The thermodynamic limit of biomanufacturing: what it is, and how to reach it

The substrate fed to a fermenter carries a quantity of chemical free energy. Some of it is conserved in biomass and product; the rest is irreversibly degraded to heat. That split is not decided by equipment, protocol, or goodwill. It is decided by the second law of thermodynamics — and it has been measured, modeled, and argued about since the 1970s.

This review assembles what is known. Where is the thermodynamic floor of a biological production process? How close do cells already operate to it? And which levers actually move a designed process toward it? The answers turn out to be more concrete — and more actionable — than the word “limit” suggests.

The cell’s energy budget

The bookkeeping starts with Stouthamer, who in 1973 computed how much ATP is needed to synthesize a gram of microbial cell material and found that polymerization — stitching amino acids, nucleotides, and sugars into macromolecules — dominates the bill. His yield constant, Y_ATP ≈ 10 g of biomass per mole of ATP, has survived fifty years of scrutiny remarkably well. Russell and Cook’s 1995 synthesis of bacterial energetics sharpened the picture: in a growing cell, protein synthesis alone consumes roughly half of the total energy budget, and as growth slows, a rising share of energy is diverted to maintenance — holding gradients, repairing damage, staying alive — which buys no biomass at all.

Maintenance is not a fudge factor but a thermodynamic quantity. Tijhuis, van Loosdrecht and Heijnen showed in 1993 that the Gibbs energy a culture must dissipate per unit time simply to persist follows a robust, temperature-dependent correlation across organisms. Von Stockar and colleagues later formalized the whole ledger: microbial growth is a set of coupled chemical reactions, and like any chemical system it can be balanced in Gibbs energy — what enters as substrate must equal what is stored plus what is dissipated. With that balance in hand, “efficiency” stops being a metaphor and becomes a measurable number.

Where the floor sits

Heijnen and van Dijken’s 1992 search for a thermodynamic description of biomass yields located the first floor: the Gibbs energy dissipated per C-mole of biomass produced is not arbitrary, but correlates with the carbon source and the anabolic route. For aerobic growth on glucose, roughly half of the feed’s Gibbs energy is conserved in biomass; the other half is dissipated. Desmond-Le Quéméner and Bouchez (2014) and Roden and Jin (2011) generalized this into a thermodynamic theory of growth: across metabolisms as different as glucose respiration and hydrogenotrophy, growth rate and yield both track the energy left over after dissipation is paid.

The most striking recent result comes from Niebel, Leupold and Heinemann (2019): cells appear to respect an upper rate limit on Gibbs energy dissipation. As glucose uptake rises, metabolism rearranges its fluxes to stay below a critical dissipation rate; when every rearrangement option is exhausted, the cell has reached its maximal growth rate. The limit, in other words, is not a single number but a frontier — a maximum of conserved energy per unit dissipated, at a given rate — and evolution has already pushed organisms hard against it.

Rate versus yield: the tradeoff physics enforces

If the frontier is real, a cell cannot have everything. The growth laws of Scott et al. (2010) showed that ribosome abundance scales with growth rate: building protein-synthesis machinery is itself a major expense, and every ribosome made is carbon and energy not going into product. Molenaar and colleagues (2009) framed metabolism as cellular economics — strategies that maximize yield and strategies that maximize rate are different strategies, and cells shift between them as conditions change.

Basan et al. (2015) supplied the mechanistic explanation for the oldest observation in fermentation: overflow metabolism. E. coli excretes acetate on excess glucose not because it is broken, but because fermentative glycolysis is proteome-cheap — it yields less ATP per glucose, but more ATP per unit of enzyme protein. Flamholz et al. (2013) had shown the same tradeoff inside glycolysis itself, where the EMP and ED pathways sit at different points on the yield-versus-protein-cost frontier, and Noor et al. (2016) demonstrated that the protein cost of a flux distribution is computable in advance. Baldazzi et al. (2023) then showed that resource allocation alone explains most of the rate–yield variability across bacterial strains. The consequence for manufacturing is blunt: maximum production rate and maximum thermodynamic efficiency cannot be had at the same operating point. Choosing where to sit on the frontier is a design decision.

Carbon ceilings

Energy is not the only currency with a floor; carbon has one too. Every pathway has a stoichiometric maximum product yield, and thermodynamics-based metabolic flux analysis (Henry, Broadbelt and Hatzimanikatis, 2007) can prune flux distributions that violate energy conservation before any experiment is run. Bogorad, Lin and Liao (2013) then proved the ceiling itself can be moved: their synthetic non-oxidative glycolysis converts glucose to acetyl-CoA with complete carbon conservation — one hundred percent of feed carbon retained, where native glycolysis throws a third of it away as CO₂. Pathway design, not luck, decides the carbon ceiling.

The substrate side is designable as well. Bar-Even and colleagues (2012) showed that the known carbon-fixation pathways are essentially the ones thermodynamics permits to be efficient, and Gleizer et al. (2019) pushed E. coli across the boundary entirely, rewiring it to build all of its biomass carbon from CO₂. The limits are real — but their position depends on the network, and the network is editable.

The machine floor

Zoom in from the network to its parts and the picture sharpens. Bar-Even et al. (2011) surveyed thousands of enzymes and found most to be only “moderately efficient”: their kinetic parameters sit orders of magnitude below the diffusion limit, shaped by physiology rather than physics. Yet perfection exists. Triose phosphate isomerase is the textbook case of a catalytically perfect enzyme (Knowles and Albery, 1977), and the F₁-ATPase rotates with a mechanical efficiency near one hundred percent (Yasuda et al., 1998) — a molecular electric motor running essentially at the thermodynamic wall. Biology can build components that touch the limit. That whole cells do not reach it is therefore not a verdict of physics, but a consequence of system-level economics: rate, robustness, and yield all draw on the same budget.

The third currency: information

A cell is a chemical reactor, but it is also a computer — and computation has its own thermodynamic price. Landauer’s principle (1961) fixes the floor: erasing one bit of information releases at least kT ln 2 of heat, about 2.9 zeptojoules at 300 K. Hopfield’s kinetic proofreading (1974) showed what that means in practice: the ribosome buys its low error rate with extra energy dissipation, and more accuracy always costs more energy. Mehta and Schwab (2012) quantified the energetic cost of the computations a cell performs when it senses and responds to its environment, and Ouldridge, Govern and ten Wolde (2017) showed that copying information — what every polymerase does at every bond — carries a fundamental dissipation per bit copied.

For biomanufacturing, the lesson is that accuracy is a line item in the energy budget, not a free good. Fidelity, proofreading, sensing, and control all dissipate. Designing toward the limit therefore includes spending on precision only where it pays — a question a design tool can answer before the fermenter ever runs.

Engineering routes toward the limit

Read across these literatures and they converge on a short list of levers:

• Subtract the unnecessary. The minimal genome JCVI-syn3.0 (Hutchison et al., 2016) stripped a bacterium to 473 genes; every function removed is maintenance and replication overhead that no longer dissipates.
• Remove the cell. Cell-free expression systems (Silverman, Karim and Jewett, 2020) abandon growth entirely and point the energy budget almost wholly at product — at the price of preparing the extract.
• Decouple growth from production. Because maintenance dissipation is thermodynamically quantifiable (Tijhuis et al., 1993), production states that minimize growth minimize the fraction of feed spent on staying alive.
• Choose the feedstock. Methylotrophic hosts such as Komagataella phaffii run a distinctive energy economy on methanol (Yurimoto, Oku and Sakai, 2011) — one of the reasons Pichia became a premier production host (Gasser et al., 2013).
• Choose the pathway. Carbon-conserving routes (Bogorad et al., 2013) and thermodynamically shaped networks (Bar-Even et al., 2012) raise the ceiling itself rather than merely approaching it.
• Measure what is dissipated. The calorimetric tradition (von Stockar et al., 2006) made dissipation measurable — and what is measurable is optimizable.

How close can a designed process get?

A working definition helps. Take η as the fraction of feed Gibbs energy that ends up conserved in product. For aerobic biomass on glucose, η sits near one half; for a secreted protein it is lower still, because carbon and energy must also build the cell that does the secreting. And η can never reach one: maintenance, control, and accuracy all require dissipation, and Niebel’s result suggests that evolution itself treats the dissipation rate — not efficiency — as the binding constraint at high growth.

The practical program is therefore not “defeat thermodynamics” but “stop wasting what thermodynamics allows”: maximize carbon conservation, minimize maintenance by design, minimize protein overhead per unit of product flux, spend on accuracy only where it pays, and measure dissipation so the loop can close. Each item is a design variable before it becomes a process variable — computable in advance, rankable in advance. That is the loop this platform is built to run: compute what can be computed, reserve slots for what must be measured, and let wet-lab data land where it belongs. The limit is not a wall at the end of the road. It is the road’s destination — and the map above is the itinerary.

References

Every claim above is anchored in the primary literature. Where PubMed indexes the paper we link the PMID; otherwise we link the DOI.

1. Stouthamer AH. A theoretical study on the amount of ATP required for synthesis of microbial cell material. Antonie van Leeuwenhoek, 1973. PMID:4148026
2. Russell JB, Cook GM. Energetics of bacterial growth: balance of anabolic and catabolic reactions. Microbiol Rev, 1995. PMID:7708012
3. Tijhuis L, van Loosdrecht MCM, Heijnen JJ. A thermodynamically based correlation for maintenance Gibbs energy requirements in aerobic and anaerobic chemotrophic growth. Biotechnol Bioeng, 1993. PMID:18613056
4. von Stockar U, et al. Thermodynamics of microbial growth and metabolism: an analysis of the current situation. J Biotechnol, 2006. PMID:16185782
5. Heijnen JJ, van Dijken JP. In search of a thermodynamic description of biomass yields for the chemotrophic growth of microorganisms. Biotechnol Bioeng, 1992. PMID:18601018
6. Desmond-Le Quéméner E, Bouchez T. A thermodynamic theory of microbial growth. ISME J, 2014. PMID:24522260
7. Roden EE, Jin Q. Thermodynamics of microbial growth coupled to metabolism of glucose, ethanol, short-chain organic acids, and hydrogen. Appl Environ Microbiol, 2011. PMID:21216913
8. Niebel B, Leupold S, Heinemann M. An upper limit on Gibbs energy dissipation governs cellular metabolism. Nat Metab, 2019. PMID:32694810
9. Scott M, et al. Interdependence of cell growth and gene expression: origins and consequences. Science, 2010. PMID:21097934
10. Molenaar D, et al. Shifts in growth strategies reflect tradeoffs in cellular economics. Mol Syst Biol, 2009. PMID:19888218
11. Basan M, et al. Overflow metabolism in Escherichia coli results from efficient proteome allocation. Nature, 2015. PMID:26632588
12. Flamholz A, et al. Glycolytic strategy as a tradeoff between energy yield and protein cost. Proc Natl Acad Sci USA, 2013. PMID:23630264
13. Noor E, et al. The protein cost of metabolic fluxes: prediction from enzymatic rate laws and cost minimization. PLoS Comput Biol, 2016. PMID:27812109
14. Baldazzi V, et al. Resource allocation accounts for the large variability of rate-yield phenotypes across bacterial strains. eLife, 2023. PMID:37255080
15. Henry CS, Broadbelt LJ, Hatzimanikatis V. Thermodynamics-based metabolic flux analysis. Biophys J, 2007. PMID:17172310
16. Bogorad IW, Lin TS, Liao JC. Synthetic non-oxidative glycolysis enables complete carbon conservation. Nature, 2013. PMID:24077099
17. Bar-Even A, et al. Thermodynamic constraints shape the structure of carbon fixation pathways. Biochim Biophys Acta, 2012. PMID:22609686
18. Gleizer S, et al. Conversion of Escherichia coli to generate all biomass carbon from CO₂. Cell, 2019. PMID:31778652
19. Bar-Even A, et al. The moderately efficient enzyme: evolutionary and physicochemical trends shaping enzyme parameters. Biochemistry, 2011. PMID:21506553
20. Knowles JR, Albery WJ. Perfection in enzyme catalysis: the energetics of triose phosphate isomerase. Acc Chem Res, 1977. doi:10.1021/ar50112a001
21. Yasuda R, Noji H, Kinosita K Jr, Yoshida M. F₁-ATPase is a highly efficient molecular motor that rotates with discrete 120° steps. Cell, 1998. PMID:9657145
22. Landauer R. Irreversibility and heat generation in the computing process. IBM J Res Dev, 1961. doi:10.1147/rd.53.0183
23. Hopfield JJ. Kinetic proofreading: a new mechanism for reducing errors in biosynthetic processes requiring high specificity. Proc Natl Acad Sci USA, 1974. PMID:4530290
24. Mehta P, Schwab DJ. Energetic costs of cellular computation. Proc Natl Acad Sci USA, 2012. PMID:23045633
25. Ouldridge TE, Govern CC, ten Wolde PR. The thermodynamics of computational copying in biochemical systems. Phys Rev X, 2017. doi:10.1103/PhysRevX.7.021004
26. Hutchison CA 3rd, et al. Design and synthesis of a minimal bacterial genome. Science, 2016. PMID:27013737
27. Silverman AD, Karim AS, Jewett MC. Cell-free gene expression: an expanded repertoire of applications. Nat Rev Genet, 2020. PMID:31780816
28. Yurimoto H, Oku M, Sakai Y. Yeast methylotrophy: metabolism, gene regulation and peroxisome homeostasis. Int J Microbiol, 2011. PMID:21754936
29. Gasser B, et al. Pichia pastoris: protein production host and model organism for biomedical research. Future Microbiol, 2013. PMID:23374125