it must take it away from another. This trade-off is the central constraint of cellular life. Scientists have long used computer models to predict how these factories work, but for a long time, these models assumed that the amount of protein available to the cell was a fixed number, like a constant salary that never changed. They also assumed that the only limit on growth was how much food the cell could grab from its environment. This paper asks a simple but profound question: what happens to the cell's internal logic when we let the protein budget change, and when we force the cell to deal with both a limited food supply and a limited protein supply at the same time?

The researchers focused on the baker's yeast, a single-celled fungus that is a standard model for understanding how cells work. They built a massive, detailed computer simulation of the yeast's metabolism, a map of thousands of chemical reactions and the specific proteins that drive them. In this model, they did not just look at one scenario. Instead, they systematically deleted genes, one by one and then in pairs, to see how the yeast would cope. In a normal experiment, if you remove a gene, the cell might grow slower. If you remove two genes, it might grow even slower, or perhaps the two missing parts cancel each other out and the cell grows fine. This relationship between two missing parts is called a genetic interaction. The team ran their simulation thousands of times, sweeping the total amount of available protein from very low levels to very high levels, and they did this both when the yeast had unlimited food and when its food intake was strictly capped.

What they found was that the amount of protein available to the cell changes the rules of the game, but only under specific conditions. When the cell has plenty of food and the only limit is how much protein it can make, changing the protein budget is like turning up the volume on a radio. Everything gets louder—both the healthy cells and the sick ones grow faster—but the relationship between them stays exactly the same. The sick cells remain sick in the same way, and the genetic interactions do not change. However, the moment the cell faces a second limit, such as a cap on how much sugar it can eat, the situation changes completely. In this state of co-limitation, where the cell is squeezed by both a lack of food and a lack of protein workers, the genetic interactions reshape themselves. The researchers discovered that as they varied the protein availability, about one out of every three genetic interactions appeared, disappeared, or even flipped from being harmful to helpful.

This flipping of the rules happens because the cell is forced to switch its entire strategy. When protein is scarce and food is limited, the yeast relies on fermentation, a quick but inefficient way to make energy that requires fewer protein workers. When protein becomes more available, the yeast can afford to switch to respiration, a slower but much more efficient process that requires a heavy investment in protein machinery. The researchers found that a specific type of genetic interaction, where two missing genes make the cell much worse off than expected, would suddenly turn into a situation where the two missing genes actually help each other out. This happens when a mutation that forces the cell to use a protein-heavy backup route meets a second mutation that removes a protein-heavy pathway. If the cell is in a protein-rich environment, the first mutation is no longer a burden because the protein is cheap. But if the cell is protein-poor, that same mutation is a disaster. The computer model showed that this switch in behavior is not random; it is a direct consequence of the cell trying to balance its protein budget against its food intake.

The study also revealed a hidden layer of connections between genes that standard models miss. In traditional models, two genes only interact if they are part of the same chemical pathway. But in this protein-constrained world, genes that are far apart in the chemical map can interact simply because they are both competing for the same limited pool of protein workers. The researchers identified pairs of genes that act as backups for each other, known as isozymes. In a standard model, deleting one of these backups does nothing because the other one takes over for free. But in this simulation, the backup costs protein to build. So, if you delete the main gene, the cell is forced to pay the protein cost for the backup, making the cell weaker. If you delete both, the cell has no choice but to find a different, more expensive route, or it simply cannot survive. This creates a new kind of weakness that only appears when the cell is struggling to manage its protein budget.

The researchers were careful to note that these findings come from a computer simulation, not a wet-lab experiment where yeast was grown in a dish. The model was calibrated to match real-world data from yeast grown in controlled environments, and the results held up when they tested the model against known biological facts, such as how yeast switches from fermentation to respiration. However, the specific numbers and the exact list of interacting genes are predictions based on the current best data about how fast yeast enzymes work. The authors suggest that these predictions can be tested in the lab by growing yeast in conditions where the food supply is carefully controlled, allowing scientists to see if the genetic interactions really do change as the protein budget shifts. The work provides a new way to think about how cells organize themselves, showing that the rules of genetic interaction are not fixed laws of nature, but flexible strategies that depend on the resources the cell has at hand. It suggests that to truly understand how a cell works, we must look at it not just as a machine with fixed parts, but as a dynamic system constantly negotiating its survival between what it eats and what it can build.

Technical Summary: Competing Constraints on Protein Availability and Nutrient Uptake Reshape Yeast Genetic Interactions

Problem Statement
Enzyme-constrained metabolic models (e.g., GECKO, MOMENT) improve upon standard Flux Balance Analysis (FBA) by coupling reaction fluxes to enzyme abundance and turnover numbers, thereby reproducing phenomena like overflow metabolism. However, these models are typically calibrated at a single, fixed value of protein availability ( $P_{tot}$ ), despite evidence that the metabolic protein pool in Saccharomyces cerevisiae varies two- to three-fold across physiological growth rates. The central question addressed is how genetic interactions (epistasis) are affected when protein availability is treated as a variable parameter rather than a fixed constant, and specifically, under what conditions this variation alters the sign or magnitude of interactions.

Methodology
The authors utilized an enzyme-constrained genome-scale model of S. cerevisiae (ecYeastGEM v8.6.2), modified to close three non-physiological secretion routes (acetaldehyde, (R,R)-2,3-butanediol, and diphosphate) to ensure physiological relevance in chemostat-like conditions. This adjusted model, termed ecYeastGEM-A, was used to compute the fitness of every single and pairwise gene deletion (approximately $4.8 \times 10^5$ linear programs per $P_{tot}$ value) across a wide sweep of protein availability ( $P_{tot} = 0.04$ to $0.88 $g gDW$ ^{-1}$).

The study systematically varied two constraints:

Protein Pool ( $P_{tot}$ ): The total mass of catalytic protein available.
Nutrient Uptake: Specifically, glucose ( $\Gamma$ ) and oxygen ( $\Omega$ ) uptake bounds.

The authors analyzed the system under three constraint configurations:

Pool-only: Protein is the sole limiting factor (the standard calibration condition).
Nutrient-only: Standard FBA conditions.
Co-limitation: Both the protein pool and a nutrient uptake bound are active simultaneously.

Genetic interactions ( $\epsilon_{ij}$ ) were defined as the deviation of double-deletion fitness from the multiplicative expectation of single deletions. The study examined how $\epsilon_{ij}$ changes as $P_{tot}$ varies, identifying "sign inversions" (changes from positive to negative epistasis or vice versa) and "protein-priced" interactions (interactions arising solely from competition for the protein pool).

Key Results

Invariance under Single Constraints:
When the protein pool is the only binding constraint (pool-only), relative fitness and genetic interactions are largely invariant to changes in $P_{tot}$ . In this regime, $P_{tot}$ acts merely as a scale factor for growth rates. Consequently, no genetic interactions change sign, and the interaction landscape remains static. This confirms that standard calibration conditions (where nutrient uptake is unconstrained) mask the dynamic nature of protein allocation.
Co-limitation Drives Reorganization:
Significant changes in genetic interactions occur only under co-limitation, where the protein pool and a nutrient uptake bound (e.g., glucose) limit growth simultaneously. In this regime, varying $P_{tot}$ alters the relative fitness of mutants and reshapes the interaction network. Approximately one-third of genetic interactions are gained or lost across the physiological range of protein availability when co-limitation is present.
Mechanisms of Interaction Change:
- Recruitment of Respiratory Machinery: Most changes in the interaction landscape are inherited from single mutants of the respiratory machinery. As protein becomes available, the wild-type cell recruits respiratory enzymes. Deletions in respiratory complexes become neutral when fermentation is dominant but harmful once respiration is recruited. Conversely, deletions in lower glycolysis (which require expensive bypasses) become less harmful as protein availability increases.
- Sign Inversion: A specific class of interactions reverses sign where the metabolic optimum switches from fermentation to respiration. The paper highlights the interaction between GPM1 (a glycolytic gene with a protein-expensive bypass) and respiratory chain subunits. In low protein (fermentation), the double deletion is aggravating (negative epistasis) because the mutant must respire to survive the glycolytic lesion, but respiration is blocked. In high protein (respiration), the interaction becomes alleviating (positive epistasis) because the respiratory deletion removes the competitor for the protein pool, making the glycolytic bypass "free."
- Protein-Priced Interactions: The study identifies interactions between genes with no direct stoichiometric coupling (e.g., isozyme pairs or genes in different pathways). These arise because the network reroutes flux around deletions, and the protein pool assigns a "price" (enzyme cost) to these detours. If two deletions compete for the same limited protein pool to fund their respective bypasses, they exhibit negative epistasis. If one deletion removes the route that the other's bypass competes with, they exhibit positive epistasis.
Strict Pairs and Isozymes:
The study identifies "strict" pairs where single mutants have no fitness defect (due to isozymes or alternative routes) but the double mutant is sick. These interactions are predicted to be negative and are absent in standard FBA because FBA assigns no cost to enzyme usage. These interactions are "protein-priced," existing only when the protein pool is constrained.

Significance and Claims
The paper claims that the assumption of fixed protein availability in current metabolic models is a significant limitation for predicting genetic interactions. The authors demonstrate that:

Nutrient supply is a direct experimental proxy: Because relative fitness depends on the ratio of nutrient supply to protein availability, manipulating nutrient supply (e.g., glucose concentration in a chemostat) allows for the experimental testing of condition-dependent epistasis without directly measuring protein pools.
Context-dependence is structural: The study predicts that genetic interaction networks are not static but are dynamically rewired by the interplay between protein availability and nutrient constraints.
Predictions for experimental validation: The authors propose four specific, testable predictions:
1. Interactions involving respiratory genes should be condition-dependent, appearing only above a critical dilution rate ( $D_{crit}$ ) where co-limitation occurs.
2. Sign inversions should be observable between fermentative and respiratory growth regimes for specific glycolytic-respiratory pairs.
3. Synthetic sickness in isozyme pairs (predicted by the model) should disappear under glucose limitation (where the pool is slack) and appear under respirofermentative conditions.
4. Interactions between the respiratory chain and cofactor-consuming biosynthetic branches (e.g., serine, proline, oxidative pentose phosphate) should emerge when protein availability is tightened.

The paper concludes that while enzyme-constrained models are powerful, their predictive power for genetic interactions is maximized only when the model accounts for the dynamic competition between protein availability and nutrient uptake, a condition that reflects the physiological reality of growing yeast.

Technical Summary: Competing Constraints on Protein Availability and Nutrient Uptake Reshape Yeast Genetic Interactions

More like this