Cells Balance Speed and Control When Budgeting Their Metabolic Enzymes

Every living cell runs a tight economic operation. Proteins are among the most expensive assets a cell can build, with protein synthesis accounting for roughly twenty percent of basal metabolic rate, so evolution has relentlessly pressured cells to squeeze the most work out of every enzyme molecule they produce. A new study published in Molecular Systems Biology by Kai Sun, Weiyan Zheng, Ziwei Dai and colleagues at the Southern University of Science and Technology now reveals a surprisingly simple mathematical rule that governs how cells distribute their enzyme budget along a metabolic pathway, and an equally elegant explanation for why real biology deliberately breaks that rule.

The team began with a deceptively basic question: if a cell has a fixed total amount of enzyme to spend on a linear pathway, how should it divide that investment among the individual reaction steps to achieve the highest possible flux? Starting from a chain of first-order reactions, and then extending the analysis to zero-order and full reversible Michaelis–Menten kinetics, the researchers used Lagrange multiplier methods to derive an explicit optimality condition. At maximal flux efficiency, the equilibrium constant of each reaction must equal what they call the catalytic-abundance quotient, or CAQ, of the adjacent reaction pair. In other words, thermodynamics and enzyme economics become locked together: reactions with larger equilibrium constants, which are less likely to stall near equilibrium, should receive proportionally greater enzyme investment.

What makes this K-CAQ relationship remarkable is its simplicity. Earlier theoretical work on metabolic control analysis, dating back to the 1970s, established that flux control coefficients should be proportional to enzyme abundances at optimal efficiency, but those classical results were formulated long before high-throughput omics data existed and were difficult to test directly. For realistic Michaelis–Menten kinetics, previous optimization problems became highly nonlinear with no analytical solution. The new derivation cuts through that complexity, producing a closed-form relationship that can be checked against real data, and it survives numerical simulation of ten-reaction pathways across thousands of randomly sampled parameter sets, where log-transformed K and CAQ values showed near-perfect agreement at the optimum regardless of enzyme saturation levels.

To test the theory against biology, the researchers assembled an unusually rich dataset spanning three evolutionarily distant organisms: human, budding yeast, and the bacterium Escherichia coli. They drew on quantitative proteomics from 378 human cancer cell lines in the Cancer Cell Line Encyclopedia, transcriptomics from more than 5,000 tumor and normal tissue samples in The Cancer Genome Atlas, and absolute protein abundance datasets for yeast and E. coli. Equilibrium constants were computed from standard Gibbs free energy changes and adjusted for intracellular cofactor concentrations such as ATP, ADP, and the NAD redox pairs, while catalytic efficiencies were estimated from kcat and Km values predicted by deep learning tools together with measured metabolite concentrations. Isozymes and cofactor dependencies, two complications that usually derail such analyses, were handled with generalized formulations of the K-CAQ equation.

The first encouraging result was that enzyme allocation itself is strikingly conserved. Across human cancer cell lines from wildly different tissues of origin, the relative abundance of metabolic enzymes showed median pairwise Spearman correlations exceeding 0.8 at both protein and transcript levels, and glycolytic enzyme proportions were similarly preserved. When the team compared the measured CAQ values to equilibrium constants for glycolysis in all three species, they found a moderate but consistent coupling, quantified by a new metric they call the Efficiency Optimality Index, the root mean squared error between log K and log CAQ. Median EOI values hovered around 2.5 in human samples, with similar patterns in yeast and bacteria. Flux efficiency maximization, it appeared, is one real objective shaping pathway design, but not the only one.

The deviations from the ideal relationship turned out to be anything but random. The largest mismatches repeatedly occurred at the same reaction pairs across species: the steps involving glucose transport, hexokinase, and phosphofructokinase in human and yeast glycolysis, and the glucose phosphotransferase step in E. coli. These are precisely the reactions long recognized as the master control points of glycolytic flux. At these steps, the measured CAQ was consistently smaller than K, meaning the enzymes were expressed at lower levels than pure efficiency maximization would predict. The researchers realized that this underinvestment is not wasteful at all. In metabolic control analysis, reducing the abundance of an enzyme increases its flux control coefficient, its leverage over the total pathway flux. By starving key control enzymes of resources, the cell concentrates regulation into a handful of sensitive, tunable steps.

This insight crystallized into a formal trade-off between two competing objectives: flux efficiency, the flux achieved per unit of total enzyme, and control efficiency, defined as the standard deviation of flux control coefficients across the pathway. Using an epsilon-constraint algorithm, the team computed the Pareto front that balances these two goals, tracing the set of enzyme allocation profiles where neither objective can be improved without sacrificing the other. When they matched physiological CAQ profiles from proteomics onto this front, the agreement was striking. The Pareto front correctly predicted the direction of deviation between K and CAQ for more than eighty percent of reaction pairs in human glycolysis, and the CAQ-matched Pareto solution correlated strongly with experimental values in all three species, with Pearson correlations between 0.62 and 0.71. The same framework extended successfully to the E. coli TCA cycle, where deviations clustered around the alpha-ketoglutarate dehydrogenase step, a known control point of that pathway.

Perhaps the most convincing validation came from metabolite concentrations. When the researchers simulated glycolysis using enzyme profiles that maximize flux efficiency alone, the predicted metabolite levels were systematically far too high and matched experiments poorly, with a root mean squared error around four orders of magnitude in log space. Enzyme profiles from the Pareto front, however, reproduced measured metabolite concentrations remarkably well, achieving correlations of 0.67 in human cells, 0.90 in yeast, and 0.69 in E. coli. Crucially, the team also showed that this agreement was not an artifact of calibrating the Pareto solution with proteomics: when they instead selected solutions using independent metabolomics data, the predictions still matched proteomics-derived CAQ profiles, with permutation tests confirming that real enzyme allocations sit significantly closer to the Pareto front than randomly shuffled ones.

Finally, the study asked what determines where along the Pareto front a given cell chooses to operate. Analyzing eight cancer types in TCGA, the researchers found that lower EOI values, indicating tighter optimization of glycolytic efficiency, were consistently associated with activation of oncogenic signaling pathways, suggesting that aggressive tumors do not merely crank up glycolysis but actively reorganize it for maximum efficiency. Conversely, higher EOI correlated with enhanced expression of ribosomal and oxidative phosphorylation genes, hinting at competition between glycolytic efficiency and other energy-consuming programs. A third pattern was perhaps the most evocative: across 16 of 19 sample groups, cells that devoted a smaller fraction of their proteome to glycolytic enzymes showed stronger efficiency optimization. Scarcity, it seems, drives cells toward the same rational economizing that shapes human decision-making under budget constraints.

The implications reach well beyond basic biochemistry. Because the K-CAQ relationship provides an explicit, testable target for pathway design, it offers metabolic engineers a rational blueprint for building synthetic pathways that avoid the chronically low efficiencies that plague engineered metabolism today. Conceptually, the work reframes metabolic regulation as a multi-objective optimization problem, echoing design principles seen elsewhere in biology, from master regulators in development to command neurons in animal behavior, where concentrating control into a few nodes simplifies the management of a complex network. The authors caution that their framework does not capture every metabolic objective, including robustness to environmental fluctuations and metabolite load minimization, and a complete picture will require higher-dimensional Pareto fronts across genome-scale networks. But the core message stands: cells are not simply maximizing throughput, nor simply maximizing controllability, but navigating a quantifiable frontier between the two, and that frontier can now be read directly from omics data.

Subject of Research: Optimality principles governing enzyme allocation in metabolic pathways

Article Title: Trade-off between flux efficiency and metabolic control shapes enzyme allocation

Article References: Sun, K., Zheng, W., Fan, W., Ding, C., Huang, D., & Dai, Z. (2026). Trade-off between flux efficiency and metabolic control shapes enzyme allocation. Molecular Systems Biology. https://doi.org/10.1038/s44320-026-00250-5

Image Credits: AI Generated

DOI: 10.1038/s44320-026-00250-5

Keywords: metabolism, enzyme allocation, flux efficiency, metabolic control analysis, glycolysis, TCA cycle, Pareto optimality, thermodynamics, proteomics, systems biology, cancer metabolism, metabolic engineering