Thermodynamic cost-controllability tradeoff in metabolic currency coupling Preprint: APS/123-QED Jumpei F. Yamagishi, Tetsuhiro S. Hatakeyama Abstract Cellular metabolism is globally regulated by various currency metabolites such as ATP, GTP, and NAD(P)H. These metabolites cycle between charged (high-energy) and uncharged (low-energy) states to mediate energy transfer. While distinct currency metabolites are associated with different metabolic functions, their charged and uncharged forms are generally interchangeable via biochemical reactions such as ATP + GDP ⇌ ADP + GTP {\rm ATP{\,+\,}GDP{\,\rightleftharpoons\,}ADP{\,+\,}GTP} and NADP + + NADH ⇌ NADPH + NAD + \rm NADP^{+}{\,+\,}NADH{\,\rightleftharpoons\,}NADPH{\,+\,}NAD^{+} . Thus, their energetic states are generally coupled and influence each other, which would hinder the independent regulation of different currency metabolites. Despite the extensive knowledge of the molecular biology of individual currency metabolites, it remains poorly understood how the coordination of various coupled currency metabolites shapes metabolic regulation, efficiency, and ultimately the evolution of organisms. Here, we present a minimal theoretical model of metabolic currency coupling and reveal a fundamental tradeoff relationship between metabolic controllability and thermodynamic cost: increasing the capacity to independently regulate multiple currency metabolites generally requires comparable abundances of those metabolites, which in turn incurs a higher entropy production rate. The tradeoff suggests that in complex environments, organisms evolutionarily favor an equal abundance of currency metabolites to enhance metabolic controllability at the expense of a higher thermodynamic cost; conversely, in simple environments, organisms evolve to have imbalanced amounts of them to reduce heat dissipation. These considerations also offer a hypothesis regarding evolutionary trends in nucleotide-pool balance and genomic GC content. I Introduction Living organisms must continually consume energy to maintain their order and avoid reaching thermodynamic equilibrium. In the 1940s, Erwin Schrödinger famously argued that life persists by avoiding equilibrium through metabolism Schrödinger 1944 ; Ornes 2017 . Since then, cellular metabolism has remained a central topic in the study of nonequilibrium thermodynamics, attracting the interest of physicists and biologists alike Von Bertalanffy 1950 ; von Stockar and Van der Wielen 2013 . Remarkably, recent advancements in stochastic thermodynamics have provided a rigorous nonequilibrium thermodynamic description for open chemical reaction networks, applicable to general metabolic reaction networks Sekimoto 2010 ; Schmiedl and Seifert 2007 ; Ge and Qian 2016 ; Rao and Esposito 2016 ; Yoshimura and Ito 2021 ; Cao and Liang 2025 . However, most theoretical efforts to date have taken one of two extremes: they either derive very general thermodynamic constraints applicable to arbitrary chemical reaction networks Schmiedl and Seifert 2007 ; Rao and Esposito 2016 ; Yoshimura and Ito 2021 , or they analyze specific, simple biochemical processes such as single-enzyme kinetics Seifert 2011 ; Rao and Peliti 2015 and molecular machines Golubeva and Imparato 2012 . A substantial conceptual gap remains between broad theoretical frameworks and detailed, context-specific biochemical studies. To reveal the key biophysical principles of metabolism, we must bridge this gap by developing a coarse-grained theory that preserves the unique characteristics of metabolic systems. One approach to appropriately coarse-graining the complexity of metabolism is to focus on a special class of compounds that globally mediate energy transfer across many metabolic reactions: coenzymes . A coenzyme is a small, non-protein organic compound that associates loosely with enzymes and participates in enzymatic reactions as a transferable carrier of chemical groups or electrons De Bolster 1997 ; Goldford et al. 2022 . Well-known examples of coenzymes include adenosine triphosphate (ATP), which carries high-energy phosphate groups, and nicotinamide adenine dinucleotide (NADH/NAD + ), which carries high-energy electrons. Coenzymes typically cycle between high-energy (charged) and low-energy (uncharged) forms Reich 1981 ; Hatakeyama and Furusawa 2017 ; West et al. 2023 . ATP, for instance, is converted by hydrolysis into its lower-energy forms ADP or AMP (adenosine diphosphate or monophosphate), releasing energy, whereas ADP can be recharged back into ATP through processes such as cellular respiration or photosynthesis. Figure 1: Metabolic currency coupling. (A) Examples of transferases and their associated currency coupling reactions. (B) Model of the coupling of currency metabolites. All reactions are reversible. Table 1 collects all the notation in one place. (C) Schematics of how the balance in currency metabolite pools affects metabolic controllability and thermodynamic cost. Coupling reactions cause the ratios [A^{\ast}]/[A] and [B^{\ast}]/[B] to shift in the same direction. The former strongly influences the latter when {[A]_{\mathrm{tot}}}\gg{[B]_{\mathrm{tot}}} . Conversely, a large {[A]_{\mathrm{tot}}}/{[B]_{\mathrm{tot}}} ratio reduces the entropy production rate required to maintain the metabolic currency coupling, which is analogous to the “friction” between different currency metabolites. (D) A theoretical connection between organismal or habitat complexity and pool sizes of multiple currency metabolites. Because of their central role in energy transfer, coenzymes are often referred to as currency metabolites or metabolic currencies Huss and Holme 2007 ; Gerlee et al. 2009 . Much like money in an economy, currency metabolites fulfill Jevons’s functions of money Jevons 1875 ; Mankiw 2013 : they serve as a medium of exchange (transferring energy or chemical groups), a unit of account (providing a characteristic energy metric across reactions), and a store of value (storing energy). The charged/uncharged ratios of currency metabolites, such as the ATP/ADP ratio, regulate metabolic activity globally and serve as macroscopic indicators of the cellular metabolic state Goldford et al. 2022 ; Metallo and Vander Heiden 2013 ; Koebmann et al. 2002 ; Bekiaris and Klamt 2023 : a higher ratio biases metabolism toward overall ATP-driven reactions over ADP-driven ones, while a lower ratio favors ADP-driven reactions. Cellular metabolic systems employ multiple different currency metabolites (e.g., ATP, GTP, NAD(P)H, and acetyl-CoA), each generally associated with specific biochemical functions Bekiaris and Klamt 2023 ; West et al. 2025 : for example, ATP is known as the primary energy currency for most biochemical reactions, while GTP works for signaling and protein synthesis processes, such as powering G-proteins and the translation of mRNA Wittinghofer and Vetter 2011 ; Mrnjavac and Martin 2025 ; among the electron carriers, NADPH predominantly drives anabolic (biosynthetic) reactions, while NADH feeds into catabolic reactions, such as cellular respiration Ying 2008 ; Cantó et al. 2015 . Cells are considered to coordinate various metabolic demands separately by allocating different currency metabolites to different tasks. Despite their distinct roles, the energetic states of different currency metabolites are thermodynamically and chemically coupled rather than isolated. In general, the charged and uncharged forms of distinct currency metabolites can be interconverted through specific enzymatic reactions Reich 1981 . For example, nucleoside diphosphate kinase catalyzes a reversible phosphate exchange between nucleoside diphosphates and triphosphates (NDPs and NTPs), effectively linking the ATP/ADP and GTP/GDP ratios through the reaction ATP + GDP \rightleftharpoons ADP + GTP Lascu and Gonin 2000 ; other transferases (EC 2) perform analogous exchanges (Fig. 1 A). Such coupling reactions will hinder the independent regulation of different currency metabolites: if the intracellular ATP/ADP ratio is increased, for instance, coupling reactions will inevitably push the GTP/GDP ratio upward as well. While extensive knowledge of the molecular biology of individual currency metabolites has been accumulated, how the coordination of various coupled currency metabolites influences metabolic regulation, efficiency, and ultimately the physiology and evolution of organisms remains poorly understood. To address this gap, we theoretically elucidate inherent, system-wide constraints on how cellular metabolism is regulated via metabolic currency coupling. In this study, we develop a minimal theoretical model of metabolic currency coupling. We thereby derive a fundamental tradeoff between metabolic controllability and thermodynamic cost: balanced currency metabolite pools enhance metabolic control at the expense of increased entropy production rate as a thermodynamic cost (see also Fig. 1 C). From a biological standpoint, this tradeoff suggests that organisms in complex environments would evolutionarily favor uniformity in currency metabolite pools to increase their metabolic controllability; in contrast, those in simpler environments would evolutionarily favor a skewed distribution of currency pools to reduce the thermodynamic cost (Fig. 1 D). We then explore biological and evolutionary implications of this tradeoff, including a hypothetical link between ATP–GTP coupling, organismal complexity, and genomic GC content. II Model As the simplest case, we consider the coupling of two currency metabolites X{\,=\,}A,B (see also Fig. 1 B and Table 1 ). Each currency metabolite is driven from its uncharged state X into its charged state X^{\ast} via a single effective reaction, \displaystyle X+F_{X}^{+}\overset{k^{+}_{X}}{\underset{k^{-}_{X}}{\rightleftharpoons}}X^{\ast}+F_{X}^{-}\quad(X=A,B). (1) Here, F_{X}^{+} coarse-grains a pool of energy‐rich “fuel” substrates (e.g., nutrients or reducing equivalents), and F_{X}^{-} represents a lumped set of lower-energy products or waste. Thus, this reversible reaction represents both catabolic (rightward) and anabolic (leftward) processes. Different currency metabolites interconvert between their charged and uncharged states as \displaystyle A+B^{\ast}\overset{k^{+}_{\mathrm{cpl}}}{\underset{k^{-}_{\mathrm{cpl}}}{\rightleftharpoons}}A^{\ast}+B. (2) For the sake of simplicity, it is assumed that the difference in standard chemical potentials between X^{\ast} and X , \Delta\mu^{\circ}_{X}{\,:=\,}\mu_{X^{\ast}}^{\circ}-\mu_{X}^{\circ}>0 , is identical for both currency metabolites X=A,B . From the local detailed balance condition, k^{+}_{\mathrm{cpl}}{\,=\,}k^{-}_{\mathrm{cpl}} then holds, which will be denoted by {k_{\mathrm{cpl}}} hereafter. Assuming mass action kinetics, the rate equation for this system is given as \displaystyle\frac{{\rm d}}{{\rm d}t}\begin{pmatrix}[A]\\ [A^{\ast}]\\ [B]\\ [B^{\ast}]\end{pmatrix} \displaystyle= \displaystyle\begin{pmatrix}-1&0&-1\\ 1&0&1\\ 0&-1&1\\ 0&1&-1\end{pmatrix}\begin{pmatrix}J_{A}\\ J_{B}\\ J_{\mathrm{cpl}}\end{pmatrix}, (3) \displaystyle J_{A} \displaystyle= \displaystyle\kappa^{+}_{A}[A]-\kappa^{-}_{A}[A^{\ast}], \displaystyle J_{B} \displaystyle= \displaystyle\kappa^{+}_{B}[B]-\kappa^{-}_{B}[B^{\ast}], \displaystyle J_{\mathrm{cpl}} \displaystyle= \displaystyle{k_{\mathrm{cpl}}}[A][B^{\ast}]-{k_{\mathrm{cpl}}}[A^{\ast}][B], where the effective rate constants for the driving reactions, defined as \kappa_{X}^{\pm}{\,:=\,}k_{X}^{\pm}[F_{X}^{\pm}] , quantify how strongly currency metabolite X is driven into its charged and uncharged state by catabolic and anabolic processes. Each reaction conserves the total pool of each currency metabolite X , i.e., \displaystyle[X]_{\mathrm{tot}}:=[X^{\ast}]+[X]\quad(X=A,B) are conserved moieties. By exploiting the symmetry between currency metabolites, we order their total pool sizes as {[A]_{\mathrm{tot}}}\,{\geq}\,{[B]_{\mathrm{tot}}} , unless otherwise stated. Table 1: Notations Symbol Description X,X^{\ast} Currency metabolite in uncharged/charged state (X{\,=\,}A{,}B) [X]_{\mathrm{tot}} Pool size of currency metabolite X {k_{\mathrm{cpl}}} Rate constant for the coupling reaction (Eq. ( 2 )) k^{\pm}_{X} Rate constant for the driving reactions (Eq. ( 1 )) F_{X}^{\pm} Fuel ( F_{X}^{+} ) and product ( F_{X}^{-} ) for the driving reactions \kappa^{\pm}_{X},\kappa_{X} Effective rate constant for the driving reactions ( \kappa_{X}^{\pm}{:=}k_{X}^{\pm}[F_{X}^{\pm}];\kappa_{X}{:=}\kappa_{X}^{+}{+}\kappa_{X}^{-} ) \Gamma_{X} Charged/Uncharged ratio of X ( \Gamma_{X}{\,:=\,}[X^{\ast}]/[X] ) \Gamma_{0} \Gamma_{X} in the strong coupling limit (Eq. ( 5 )) e^{\pm}_{XY} Elasticity of \Gamma_{X} by \kappa_{Y}^{\pm} (Eq. ( 6 )) e_{X} Geometric mean of e^{\pm}_{XX} (Eq. ( 13 )) e Mean elasticity (Eq. ( 14 )) \tilde{e}^{\pm}_{X} e^{\pm}_{YX} in strong coupling limit \sigma_{\mathrm{cpl}} Entropy production rate (EPR) of the coupling reaction III Results The present study focuses on the charged/uncharged ratios of currency metabolites X , \Gamma_{X}{\,:=\,}[X^{\ast}]/[X] . These ratios determine the macroscopic state of intracellular metabolism: when \Gamma_{X} increases (decreases), the net fluxes of X^{\ast} -driven reactions are preferentially promoted (suppressed) overall. First, \Gamma_{X} at the steady state, denoted by \Gamma_{X}^{\mathrm{st}} , is calculated as (see also Appendix A ): \displaystyle\Gamma_{A}^{\mathrm{st}}=\frac{[A^{\ast}]^{\mathrm{st}}}{[A]^{\mathrm{st}}}=\frac{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{+}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{+}+\kappa_{A}^{+}\kappa_{B}}{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{-}+\kappa_{A}^{-}\kappa_{B}}, \displaystyle\Gamma_{B}^{\mathrm{st}}=\frac{[B^{\ast}]^{\mathrm{st}}}{[B]^{\mathrm{st}}}=\frac{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{+}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{+}+\kappa_{A}\kappa_{B}^{+}}{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{-}+\kappa_{A}\kappa_{B}^{-}}, (4) where \kappa_{X}{\,:=\,}\kappa_{X}^{+}{\,+\,}\kappa_{X}^{-} (see also Fig. 2 A). Without currency coupling (i.e., {k_{\mathrm{cpl}}}{\,=\,}0 ) \Gamma_{X}^{\mathrm{st}}{\,=\,}\kappa_{X}^{+}/\kappa_{X}^{-} holds, as the coupling reaction deviates \Gamma_{X}^{\mathrm{st}} from the \kappa_{X}^{+}/\kappa_{X}^{-} ratio. This deviation depends on the pool size [Y]_{\mathrm{tot}} of the other currency metabolite Y\,({\neq}\,X) as well as the coupling constant {k_{\mathrm{cpl}}} . In the limit of weak currency coupling, {k_{\mathrm{cpl}}}{\,\ll\,}\kappa_{X}^{\pm}/{[X]_{\mathrm{tot}}} , \displaystyle\Gamma_{X}^{\mathrm{st}}=\frac{\kappa_{X}^{+}}{\kappa_{X}^{-}}+{k_{\mathrm{cpl}}}{[Y]_{\mathrm{tot}}}\frac{\kappa_{Y}^{-}}{\kappa_{Y}\kappa_{X}^{-}}\left(\frac{\kappa_{Y}^{+}}{\kappa_{Y}^{-}}-\frac{\kappa_{X}^{+}}{\kappa_{X}^{-}}\right)+O(k^{2}_{\mathrm{cpl}}). As [Y]_{\mathrm{tot}} increases, \Gamma_{X}^{\mathrm{st}} deviates more from {\kappa_{X}^{+}}/{\kappa_{X}^{-}} toward {\kappa_{Y}^{+}}/{\kappa_{Y}^{-}} . By contrast, in the strong coupling limit, {k_{\mathrm{cpl}}}\,{\gg}\,\kappa_{X}^{\pm}/{[X]_{\mathrm{tot}}} , both steady‐state ratios \Gamma_{A}^{\mathrm{st}} and \Gamma_{B}^{\mathrm{st}} collapse onto a single value as \displaystyle\Gamma_{X}^{\mathrm{st}}\to\Gamma_{0}:=\frac{{[A]_{\mathrm{tot}}}\kappa_{A}^{+}+{[B]_{\mathrm{tot}}}\kappa_{B}^{+}}{{[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{[B]_{\mathrm{tot}}}\kappa_{B}^{-}}\quad(X=A,B), (5) which is a weighted average of the individual driving‐force ratios \kappa_{X}^{+}/\kappa_{X}^{-} with weights {[X]_{\mathrm{tot}}} . With strong currency coupling, the relative size of the currency pools determines the charged/uncharged ratio, \Gamma_{X}^{\mathrm{st}} . Figure 2: Elasticity of the charged/uncharged ratios with metabolic currency coupling. (A) Dependence of charged/uncharged ratio \Gamma_{X}^{\mathrm{st}}\,(X{\,=\,}A,B) on \kappa_{B}^{+} . Colored lines show the cases with {k_{\mathrm{cpl}}}{\,=\,}1,{[B]_{\mathrm{tot}}}{\,=\,}1 and different {[A]_{\mathrm{tot}}} , while gray line shows the no coupling case. (B) Dependence of self-elasticity e^{+}_{BB} on {[A]_{\mathrm{tot}}}\kappa_{A}^{+}/{[B]_{\mathrm{tot}}}\kappa_{B}^{+} . Colored lines show different coupling strength {k_{\mathrm{cpl}}} and black line shows the large coupling limit, \tilde{e}^{+}_{B} (Eq. ( 11 )). (C) Relationship between {[A]_{\mathrm{tot}}}/{[B]_{\mathrm{tot}}} and mean elasticity e . The black line shows the strong coupling limit. In panels (B) and (C), \kappa_{B}^{+}{\,=\,}1/3 and {[A]_{\mathrm{tot}}}{\,+\,}{[B]_{\mathrm{tot}}}{\,=\,}2 are fixed. The other parameters are set as \kappa_{A}^{+}{\,=\,}2/3 , \kappa_{A}^{-}{\,=\,}1/3 , and \kappa_{B}^{-}{\,=\,}2/3 . III.1 Controllability of charged/uncharged ratios The adaptive regulation of cellular metabolism requires the appropriate modulation of the charged/uncharged ratios \Gamma_{X} in response to changing environmental conditions such as nutrient availability. To quantify this controllability, we adopt the concept of elasticity from economics Samuelson 1947 , measuring the susceptibility of the steady-state charged/uncharged ratios \Gamma_{X}^{\mathrm{st}} to variations in the effective rate constant for a driving reaction, \kappa_{Y}^{\pm} : \displaystyle e^{\pm}_{XY}:=\pm\frac{\partial\log\Gamma_{X}^{\mathrm{st}}}{\partial\log\kappa_{Y}^{\pm}}\quad(X,Y=A,B). (6) where it is defined in logarithmic form to be dimensionless. As \Gamma_{X}^{\mathrm{st}} ( 1/\Gamma_{X}^{\mathrm{st}} ) increases linearly with the driving strength \kappa_{X}^{+} ( \kappa_{X}^{-} ) as shown in Eq. ( 4 ), each elasticity e^{\pm}_{XY} takes a constant, dimensionless value between 0 and 1 : \displaystyle e^{\pm}_{AA}=\; \displaystyle\frac{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{\pm}+\kappa_{A}^{\pm}\kappa_{B}}{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{\pm}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{\pm}+\kappa_{A}^{\pm}\kappa_{B}}, (7) \displaystyle e^{\pm}_{AB}=\; \displaystyle\frac{{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{\pm}}{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{\pm}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{\pm}+\kappa_{A}^{\pm}\kappa_{B}}, (8) \displaystyle e^{\pm}_{BA}=\; \displaystyle\frac{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{\pm}}{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{\pm}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{\pm}+\kappa_{A}\kappa_{B}^{\pm}}, (9) \displaystyle e^{\pm}_{BB}=\; \displaystyle\frac{{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{\pm}+\kappa_{A}\kappa_{B}^{\pm}}{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{\pm}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{\pm}+\kappa_{A}\kappa_{B}^{\pm}}. (10) Note that the elasticity of \Gamma^{\mathrm{st}}_{X} against different driving reactions satisfies e^{\pm}_{XA}{\,+\,}e^{\pm}_{XB}{\,=\,}1\;(X{\,=\,}A,B) . Equations ( 7 )-( 10 ) show that elasticity is governed not only by the rate constants for the driving and coupling reactions, \kappa_{X}^{\pm} and {k_{\mathrm{cpl}}} , but also by the currency metabolite pool sizes, {[X]_{\mathrm{tot}}} (Fig. 2 B). In particular, each elasticity e^{\pm}_{YX} depends on the relative magnitude of {[X]_{\mathrm{tot}}}\kappa_{X}^{\pm} . This is particularly evident in the strong coupling limit, {k_{\mathrm{cpl}}}\,{\gg}\,\kappa_{X}^{\pm}/{[X]_{\mathrm{tot}}} : \displaystyle e^{\pm}_{YX}\to\tilde{e}^{\pm}_{X}:=\frac{[X]_{\mathrm{tot}}\kappa_{X}^{\pm}}{{[A]_{\mathrm{tot}}}\kappa_{A}^{\pm}+{[B]_{\mathrm{tot}}}\kappa_{B}^{\pm}}. (11) This \tilde{e}^{\pm}_{X} satisfies \displaystyle e^{\pm}_{XX}\geq\tilde{e}^{\pm}_{X}\geq e^{\pm}_{YX}\quad(Y\,{\neq}\,X), (12) and it thus provides the lower (upper) bound of the dimensionless elasticity e^{\pm}_{XX} ( e^{\pm}_{YX} ) (see also Fig. 2 B). Remarkably, when the currency metabolite pools become imbalanced ( {[A]_{\mathrm{tot}}}\,{\gg}\,{[B]_{\mathrm{tot}}} ), the less abundant currency B loses susceptibility to its own drives \kappa_{B}^{\pm} and becomes predominantly slaved to \Gamma_{A}^{\mathrm{st}} and \kappa_{A}^{\pm} (Fig. 2 AB). In other words, the model effectively behaves as a single-degree-of-freedom system. Intuitively, this occurs because the more abundant currency metabolite A acts as a chemical bath against the less abundant currency metabolite B (Fig. 1 C). Taking the geometric mean of “self-elasticity” e^{\pm}_{XX} yields a single measure of the controllability of each \Gamma_{X}^{\mathrm{st}} : \displaystyle e_{X}:=\sqrt{e^{+}_{XX}e^{-}_{XX}}\quad(X=A,B). (13) Furthermore, the geometric mean of e_{X} yields a single measure of the system-wide controllability: \displaystyle e:=\sqrt{e_{A}e_{B}}=\left(e^{+}_{AA}e^{-}_{AA}e^{+}_{BB}e^{-}_{BB}\right)^{1/4}. (14) Consistent with the above discussion, imbalanced currency metabolite pools ( {[A]_{\mathrm{tot}}}\,{\gg}\,{[B]_{\mathrm{tot}}} ) bias e^{\pm}_{XX} , and consequently e_{X} , thereby reducing mean elasticity e (Fig. 2 C). Last but not least, in addition to controllability in terms of elasticity, the achievable ranges of the charged/uncharged ratios \Gamma_{X}^{\mathrm{st}} are constrained by the {[\mathrm{ATP}]}_{\mathrm{tot}}/{[\mathrm{GTP}]}_{\mathrm{tot}} ratio (see Appendix B for details). To achieve the wide range of charged/uncharged ratios observed in vivo Liesa and Shirihai 2013 , cells must regulate not only the driving reactions but also the pool sizes of currency metabolites. III.2 Entropy production rate as a thermodynamic cost Cellular metabolism requires energy to homeostatically maintain their nonequilibrium states. The coupling of multiple currency metabolites would result in an increase in entropy production rate (EPR) as a thermodynamic cost Himeoka and Kaneko 2014 ; Droste et al. 2024 . The EPR associated with coupling reaction, \sigma_{\mathrm{cpl}} , is given by Yoshimura and Ito 2021 \displaystyle\sigma_{\mathrm{cpl}}:=\; \displaystyle\left({k_{\mathrm{cpl}}}[A][B^{\ast}]-{k_{\mathrm{cpl}}}[A^{\ast}][B]\right)\log\frac{[A][B^{\ast}]}{[A^{\ast}][B]} \displaystyle=\; \displaystyle{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}{[B]_{\mathrm{tot}}}\frac{\Gamma_{B}-\Gamma_{A}}{(\Gamma_{A}+1)(\Gamma_{B}+1)}\log\frac{\Gamma_{B}}{\Gamma_{A}}. It increases with both the difference between \Gamma_{X} and the product of currency pool sizes, {k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}{[B]_{\mathrm{tot}}} . Intuitively, the EPR associated with currency coupling can be viewed as the effective “friction” between different currency metabolites (Fig. 1 C). Examining the steady-state (housekeeping) EPR \sigma_{{\mathrm{cpl}}}^{\mathrm{st}} in the strong coupling limit ( {k_{\mathrm{cpl}}}\,{\gg}\,\kappa_{X}^{\pm}/{[X]_{\mathrm{tot}}} ) provides further implications into the thermodynamic cost of currency coupling. In this limit, the housekeeping EPR of currency coupling is given by (see Appendix A for derivation): \displaystyle\sigma^{\mathrm{st}}_{\mathrm{cpl}}\simeq\; \displaystyle\frac{1}{{k_{\mathrm{cpl}}}}\frac{(\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-})^{2}{[A]_{\mathrm{tot}}}{[B]_{\mathrm{tot}}}}{({[A]_{\mathrm{tot}}}\kappa_{A}^{+}+{[B]_{\mathrm{tot}}}\kappa_{B}^{+})({[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{[B]_{\mathrm{tot}}}\kappa_{B}^{-})} \displaystyle\propto\; \displaystyle\frac{1}{{k_{\mathrm{cpl}}}}\frac{\frac{{[A]_{\mathrm{tot}}}}{{[B]_{\mathrm{tot}}}}}{\left(\frac{{[A]_{\mathrm{tot}}}}{{[B]_{\mathrm{tot}}}}\kappa_{A}^{+}+\kappa_{B}^{+}\right)\left(\frac{{[A]_{\mathrm{tot}}}}{{[B]_{\mathrm{tot}}}}\kappa_{A}^{-}+\kappa_{B}^{-}\right)}, (15) which depends on the {[A]_{\mathrm{tot}}}/{[B]_{\mathrm{tot}}} ratio, but is independent of the total abundance of currency metabolites, {[A]_{\mathrm{tot}}}{\,+\,}{[B]_{\mathrm{tot}}} . It is maximized at {[A]_{\mathrm{tot}}}/{[B]_{\mathrm{tot}}}{\,=\,}\left.\sqrt{\kappa_{B}^{+}\kappa_{B}^{-}}\middle/\sqrt{\kappa_{A}^{+}\kappa_{A}^{-}}\right. and decreases as the system deviates from this balanced {[A]_{\mathrm{tot}}}/{[B]_{\mathrm{tot}}} ratio (Fig. 3 A). A similar approximation-based analysis also applies to the weak coupling limit ( {k_{\mathrm{cpl}}}{\,\ll\,}\kappa_{X}^{\pm}/{[X]_{\mathrm{tot}}} ), yielding a qualitatively consistent result: EPR \sigma_{{\mathrm{cpl}}}^{\mathrm{st}} increases as the {[A]_{\mathrm{tot}}}/{[B]_{\mathrm{tot}}} ratio approaches 1 (see Appendix A for details). Figure 3: Thermodynamic cost of currency coupling. (A) Dependence of {k_{\mathrm{cpl}}}\sigma_{\mathrm{cpl}}^{\mathrm{st}} on {[A]_{\mathrm{tot}}}/{[B]_{\mathrm{tot}}} with relatively strong currency coupling. The black curve represents the strong coupling limit, Eq. ( III.2 ). For a consistent comparison, the system size is kept constant by setting {[A]_{\mathrm{tot}}}{\,+\,}{[B]_{\mathrm{tot}}}{\,=\,}2 while {[A]_{\mathrm{tot}}}/{[B]_{\mathrm{tot}}} is varied; parameters are set as \kappa_{A}^{+}{\,=\,}2/3,\kappa_{A}^{-}{\,=\,}1/3 , \kappa_{B}^{+}{\,=\,}1/3,\kappa_{B}^{-}{\,=\,}2/3 . (B) Relationship between the scaled EPR of the coupling reaction, \frac{{k_{\mathrm{cpl}}}\sqrt{\kappa_{A}^{+}\kappa_{A}^{-}\kappa_{B}^{+}\kappa_{B}^{-}}}{(\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-})^{2}}\sigma^{\mathrm{st}}_{\mathrm{cpl}} , and mean elasticity e . {k_{\mathrm{cpl}}}{\,=\,}10 . The parameters \kappa_{X}^{\pm} and {[X]_{\mathrm{tot}}} are randomly sampled from a uniform distribution within the range [0,1] . The black curve represents the strong coupling limit, Eq. ( 16 ). III.3 Tradeoff between elasticity and thermodynamic cost With strong currency coupling, from Eqs. ( 11 ) and ( III.2 ), the housekeeping EPR \sigma^{\mathrm{st}}_{\mathrm{cpl}} and elasticity are related as: \displaystyle\sigma^{\mathrm{st}}_{\mathrm{cpl}}\simeq\frac{1}{{k_{\mathrm{cpl}}}}\frac{(\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-})^{2}}{\sqrt{\kappa_{A}^{+}\kappa_{A}^{-}\kappa_{B}^{+}\kappa_{B}^{-}}}e^{2}. (16) It shows a tradeoff such that a higher (lower) mean elasticity is associated with a larger (smaller) EPR. We also numerically confirmed a positive correlation between e and \sigma^{\mathrm{st}}_{\mathrm{cpl}} even outside the strong coupling limit (Fig. 3 B); thus, this tradeoff could generally appear in systems with a coupling of different currency metabolites. In addition, the following tradeoff inequality holds for arbitrary {k_{\mathrm{cpl}}} (see Appendix A for the derivation): \displaystyle\frac{(\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-})^{2}}{{k_{\mathrm{cpl}}}\sqrt{\kappa_{A}^{+}\kappa_{A}^{-}\kappa_{B}^{+}\kappa_{B}^{-}}}e^{2}\geq\sigma^{\mathrm{st}}_{\mathrm{cpl}}. (17) The tradeoff suggests that greater uniformity in the currency metabolite pools could enhance the controllability of cellular metabolism and thus increase the effective degrees of freedom, at the expense of a higher thermodynamic cost (Fig. 1 C). Therefore, organisms inhabiting more complex environments, which require adaptive regulation of metabolic functions in response to higher-dimensional environmental cues, e.g., independent regulation of glycolysis via ATP, citric acid cycle via GTP, and pentose phosphate pathway via NADPH West et al. 2025 , are expected to favor more balanced currency pools during evolution and adaptation (Fig. 1 D). Figure 4: Statistical relationship between organismal complexity and GC content. (A) Correlation between genomic properties (genome size or the number of coding genes) and GC content. (B) Correlation between proteomic properties (the number of protein families or domains) and GC content. The data are obtained from dataset Madin et al. 2020 and Pfam Mistry et al. 2021 for 496 bacteria and archaea. III.4 Influence of currency coupling on evolution of GC content The above tradeoff has various biological implications, including a possible link between ATP–GTP coupling and an evolutionary trend in genomic GC content (see also Fig. 1 D). Genomic GC content refers to the percentage of guanine (G) and cytosine (C) in a genome. Interestingly, the GC content of bacterial genomes is highly diverse, ranging from 15\% to 75\% across species Rocha and Danchin 2002 ; Madin et al. 2020 . Although several factors have been proposed as contributing to this diversity, such as the differential stability due to the number of hydrogen bonds and sugar-acid preference arising from codon usage bias Sueoka 1961 ; Gralka et al. 2023 , the origin of this diversity remains unknown. Here, we focus on the statistical relationship between genomic GC content and environmental complexity: GC content generally correlates with various indicators of organismal complexity (Fig. 4 ). Fig. 4 A shows a positive correlation between genomic GC content and the genome size or the number of coding genes—widely-accepted indicators of organismal complexity, or the required complexity of physiological regulation Nelson et al. 2004 ; Sharov 2006 —across diverse microorganisms, as previously reported Almpanis et al. 2018 . GC content also positively correlates with the number of protein families or domains, a recently proposed quantitative indicator of organismal complexity Alvarez-Ponce and Krishnamurthy 2025 (Fig. 4 B). From an evolutionary perspective, it has been reported that organisms evolve to become GC-poor and AT-rich under simple environmental conditions Foerstner et al. 2005 ; Almpanis et al. 2018 . Endosymbiotic organisms, which inhabit more stable and simplified environments compared to free-living species, evolve to become GC-poor Rocha and Danchin 2002 ; McCutcheon et al. 2019 ; similarly, mitochondria and chloroplasts typically have GC-poor genomes Zhang et al. 2012 . Notably, observations from endosymbiotic evolution McCutcheon et al. 2019 suggest the existence of a mechanism that evolutionarily favors low GC content in simple environments, independently of genome size: a sharp reduction in genome size due to gene loss is followed by a progressive decrease in GC content (see Fig. 3 of ref. McCutcheon et al. 2019 ); in this regime, genome size and GC content are no longer correlated, but GC content continues to decrease as adaptation to “simpler” habitats increases. Our theory on metabolic currency coupling provides a hypothesis on the evolutionary tendency of GC-poor microbes to inhabit simple environments. NTPs are the building blocks of RNA, which is the second most abundant substance in cells after proteins Milo and Phillips 2015 . Therefore, the intracellular ATP/GTP ratio may be positively correlated with genomic GC content Dietel et al. 2019 (see also Appendix C ). From the perspective of the above controllability–cost tradeoff due to currency coupling, GC-rich microbes should tend to have higher metabolic controllability but higher thermodynamic cost, and vice versa . Although GC content may be affected by many other factors, this suggests that microbes in complex habitats should evolve to increase their GC content to enhance metabolic controllability despite the added energetic burden, while microbes in simple habitats should reduce their GC content to lower thermodynamic cost at the expense of controllability (see also Fig. 1 D). IV Discussion This work moves beyond the naïve view that distinct currency metabolites simply regulate distinct functions, instead providing a theoretical basis that quantitatively links the abundances and distributions of multiple currency metabolites to their physiological roles in managing controllability and energetic cost. By developing a minimal model of metabolic currency coupling that does not rely on specific molecules or reactions, we uncover a fundamental tradeoff between the metabolic controllability and the thermodynamic cost: greater controllability of cellular energetic states demands a higher thermodynamic cost, and vice versa . This theory further applies to diverse contexts in both evolution and adaptation. As an evolutionary example, we have argued how metabolic currency coupling could influence the evolutionary trend toward more GC-rich microbes in more complex environments. At the level of non-genetic adaptation, our theoretical results could be experimentally tested by manipulating the pool size of a currency metabolite: e.g., the addition of guanosine, which is converted to GTP via the salvage pathway, can increase GTP levels Kriel et al. 2012 . Note here that, although our minimal model assumes a constant coupling strength {k_{\mathrm{cpl}}} for simplicity, parameter values including {k_{\mathrm{cpl}}} in reality vary dynamically, depending on enzyme and metabolite abundances or growth conditions. An order-of-magnitude estimation of the parameters in our model suggests that {k_{\mathrm{cpl}}} and \kappa^{+}_{\rm ATP}/{[\mathrm{ATP}]}_{\mathrm{tot}} are typically of comparable magnitude, both on the order of 10^{3}\text{--}10^{4}\;{\rm[/M/s]} under physiological conditions (see Appendix D for details); this, in turn, implies that both strong and weak currency-coupling regimes can arise in natural cells as well as in experimentally manipulated systems. Our minimal model simplifies substrates and anabolic products into a single variable [F_{X}^{+}] to capture the essence of metabolic currency coupling. This leads to the unrealistic feature that the total EPR becomes zero in the case without currency coupling ( {k_{\mathrm{cpl}}}{\,=\,}0 ). Such limitations could be resolved by introducing a more detailed and realistic model which incorporates the dynamic change in coupling strength {k_{\mathrm{cpl}}} and distinguishes between catabolism and anabolism, without altering our core findings. For example, the present model can be easily extended to systems with the coupling of n{\,\geq\,}3 currency metabolites, where a similar tradeoff between elasticity and EPR exists (Appendix E ). The present study theoretically examines the monetary economics of energy metabolism, thereby revealing a key tradeoff between metabolic controllability and thermodynamic cost. As viewing biology through the lens of economics has already yielded valuable insight The Economic Cell Collective 2026 ; Yamagishi and Hatakeyama 2021 ; Yamagishi and Hatakeyama 2023 ; Yamagishi and Hatakeyama 2025 , we anticipate further intersections between cellular metabolism and monetary economics. For example, the economic theory and observation that a new denomination of currency emerges to reduce temporal and spatial ‘friction’ in the economy Kuroda 2020 seems consistent with the recent theory that the redundancy of currency metabolites (coenzymes) facilitates a parsimonious usage of the proteome Goldford et al. 2022 . We anticipate that further integrating nonequilibrium chemical thermodynamics, systems biology of metabolism, and mathematical economics will uncover universal constraints and design principles governing the economy of cellular metabolism. Acknowledgements. We would like to thank Ryuna Nagayama for his careful reading of the manuscript. We would also like to acknowledge Kohei Yoshimura, Yuichi Wakamoto, Ken-ichiro F. Kamei, Naoki Konno, and Takuma Ōnishi for helpful discussions and useful comments. J. F. Y. is supported by the RIKEN Research Fund for Special Postdoctoral Researcher (Project Code: 202501094096), JST ACT-X Grant Number JPMJAX25LG, and the Masason Foundation. T. S. H. is supported by JSPS KAKENHI Grant Number JP21K15048. Appendix A Properties of steady states The steady state of the present model is calculated as \displaystyle[B^{\ast}]^{\mathrm{st}}=\; \displaystyle\frac{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa^{+}_{A}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa^{+}_{B}+\kappa_{A}\kappa^{+}_{B}}{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}+\kappa_{A}\kappa_{B}}{[B]_{\mathrm{tot}}}, \displaystyle[B]^{\mathrm{st}}=\; \displaystyle\frac{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa^{-}_{A}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa^{-}_{B}+\kappa_{A}\kappa^{-}_{B}}{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}+\kappa_{A}\kappa_{B}}{[B]_{\mathrm{tot}}}, \displaystyle[A^{\ast}]^{\mathrm{st}}=\; \displaystyle\frac{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa^{+}_{A}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa^{+}_{B}+\kappa^{+}_{A}\kappa_{B}}{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}+\kappa_{A}\kappa_{B}}{[A]_{\mathrm{tot}}}, \displaystyle[A]^{\mathrm{st}}=\; \displaystyle\frac{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa^{-}_{A}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa^{-}_{B}+\kappa^{-}_{A}\kappa_{B}}{{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}+\kappa_{A}\kappa_{B}}{[A]_{\mathrm{tot}}}. (18) \Gamma_{X}^{\mathrm{st}}{\,:=\,}[X^{\ast}]^{\mathrm{st}}/[X]^{\mathrm{st}} ( 1/\Gamma_{X}^{\mathrm{st}} ) increases linearly with \kappa_{X}^{+} ( \kappa_{X}^{-} ), the effective rate constant of the forward (reverse) driving reaction for X (see Eq. ( 4 )). The entropy production rate (EPR) of the driving reaction for currency metabolite X{\,=\,}A,B is given as follows Yoshimura and Ito 2021 : \displaystyle\sigma_{X}:=\; \displaystyle\left(\kappa^{-}_{X}{[X^{\ast}]}-\kappa^{+}_{X}{[X]}\right)\log\frac{\kappa^{-}_{X}{[X^{\ast}]}}{\kappa^{+}_{X}{[X]}} \displaystyle=\; \displaystyle\frac{{[X]_{\mathrm{tot}}}\kappa^{-}_{X}}{\Gamma_{X}+1}\left(\Gamma_{X}-\Gamma^{\kappa}_{X}\right)\log\frac{\Gamma_{X}}{\Gamma^{\kappa}_{X}}, where \Gamma^{\kappa}_{X}{\,:=\,}\kappa_{X}^{+}/\kappa_{X}^{-} . That is, the entropy production due to the driving reaction for X arises from the difference between the charged/uncharged ratio \Gamma_{X} of currency metabolite X and the bias \Gamma^{\kappa}_{X}{\,=\,}\kappa_{X}^{+}/\kappa_{X}^{-} in the effective rate constants of forward and reverse driving reactions. Weak currency coupling limit. In the limit of weak currency coupling ( {k_{\mathrm{cpl}}}{\,\ll\,}\kappa_{X}^{\pm}/{[X]_{\mathrm{tot}}} ), the present model is reduced to a linear dynamical system. The steady state, ( 18 ), is then approximated as: \displaystyle[B^{\ast}]^{\mathrm{st}}\simeq\; \displaystyle{[B]_{\mathrm{tot}}}\left(\frac{\kappa_{B}^{+}}{\kappa_{B}}+{k_{\mathrm{cpl}}}\frac{{[A]_{\mathrm{tot}}}}{\kappa_{B}}\left(\frac{\kappa_{A}^{+}}{\kappa_{A}}-\frac{\kappa_{B}^{+}}{\kappa_{B}}\right)\right), \displaystyle[B]^{\mathrm{st}}\simeq\; \displaystyle{[B]_{\mathrm{tot}}}\left(\frac{\kappa_{B}^{-}}{\kappa_{B}}+{k_{\mathrm{cpl}}}\frac{{[A]_{\mathrm{tot}}}}{\kappa_{B}}\left(\frac{\kappa_{A}^{-}}{\kappa_{A}}-\frac{\kappa_{B}^{-}}{\kappa_{B}}\right)\right), \displaystyle[A^{\ast}]^{\mathrm{st}}\simeq\; \displaystyle{[A]_{\mathrm{tot}}}\left(\frac{\kappa_{A}^{+}}{\kappa_{A}}+{k_{\mathrm{cpl}}}\frac{{[B]_{\mathrm{tot}}}}{\kappa_{A}}\left(\frac{\kappa_{B}^{+}}{\kappa_{B}}-\frac{\kappa_{A}^{+}}{\kappa_{A}}\right)\right), \displaystyle[A]^{\mathrm{st}}\simeq\; \displaystyle{[A]_{\mathrm{tot}}}\left(\frac{\kappa_{A}^{-}}{\kappa_{A}}+{k_{\mathrm{cpl}}}\frac{{[B]_{\mathrm{tot}}}}{\kappa_{A}}\left(\frac{\kappa_{B}^{-}}{\kappa_{B}}-\frac{\kappa_{A}^{-}}{\kappa_{A}}\right)\right). Notably, the leading-order behavior of [X^{\ast}]^{\mathrm{st}}/{[X]_{\mathrm{tot}}} and [X]^{\mathrm{st}}/{[X]_{\mathrm{tot}}} exhibits no explicit dependence on {[X]_{\mathrm{tot}}} . These are primarily governed by the pool size of the other currency metabolite, {[Y]_{\mathrm{tot}}}\,(Y{\,\neq\,}X) , with the effect of {[X]_{\mathrm{tot}}} appearing only at second order. The same holds for the charged/uncharged ratio and self-elasticity: e.g., in the case of currency metabolite B , \displaystyle\Gamma_{B}^{\mathrm{st}}\simeq\; \displaystyle\Gamma^{\kappa}_{B}+{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\frac{\Gamma^{\kappa}_{B}+1}{\Gamma^{\kappa}_{A}+1}(\Gamma^{\kappa}_{A}-\Gamma^{\kappa}_{B}), \displaystyle e^{\pm}_{BB}=\; \displaystyle 1-{k_{\mathrm{cpl}}}\frac{{[A]_{\mathrm{tot}}}\kappa_{A}^{\pm}}{\kappa_{A}\kappa_{B}^{\pm}}+O(k_{\mathrm{cpl}}^{2}). This indicates that \Gamma_{B}^{\mathrm{st}} is determined by a balance between the effective rate constants, \kappa_{B}^{\pm} , of the driving reaction for currency metabolite B and the driving strength, {k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}} , of the coupling reaction by the other currency metabolite A . As a result, \sigma^{\mathrm{st}}_{i} ( i=A,B,{\mathrm{cpl}} ) are approximated as: \displaystyle\sigma^{\mathrm{st}}_{A}\simeq k_{\mathrm{cpl}}^{2}{[A]_{\mathrm{tot}}}\frac{\kappa_{A}^{+}\kappa_{A}^{-}}{\kappa_{A}}\left(\frac{{[B]_{\mathrm{tot}}}}{\kappa_{B}}\left(\frac{\kappa_{B}^{+}}{\kappa_{A}^{+}}-\frac{\kappa_{B}^{-}}{\kappa_{A}^{-}}\right)\right)^{2}, \displaystyle\sigma^{\mathrm{st}}_{B}\simeq k_{\mathrm{cpl}}^{2}{[B]_{\mathrm{tot}}}\frac{\kappa_{B}^{+}\kappa_{B}^{+}}{\kappa_{B}}\left(\frac{{[A]_{\mathrm{tot}}}}{\kappa_{A}}\left(\frac{\kappa_{A}^{+}}{\kappa_{B}^{+}}-\frac{\kappa_{A}^{-}}{\kappa_{B}^{-}}\right)\right)^{2}, \displaystyle\sigma^{\mathrm{st}}_{\mathrm{cpl}}\simeq{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}{[B]_{\mathrm{tot}}}\frac{\kappa^{-}_{A}\kappa^{+}_{B}-\kappa_{A}^{+}\kappa_{B}^{-}}{\kappa_{A}\kappa_{B}}\log\frac{\kappa^{-}_{A}\kappa^{+}_{B}}{\kappa^{+}_{A}\kappa^{-}_{B}}. (19) The non-negative term in ( 19 ), \frac{\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-}}{\kappa_{A}\kappa_{B}}\log\frac{\kappa_{A}^{-}\kappa_{B}^{+}}{\kappa_{A}^{+}\kappa_{B}^{-}} , quantifies the extent of asymmetry between the effective rate constants of driving reactions, \kappa_{A}^{\pm} and \kappa_{B}^{\pm} . When normalized by the square of the total currency pool {C_{\mathrm{tot}}}{\,:=\,}{[A]_{\mathrm{tot}}}{\,+\,}{[B]_{\mathrm{tot}}} , the housekeeping EPR of currency coupling, \sigma_{{\mathrm{cpl}}}^{\mathrm{st}}/{C_{\mathrm{tot}}}^{2}\propto{k_{\mathrm{cpl}}}r/(1+r)^{2} with r{\,:=\,}{[A]_{\mathrm{tot}}}/{[B]_{\mathrm{tot}}} , is maximized at r=1 . Note here that the total EPR, defined as \sigma_{\mathrm{tot}}{\,:=\,}\sum_{i=A,B,{\mathrm{cpl}}}\sigma_{i} , satisfies \sigma^{\mathrm{st}}_{\mathrm{tot}}=\sigma^{\mathrm{st}}_{\mathrm{cpl}}+O(k_{\mathrm{cpl}}^{2}) and is primarily governed by the coupling reaction. Fig. 5 numerically shows that the above approximation ( 19 ) holds well for small {k_{\mathrm{cpl}}} . As {k_{\mathrm{cpl}}} increases, the (scaled) EPR \sigma_{\mathrm{cpl}}^{\mathrm{st}}/{k_{\mathrm{cpl}}} decreases overall and deviates from ( 19 ), but it is still maximized at {[A]_{\mathrm{tot}}}/{[B]_{\mathrm{tot}}}{\,=\,}1 . Figure 5: Dependence of \sigma_{\mathrm{cpl}}^{\mathrm{st}}/{k_{\mathrm{cpl}}} on the {[A]_{\mathrm{tot}}}/{[B]_{\mathrm{tot}}} ratio with relatively weak currency coupling (colored points). The gray curve represents ( 19 ). For a consistent comparison, the system size is kept constant by setting {C_{\mathrm{tot}}}{\,=\,}{[A]_{\mathrm{tot}}}{\,+\,}{[B]_{\mathrm{tot}}}{\,=\,}2 while varying {[A]_{\mathrm{tot}}}/{[B]_{\mathrm{tot}}} ; parameters are set as \kappa_{A}^{+}{\,=\,}2/3,\kappa_{A}^{-}{\,=\,}1/3 , \kappa_{B}^{+}{\,=\,}1/3,\kappa_{B}^{-}{\,=\,}2/3 . Strong currency coupling limit. In the limit of strong currency coupling ( {k_{\mathrm{cpl}}}\,{\gg}\,\kappa_{X}^{\pm}/{[X]_{\mathrm{tot}}} ), the charged/uncharged ratios at the steady state are approximated as: \displaystyle\Gamma_{A}^{\mathrm{st}}=\Gamma_{0}+\frac{\kappa_{B}\left(\kappa_{A}^{+}-\Gamma_{0}\kappa_{A}^{-}\right)}{{k_{\mathrm{cpl}}}({[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{[B]_{\mathrm{tot}}}\kappa_{B}^{-})}+O(k^{-2}_{\mathrm{cpl}}), \displaystyle\Gamma_{B}^{\mathrm{st}}=\Gamma_{0}+\frac{\kappa_{A}\left(\kappa_{B}^{+}-\Gamma_{0}\kappa_{B}^{-}\right)}{{k_{\mathrm{cpl}}}({[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{[B]_{\mathrm{tot}}}\kappa_{B}^{-})}+O(k^{-2}_{\mathrm{cpl}}). Accordingly, \displaystyle\Gamma_{B}^{\mathrm{st}}-\Gamma_{A}^{\mathrm{st}}\simeq(\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-})\frac{1+\Gamma_{0}}{{k_{\mathrm{cpl}}}({[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{[B]_{\mathrm{tot}}}\kappa_{B}^{-})}, \displaystyle\frac{\Gamma_{B}^{\mathrm{st}}}{\Gamma_{A}^{\mathrm{st}}}\simeq 1+\frac{({[A]_{\mathrm{tot}}}\kappa_{A}+{[B]_{\mathrm{tot}}}\kappa_{B})\left(\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-}\right)}{{k_{\mathrm{cpl}}}({[A]_{\mathrm{tot}}}\kappa_{A}^{+}+{[B]_{\mathrm{tot}}}\kappa_{B}^{+})({[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{[B]_{\mathrm{tot}}}\kappa_{B}^{-})}. With the definition of \sigma^{\mathrm{st}}_{\mathrm{cpl}} , this approximation leads to Eq. ( III.2 ): \displaystyle\sigma^{\mathrm{st}}_{\mathrm{cpl}}\simeq\; \displaystyle\frac{1}{{k_{\mathrm{cpl}}}}\frac{(\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-})^{2}{[A]_{\mathrm{tot}}}{[B]_{\mathrm{tot}}}}{({[A]_{\mathrm{tot}}}\kappa_{A}^{+}+{[B]_{\mathrm{tot}}}\kappa_{B}^{+})({[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{[B]_{\mathrm{tot}}}\kappa_{B}^{-})} \displaystyle\propto\; \displaystyle\frac{1}{{k_{\mathrm{cpl}}}}\frac{\frac{{[A]_{\mathrm{tot}}}}{{[B]_{\mathrm{tot}}}}}{\left(\frac{{[A]_{\mathrm{tot}}}}{{[B]_{\mathrm{tot}}}}\kappa_{A}^{+}+\kappa_{B}^{+}\right)\left(\frac{{[A]_{\mathrm{tot}}}}{{[B]_{\mathrm{tot}}}}\kappa_{A}^{-}+\kappa_{B}^{-}\right)}. (20) Moreover, from Eq. ( A ) and Eq. ( 12 ), \sigma^{\mathrm{st}}_{\mathrm{cpl}} and \tilde{e}_{X} are related as \displaystyle\sigma^{\mathrm{st}}_{\mathrm{cpl}}\simeq\frac{(\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-})^{2}}{{k_{\mathrm{cpl}}}\sqrt{\kappa_{A}^{+}\kappa_{A}^{-}\kappa_{B}^{+}\kappa_{B}^{-}}}\sqrt{\tilde{e}^{+}_{A}\tilde{e}^{-}_{A}}\sqrt{\tilde{e}^{+}_{B}\tilde{e}^{-}_{B}} (21) which leads to Eq. ( 16 ). This equation reveals a tradeoff: higher (lower) approximate elasticities are associated with larger (smaller) EPR. Given the constraint that the elasticities satisfy \tilde{e}^{\pm}_{A}+\tilde{e}^{\pm}_{B}=1 , the geometric mean \tilde{e}_{A}\tilde{e}_{B} of the elasticities becomes larger when the elasticities \tilde{e}_{X} are evenly distributed. Tradeoff inequality. Let us prove the inequality ( 17 ). As e=\left(e^{+}_{AA}e^{-}_{AA}e^{+}_{BB}e^{-}_{BB}\right)^{1/4}\geq\left(\tilde{e}^{+}_{A}\tilde{e}^{-}_{A}\tilde{e}^{+}_{B}\tilde{e}^{-}_{B}\right)^{1/4} holds from Eq. ( 12 ), the left-hand side of ( 17 ) satisfies \displaystyle\frac{(\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-})^{2}}{{k_{\mathrm{cpl}}}\sqrt{\kappa_{A}^{+}\kappa_{A}^{-}\kappa_{B}^{+}\kappa_{B}^{-}}}e^{2}\geq\frac{(\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-})^{2}}{{k_{\mathrm{cpl}}}\sqrt{\kappa_{A}^{+}\kappa_{A}^{-}\kappa_{B}^{+}\kappa_{B}^{-}}}\sqrt{\tilde{e}^{+}_{A}\tilde{e}^{-}_{A}}\sqrt{\tilde{e}^{+}_{B}\tilde{e}^{-}_{B}} \displaystyle=\frac{{[A]_{\mathrm{tot}}}{[B]_{\mathrm{tot}}}}{{k_{\mathrm{cpl}}}}\frac{(\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-})^{2}}{({[A]_{\mathrm{tot}}}\kappa_{A}^{+}+{[B]_{\mathrm{tot}}}\kappa_{B}^{+})({[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{[B]_{\mathrm{tot}}}\kappa_{B}^{-})}. (22) Also, from \log x\leq x-1 , the right-hand side of ( 17 ) satisfies \displaystyle\sigma^{\mathrm{st}}_{\mathrm{cpl}} \displaystyle={k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}{[B]_{\mathrm{tot}}}\frac{\Gamma^{\mathrm{st}}_{B}-\Gamma^{\mathrm{st}}_{A}}{(\Gamma^{\mathrm{st}}_{A}+1)(\Gamma^{\mathrm{st}}_{B}+1)}\log\frac{\Gamma^{\mathrm{st}}_{B}}{\Gamma^{\mathrm{st}}_{A}} \displaystyle\leq{k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}{[B]_{\mathrm{tot}}}\frac{\Gamma^{\mathrm{st}}_{B}-\Gamma^{\mathrm{st}}_{A}}{(\Gamma^{\mathrm{st}}_{A}+1)(\Gamma^{\mathrm{st}}_{B}+1)}\left(\frac{\Gamma^{\mathrm{st}}_{B}}{\Gamma^{\mathrm{st}}_{A}}-1\right) \displaystyle={k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}{[B]_{\mathrm{tot}}}\frac{(\Gamma^{\mathrm{st}}_{B}-\Gamma^{\mathrm{st}}_{A})^{2}}{\Gamma^{\mathrm{st}}_{A}(\Gamma^{\mathrm{st}}_{A}+1)(\Gamma^{\mathrm{st}}_{B}+1)}. (23) Noting that \displaystyle\Gamma^{\mathrm{st}}_{B}-\Gamma^{\mathrm{st}}_{A}= \displaystyle\frac{(\kappa_{A}^{-}\kappa_{B}^{+}-\kappa_{A}^{+}\kappa_{B}^{-})}{({k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{-}+\kappa_{A}^{-}\kappa_{B})} \displaystyle\times\frac{({k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}+\kappa_{A}\kappa_{B})}{({k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{-}+\kappa_{A}\kappa_{B}^{-})}, (24) it is sufficient to prove \displaystyle\frac{1}{{k_{\mathrm{cpl}}}^{2}({[A]_{\mathrm{tot}}}\kappa_{A}^{+}+{[B]_{\mathrm{tot}}}\kappa_{B}^{+})({[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{[B]_{\mathrm{tot}}}\kappa_{B}^{-})} \displaystyle\geq\frac{1}{({k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{+}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{+}+\kappa_{A}^{+}\kappa_{B})} \displaystyle\quad\times\frac{1}{({k_{\mathrm{cpl}}}{[A]_{\mathrm{tot}}}\kappa_{A}^{-}+{k_{\mathrm{cpl}}}{[B]_{\mathrm{tot}}}\kappa_{B}^{-}+\kappa_{A}\kappa_{B}^{-})}. (25) As ( A ) is evident, we obtain Eq. ( 17 ). Appendix B Bounds on \Gamma_{X} modulation range In addition to controllability in terms of elasticity, the achievable ranges of the charged/uncharged ratios \Gamma_{X}^{\mathrm{st}} are fundamentally constrained by finite driving strength, \kappa_{X}^{\pm} . If the total driving strength \kappa_{X}{\,=\,}\kappa_{X}^{+}+\kappa_{X}^{-} has an upper bound \kappa_{X}^{\max} , \Gamma_{X}^{\mathrm{st}} is bounded as \Gamma_{X}^{\max}{\,\geq\,}\Gamma_{X}^{\mathrm{st}}{\,\geq\,}\Gamma_{X}^{\min} with \displaystyle\Gamma_{X}^{\max} \displaystyle= \displaystyle\frac{\kappa_{Y}^{+}}{\kappa_{Y}^{-}}+\frac{{k_{\mathrm{cpl}}}{[X]_{\mathrm{tot}}}+\kappa_{Y}}{{k_{\mathrm{cpl}}}{[Y]_{\mathrm{tot}}}\kappa_{Y}^{-}}\kappa_{X}^{\max}, \displaystyle\Gamma_{X}^{\min} \displaystyle= \displaystyle\left(\frac{\kappa_{Y}^{-}}{\kappa_{Y}^{+}}+\frac{{k_{\mathrm{cpl}}}{[X]_{\mathrm{tot}}}+\kappa_{Y}}{{k_{\mathrm{cpl}}}{[Y]_{\mathrm{tot}}}\kappa_{Y}^{+}}\kappa_{X}^{\max}\right)^{-1}.\quad(Y\,{\neq}\,X) These upper and lower bounds are governed by the relative magnitudes of {[X]_{\mathrm{tot}}}\kappa_{X}^{\pm} and {[Y]_{\mathrm{tot}}}\kappa_{Y}^{\pm} . When the currency metabolite pools are imbalanced ( {[A]_{\mathrm{tot}}}\,{\gg}\,{[B]_{\mathrm{tot}}} ), the charged/uncharged ratio \Gamma_{B}^{\mathrm{st}} of the less abundant currency metabolite B is confined to a narrower range \Gamma_{B}^{\max}/\Gamma_{B}^{\min} (Fig. 6 ), matching the trend observed in the elasticities. Therefore, to achieve the wide range of charged/uncharged ratios observed in vivo , cells must regulate not only the bias \kappa_{X}^{+}/\kappa_{X}^{-} of the driving reactions for currency metabolite X but also the pool sizes {[X]_{\mathrm{tot}}} of currency metabolites. Figure 6: Dependence of (A) \Gamma_{B}^{\mathrm{st}} modulation range, [\Gamma_{B}^{\min},\Gamma_{B}^{\max}] , and (B) its width, \Gamma_{B}^{\max}/\Gamma_{B}^{\min} , on {[A]_{\mathrm{tot}}} . \kappa_{B}{\,=\,}1 , {k_{\mathrm{cpl}}}{\,=\,}1 , and {[B]_{\mathrm{tot}}}{\,=\,}1 are fixed. Appendix C On the correlation between the intracellular ATP/GTP ratio and GC content The intracellular concentration [{\rm NTP}] of NTPs (i.e., ATP and GTP) is determined by the balance between their production flux P_{\rm NTP} and their conversion and dilution: \displaystyle\frac{{\rm d}}{{\rm d}t}[{\rm NTP}]=P_{\rm NTP}-k_{\rm NTP}[{\rm NTP}]. Then, their steady-state concentration is \displaystyle[{\rm NTP}]^{\mathrm{st}}=\frac{P_{\rm NTP}}{k_{\rm NTP}}. One might reasonably expect the production flux of GTP and guanine to increase with genomic GC content, as supported by experimental evidence Dietel et al. 2019 . We therefore assume that the production fluxes of ATP and GTP, P_{\rm ATP} and P_{\rm GTP} , positively correlate with the genomic AT and GC contents, respectively. Furthermore, if P_{\rm NTP} is decomposed into the flux used for RNA incorporation and that for de novo synthesis for currency metabolites, it is natural to assume that only the former is proportional to AT or GC content. Nevertheless, because the pool of free NTPs derived from RNA turnover can greatly exceed that from de novo synthesis, P_{\rm ATP} and P_{\rm GTP} are still expected to correlate positively with AT and GC content, respectively. According to ref. Milo and Phillips 2015 , the weight of RNA per cell is about 60 fg, and thus that of ATP or GTP is about 15 fg. In contrast, the intracellular concentrations of ATP and GTP are about 3 mM and 1 mM, respectively Danchin et al. 1984 ; Jewett et al. 2009 ; Deng et al. 2021 ; because the molecular weights of ATP and GTP are 507 and 523 \rm g/mol and the volume of microbes such as E. coli is about 1\,{\rm\mu m^{3}} , the weights of (free) ATP and GTP are about 1.5 fg and 0.5 fg, respectively. That is, the intracellular abundance of guanine molecules for RNA is higher by more than an order of magnitude. Appendix D Coupling of n\geq 3 Currency Metabolites As an extension of the minimal model discussed in the main text, we consider a system with n\geq 3 kinds of currency metabolites, X_{i}\;(i=1,2,\cdots,n) . Each metabolite X_{i} is driven by a reaction with effective rate constants \kappa^{\pm}_{i} : \displaystyle X_{i}\underset{\kappa^{+}_{i}}{\overset{\kappa^{-}_{i}}{{\rightleftharpoons}}}X_{i}^{\ast}\quad(i=1,2,\cdots,n). These currency metabolites are also assumed to be coupled through exchange reactions of the form 1 1 1 The following analysis can be applied to systems in which the coupling constant varies among currency metabolite pairs. : \displaystyle X_{i}+X_{j}^{\ast}\underset{{k_{\mathrm{cpl}}}}{\overset{{k_{\mathrm{cpl}}}}{{\rightleftharpoons}}}X_{i}^{\ast}+X_{j}. Assuming mass-action kinetics, the rate equation is given by \displaystyle\dot{[X^{\ast}_{i}]}=\; \displaystyle\kappa^{+}_{i}[X_{i}]-\kappa^{-}_{i}[X^{\ast}_{i}]+\sum_{i\neq j}{k_{\mathrm{cpl}}}([X_{i}][X^{\ast}_{j}]-[X^{\ast}_{i}][X_{j}]) \displaystyle=\; \displaystyle\kappa^{+}_{i}[X_{i}]_{\mathrm{tot}}-\kappa_{i}[X^{\ast}_{i}] \displaystyle+\sum_{j\neq i}{k_{\mathrm{cpl}}}([X_{i}]_{\mathrm{tot}}[X^{\ast}_{j}]-[X^{\ast}_{i}][X_{j}]_{\mathrm{tot}}), (26) where \kappa_{i}:=\kappa^{+}_{i}+\kappa^{-}_{i} (i=1,2,\cdots,n) . Steady state. From ( 26 ), the steady state [X^{\ast}_{i}]^{\mathrm{st}} satisfies \displaystyle\left({\bf D}+{\bf R}\right){\bf[X^{\ast}]^{\mathrm{st}}}=\{\kappa_{i}^{+}[X_{i}]_{\mathrm{tot}}\}_{i}, where a diagonal matrix {\bf D} and a rank-one matrix {\bf R} are defined as \displaystyle{\bf D}:={\rm diag}(\kappa_{i}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}})_{i},\quad{\bf R}:=-\{{k_{\mathrm{cpl}}}[X_{i}]_{\mathrm{tot}}\}_{i}\cdot{\bf 1}^{\top}, with {C_{\mathrm{tot}}}{\,:=\,}\sum_{j}[X_{j}]_{\mathrm{tot}} . Therefore, we have: \displaystyle{\bf[X^{\ast}]}^{\mathrm{st}}=({\bf D}+{\bf R})^{-1}\{\kappa_{i}^{+}[X_{i}]_{\mathrm{tot}}\}_{i}. Here, the inverse can be evaluated using the matrix inversion lemma (Woodbury matrix identity): \displaystyle({\bf D}+{\bf R})^{-1}={\bf D}^{-1}+ \displaystyle\;\;{\bf D}^{-1}\{{k_{\mathrm{cpl}}}[X_{i}]_{\mathrm{tot}}\}_{i}(1-{\bf 1}^{\top}{\bf D}^{-1}\{{k_{\mathrm{cpl}}}[X_{i}]_{\mathrm{tot}}\}_{i})^{-1}{\bf 1}^{\top}{\bf D}^{-1}. Accordingly, we obtain: \displaystyle[X^{\ast}_{i}]^{\mathrm{st}}=\frac{[X_{i}]_{\mathrm{tot}}}{\kappa_{i}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}}}\left(\kappa_{i}^{+}+{k_{\mathrm{cpl}}}\frac{\sum_{j}\frac{\kappa_{j}^{+}[X_{j}]_{\mathrm{tot}}}{\kappa_{j}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}}}}{1-\sum_{j}\frac{{k_{\mathrm{cpl}}}[X_{j}]_{\mathrm{tot}}}{\kappa_{j}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}}}}\right), \displaystyle[X_{i}]^{\mathrm{st}}=\frac{[X_{i}]_{\mathrm{tot}}}{\kappa_{i}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}}}\left(\kappa_{i}^{-}+{k_{\mathrm{cpl}}}\frac{\sum_{j}\frac{\kappa_{j}^{-}[X_{j}]_{\mathrm{tot}}}{\kappa_{j}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}}}}{1-\sum_{j}\frac{{k_{\mathrm{cpl}}}[X_{j}]_{\mathrm{tot}}}{\kappa_{j}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}}}}\right), \displaystyle\Gamma_{i}^{\mathrm{st}}:=\frac{[X^{\ast}_{i}]^{\mathrm{st}}}{[X_{i}]^{\mathrm{st}}}=\frac{\kappa_{i}^{+}+{k_{\mathrm{cpl}}}\sum_{j\neq i}\frac{(\kappa_{j}^{+}-\kappa_{i}^{+})[X_{j}]_{\mathrm{tot}}}{\kappa_{j}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}}}}{\kappa_{i}^{-}+{k_{\mathrm{cpl}}}\sum_{j\neq i}\frac{(\kappa_{j}^{-}-\kappa_{i}^{-})[X_{j}]_{\mathrm{tot}}}{\kappa_{j}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}}}}. (27) In the strong coupling limit, \displaystyle\Gamma_{i}^{\mathrm{st}}= \displaystyle\;\Gamma_{0}+k_{\mathrm{cpl}}^{-1}\frac{\sum_{j\neq i}\kappa_{j}[X_{j}]_{\mathrm{tot}}\left((\kappa_{i}^{+}-\kappa_{j}^{+})-\Gamma_{0}(\kappa_{i}^{-}-\kappa_{j}^{-})\right)}{{C_{\mathrm{tot}}}\sum_{j}[X_{j}]_{\mathrm{tot}}\kappa_{j}^{-}} \displaystyle+O(k_{\mathrm{cpl}}^{-2}), (28) with \displaystyle\Gamma_{0}:=\frac{\sum_{j}[X_{j}]_{\mathrm{tot}}\kappa_{j}^{+}}{\sum_{j}[X_{j}]_{\mathrm{tot}}\kappa_{j}^{-}}. Elasticity. The dimensionless elasticity is defined as \displaystyle e^{\pm}_{ij}:=\pm\frac{\partial\log\Gamma_{i}^{\mathrm{st}}}{\partial\log\kappa^{\pm}_{j}}. Then, \displaystyle e^{\pm}_{ii}=1-\frac{{k_{\mathrm{cpl}}}\sum_{l\neq i}\frac{\kappa_{l}^{\pm}[X_{l}]_{\mathrm{tot}}}{\kappa_{l}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}}}}{\kappa_{i}^{\pm}+{k_{\mathrm{cpl}}}\sum_{l\neq i}\frac{(\kappa_{l}^{\pm}-\kappa_{i}^{\pm})[X_{l}]_{\mathrm{tot}}}{\kappa_{l}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}}}}, \displaystyle e^{\pm}_{ij}=\frac{{k_{\mathrm{cpl}}}\frac{\kappa^{\pm}_{j}[X_{j}]_{\mathrm{tot}}}{\kappa_{j}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}}}}{\kappa_{i}^{\pm}+{k_{\mathrm{cpl}}}\sum_{l\neq i}\frac{(\kappa_{l}^{\pm}-\kappa_{i}^{\pm})[X_{l}]_{\mathrm{tot}}}{\kappa_{l}+{k_{\mathrm{cpl}}}{C_{\mathrm{tot}}}}}\;\;(j\neq i). Accordingly, \sum_{j}e^{\pm}_{ij}=1 . In the strong coupling limit, \displaystyle e^{\pm}_{ij}\to\tilde{e}_{l}:=\frac{[X_{j}]_{\mathrm{tot}}\kappa^{\pm}_{j}}{\sum_{l}[X_{l}]_{\mathrm{tot}}\kappa_{l}^{\pm}}. EPR and tradeoff. The EPR of all coupling reactions is given by \displaystyle\sigma_{\mathrm{cpl}}^{(n)}:=\; \displaystyle\sum_{i>j}\sigma_{(i,j)}^{\mathrm{st}} \displaystyle=\; \displaystyle\sum_{i>j}{k_{\mathrm{cpl}}}[X_{i}]_{\mathrm{tot}}[X_{j}]_{\mathrm{tot}}\frac{\Gamma_{i}^{\mathrm{st}}-\Gamma_{j}^{\mathrm{st}}}{(\Gamma_{i}^{\mathrm{st}}+1)(\Gamma_{j}^{\mathrm{st}}+1)}\log\frac{\Gamma_{i}^{\mathrm{st}}}{\Gamma_{j}^{\mathrm{st}}}. From ( 28 ), in the strong coupling limit, each \sigma_{(i,j)}^{\mathrm{st}} can be approximated as: \displaystyle\sigma_{(i,j)}^{\mathrm{st}} \displaystyle\simeq\frac{k_{\mathrm{cpl}}^{-1}[X_{i}]_{\mathrm{tot}}[X_{j}]_{\mathrm{tot}}}{(1+\Gamma_{0})^{2}\Gamma_{0}}\left(\frac{d\Gamma_{i}^{\mathrm{st}}}{dk_{\mathrm{cpl}}^{-1}}\big|_{k_{\mathrm{cpl}}^{-1}\to 0}-\frac{d\Gamma_{j}^{\mathrm{st}}}{dk_{\mathrm{cpl}}^{-1}}\big|_{k_{\mathrm{cpl}}^{-1}\to 0}\right)^{2} \displaystyle=\frac{[X_{i}]_{\mathrm{tot}}[X_{j}]_{\mathrm{tot}}}{k_{\mathrm{cpl}}(\sum_{l}[X_{l}]_{\mathrm{tot}}\kappa_{l}^{+})(\sum_{l}[X_{l}]_{\mathrm{tot}}\kappa_{l}^{-})}\times \displaystyle\hskip 9.24994pt\left(\frac{\sum_{l}[X_{l}]_{\mathrm{tot}}\left((\kappa_{i}^{+}-\kappa_{j}^{+})\kappa_{l}^{-}-(\kappa_{i}^{-}-\kappa_{j}^{-})\kappa_{l}^{+}\right)}{{C_{\mathrm{tot}}}}\right)^{2}. That is, \displaystyle\sigma_{(i,j)}^{\mathrm{st}}\simeq\frac{\left((\kappa_{i}^{+}-\kappa_{j}^{+})\overline{\kappa^{-}}-(\kappa_{i}^{-}-\kappa_{j}^{-})\overline{\kappa^{+}}\right)^{2}}{{k_{\mathrm{cpl}}}\sqrt{\kappa_{i}^{+}\kappa_{i}^{-}}\sqrt{\kappa_{j}^{+}\kappa_{j}^{-}}}\sqrt{\tilde{e}^{+}_{i}\tilde{e}^{-}_{i}}\sqrt{\tilde{e}^{+}_{j}\tilde{e}^{-}_{j}}, (29) where \overline{\kappa^{\pm}}:=(\sum_{l}[X_{l}]_{\mathrm{tot}}\kappa_{l}^{\pm})/(\sum_{l}[X_{l}]_{\mathrm{tot}}) is the averaged effective rate constant of all driving reactions. Therefore, a tradeoff between the EPR \sigma_{\mathrm{cpl}}^{(n)} and elasticity in the strong coupling limit (analogous to Eq. ( 16 )) still holds, although it is biased by the driving reactions for currency metabolites other than X_{i} and X_{j} . Noting the constraint \sum_{i}\tilde{e}^{\pm}_{i}=1 , ( 29 ) implies that the more uniformly large the values of \tilde{e}^{\pm}_{i} are, the greater the total EPR contributed by all coupling reactions: i.e., an intrinsic tradeoff between global elasticity and thermodynamic cost. The above results are consistent with the case of n=2 in the main text. Table 2: Order estimation in the case of ATP–GTP coupling. Symbol Description Estimated value Refs. {[\mathrm{ATP}]}_{\mathrm{tot}} Pool size of ATP 1\text{--}5 [mM] Danchin et al. 1984 ; Jewett et al. 2009 ; Deng et al. 2021 ; Albe et al. 1990 {[\mathrm{GTP}]}_{\mathrm{tot}} Pool size of GTP 0.1\text{--}1 [mM] Danchin et al. 1984 ; Jewett et al. 2009 ; Deng et al. 2021 {k_{\mathrm{cpl}}} Rate constant for the coupling reaction \sim 10^{4} [/M/s] Zala et al. 2017 ; Tokarska-Schlattner et al. 2008 ; Pollack et al. 2002 \kappa^{\pm}_{\rm ATP} Effective rate constant for the driving reaction {\rm ADP}\rightleftharpoons{\rm ATP} \sim 10^{1} [/s] Deng et al. 2021 ; Albe et al. 1990 \Gamma_{\rm ATP} {[\mathrm{ATP}]}/[{\rm ADP}] ratio 1\text{--}10^{2} Traut 1994 ; Tantama et al. 2013 \Gamma_{\rm GTP} {[\mathrm{GTP}]}/[{\rm GDP}] ratio 1\text{--}10^{2} Colombo et al. 1998 ; Traut 1994 r {[\mathrm{ATP}]}_{\mathrm{tot}}/{[\mathrm{GTP}]}_{\mathrm{tot}} ratio 1\,\text{--}\;{>\,}10 Danchin et al. 1984 Appendix E Order estimation of model parameter values It is thought that cells generally maintain substantial levels of enzymes responsible for metabolic currency coupling. Nucleoside diphosphate kinase (NDPK), for example, is considered a housekeeping enzyme that is constitutively expressed in many types of cells to maintain nucleotide homeostasis Yu et al. 2017 . To evaluate the biological relevance of our theory, particularly the controllability–EPR tradeoff relationship, we estimate typical values of relevant parameters and variables, using ATP–GTP coupling as an example (summarized in Table 2 ). Effective rate constant of driving reaction. The total ATP synthesis flux in a microbe is estimated as Deng et al. 2021 \displaystyle\kappa^{+}_{\rm ATP}{[\mathrm{ADP}]}\simeq 10^{-2}\;{\rm[M/s]}, and the intracellular ADP concentration in E. coli is measured as Albe et al. 1990 \displaystyle{[\mathrm{ADP}]}\simeq 10^{-3}\;{\rm[M]}. Therefore, the effective rate constant for the driving reaction, {\rm ADP\leftrightarrow ATP} , should be estimated as \displaystyle\kappa^{+}_{\rm ATP}\simeq 10^{1}\;{\rm[/s]}. Given that {[\mathrm{ATP}]}_{\mathrm{tot}}{\,\simeq\,}3\times 10^{-3}\;{\rm[M]}~\cite[cite]{\@@bibref{Authors Phrase1YearPhrase2}{albe1990cellular}{\@@citephrase{(}}{\@@citephrase{)}}} , \displaystyle\kappa^{+}_{\rm ATP}/{[\mathrm{ATP}]}_{\mathrm{tot}}\simeq 10^{3}\text{--}10^{4}\;{\rm[/M/s]}. (30) Coupling constant {k_{\mathrm{cpl}}} in ping-pong mechanism. NDPK catalyzes the following reactions: \displaystyle{\rm NTP}+{\rm E}\overset{k^{+}_{N}}{\underset{k^{-}_{N}}{\rightleftharpoons}}{\rm NDP}+{\rm EP}\quad{\rm(N=A,G)}, (31) in the case of ATP–GTP coupling. Through this ping-pong mechanism, the charged/uncharged ratios, \Gamma_{\rm ATP} and \Gamma_{\rm GTP} , are driven toward equilibrium, providing an effective coupling between them. Assuming that reaction ( 31 ) follows mass action kinetics, the steady-state concentrations of E and EP are calculated as \displaystyle{\rm[E]}^{\mathrm{st}} \displaystyle= \displaystyle\frac{k_{A}{[\mathrm{ADP}]}+k_{G}{[\mathrm{GDP}]}}{k_{A}{[\mathrm{ATP}]}_{\mathrm{tot}}+k_{G}{[\mathrm{GTP}]}_{\mathrm{tot}}}{E_{\mathrm{tot}}}, \displaystyle{\rm[EP]}^{\mathrm{st}} \displaystyle= \displaystyle\frac{k_{A}{[\mathrm{ATP}]}+k_{G}{[\mathrm{GTP}]}}{k_{A}{[\mathrm{ATP}]}_{\mathrm{tot}}+k_{G}{[\mathrm{GTP}]}_{\mathrm{tot}}}{E_{\mathrm{tot}}}, where {E_{\mathrm{tot}}} denotes the total amount of NDPK, and we assume k_{N}{\,:=\,}k^{+}_{N}{\,=\,}k^{-}_{N} . Accordingly, we obtain the following dynamics of {{[\mathrm{ATP}]}} : \displaystyle\frac{d}{dt}{{[\mathrm{ATP}]}}=\; \displaystyle k_{A}{[\mathrm{ADP}]}{\rm[EP]}^{\mathrm{st}}-k_{A}{[\mathrm{ATP}]}{\rm[E]}^{\mathrm{st}} \displaystyle+\kappa_{A}^{+}{[\mathrm{ADP}]}-\kappa_{A}^{-}{[\mathrm{ATP}]} \displaystyle=\; \displaystyle k_{A}{E_{\mathrm{tot}}}\frac{k_{G}{[\mathrm{ADP}]}{[\mathrm{GTP}]}-k_{G}{[\mathrm{ATP}]}{[\mathrm{GDP}]}}{k_{A}{[\mathrm{ATP}]}_{\mathrm{tot}}+k_{G}{[\mathrm{GTP}]}_{\mathrm{tot}}} \displaystyle+\kappa_{A}^{+}{[\mathrm{ADP}]}-\kappa_{A}^{-}{[\mathrm{ATP}]}. It follows \displaystyle{k_{\mathrm{cpl}}}=\frac{k_{A}k_{G}{E_{\mathrm{tot}}}}{k_{A}{[\mathrm{ATP}]}_{\mathrm{tot}}+k_{G}{[\mathrm{GTP}]}_{\mathrm{tot}}}. (32) From refs. Zala et al. 2017 ; Tokarska-Schlattner et al. 2008 ; Pollack et al. 2002 , \displaystyle k_{\rm cat} \displaystyle\simeq \displaystyle 10^{3}\;[{\rm/s}], \displaystyle K_{\rm M,ATP} \displaystyle\simeq \displaystyle 10^{-4}\text{--}10^{-3}\;[{\rm M}], \displaystyle K_{\rm M,GDP} \displaystyle\simeq \displaystyle 10^{-4}\;[{\rm M}], \displaystyle{E_{\mathrm{tot}}} \displaystyle\simeq \displaystyle 10^{-6}\;[{\rm M}], and thus, \displaystyle k_{N}\simeq k_{\rm cat}/K_{\rm M,N}\simeq 10^{6}\text{--}10^{7}{\rm[/M/s]}. (33) This is consistent with ref. Schaertl et al. 1998 , which states that the second order rate constants k_{N} for phosphorylation by natural NTPs are between 0.7\text{--}13\times 10^{6}\;{\rm[M^{-1}s^{-1}]} . Finally, from ( 32 ), \displaystyle{k_{\mathrm{cpl}}}\simeq 10^{3}\text{--}10^{4}{\rm[/M/s]}. (34) From Eqs. ( 30 ) and ( 34 ), {k_{\mathrm{cpl}}} and \kappa^{+}_{\rm ATP}/{[\mathrm{ATP}]}_{\mathrm{tot}} are typically in the same order of magnitude, while they depend largely on the amount of enzymes and the growth conditions. This order-of-magnitude estimation suggests that {k_{\mathrm{cpl}}} and \kappa_{\rm ATP}^{\pm}/[{\rm ATP}]_{\mathrm{tot}} are of comparable magnitude with typical NDPK concentrations, implying that both strong ( {k_{\mathrm{cpl}}}\,{\gg}\,\kappa_{\rm ATP}^{\pm}/[{\rm ATP}]_{\mathrm{tot}} ) and weak ( {k_{\mathrm{cpl}}}\,{\ll}\,\kappa_{\rm ATP}^{\pm}/[{\rm ATP}]_{\mathrm{tot}} ) currency coupling may occur in natural microbial systems. In particular, under nutrient-limited conditions, the strong coupling limit will be more likely.