From toxic waste to beneficial nutrient: acetate boosts E. coli growth at low glycolytic flux

Acetate, a major by-product of glycolytic metabolism in Escherichia coli and many other microorganisms, has long been considered a toxic waste compound that inhibits microbial growth. This counterproductive auto-inhibition represents a major problem in biotechnology and has puzzled the scientific community for decades. Recent studies have revealed that acetate is also a co-substrate of glycolytic nutrients and a global regulator of E. coli metabolism and physiology. Here, we used a systems biology strategy to investigate the mutual regulation of glycolytic and acetate metabolism. Computational and experimental results demonstrate that reducing the glycolytic flux enhances co-utilization of acetate with glucose. Acetate metabolism thus compensates for the reduction in glycolytic flux and eventually buffers carbon uptake so that acetate, far from being toxic, actually enhances E. coli growth under these conditions. We validated this mechanism using three orthogonal strategies: chemical inhibition of glucose uptake, glycolytic mutant strains, and alternative substrates with a natively low glycolytic flux. Acetate makes E. coli more robust to glycolytic perturbations and is a valuable nutrient, with a beneficial effect on microbial growth.


Introduction
Metabolism is the fundamental biochemical process that converts nutrients into energy and cellular building blocks. The preferred nutrients of many organisms are glycolytic substrates, such as glucose, which are oxidized into pyruvate by glycolysis. In the presence of oxygen, pyruvate can then be fully oxidized into carbon dioxide and water via respiratory pathways involving the tricarboxylic acid (TCA) cycle and oxidative phosphorylation. When oxygen is limiting, pyruvate is incompletely oxidized through non-respiratory pathways, and fermentation by-products are released into the environment. Despite fermentation being less efficient than respiration, fermentation by-products are still released in the presence of oxygen by most microorganisms, including bacteria (Harden, 1901), yeasts (Pasteur, 1857) and mammalian cells (Warburg et al, 1927).
Acetate is one of the most common by-products of glycolytic metabolism in Escherichia coli and many other microorganisms (Harden, 1901;Kutscha & Pflügl, 2020;Trcek et al, 2015). The established view is that acetate and other fermentation by-products are toxic and inhibit microbial growth (Beck et al, 2022;Enjalbert et al, 2017;Hollowell & Wolin, 1965;Kutscha & Pflügl, 2020;Luli & Strohl, 1990;Pinhal et al, 2019;Roe et al, 1998;Roe et al, 2002;Russel, 1992;Trcek et al., 2015;Warnecke & Gill, 2005). Why E. coli should produce a self-toxic byproduct is an intriguing question, particularly since E. coli's natural environment, the gut, is acetate-rich. Several explanations for the toxic effect of acetate have been proposed, the classical one being the uncoupling effect of organic acids (Russel, 1992), with acetic acid freely diffusing through the membrane before dissociating intracellularly into acetate and a proton due to its low pKa, and the energy required to expel the excess protons from the cell to maintain pH homeostasis being detrimental to growth. Another hypothesis is that the presence of acetate in the cytoplasm may affect the anion balance of the cell (Roe et al., 1998), reducing the pool of other anions -primarily glutamate -available to maintain osmotic pressure and electroneutrality and thereby impeding cell function. Acetate has also been reported to inhibit a step in the methionine biosynthesis pathway (Roe et al., 2002), reducing the intracellular methionine pool and concomitantly increasing the concentration of the toxic intermediate homocysteine. Finally, acetate may modulate the accumulation of acetylphosphate (Weinert et al, 2013;Wolfe et al, 2003), an intermediate of the acetate pathway and a signaling metabolite that can phosphorylate and acetylate enzymes and regulatory proteins with broad physiological consequences. However, none of these studies conclusively explains how acetate inhibits microbial growth (Pinhal et al., 2019). The physiological role of fermentation under aerobic conditions also remains unclear, especially since it leads to the accumulation of toxic by-products at the expense of biomass production.
Acetate overflow and its impact on E. coli are also important issues in biotechnology, where acetate is both produced from glycolytic carbon sources and found in many renewable resources (Klinke et al, 2004;Kutscha & Pflügl, 2020;Palmqvist & Hahn-Hägerdal, 2000;Sun et al, 2021;Trcek et al., 2015). As well as inhibiting growth, acetate production diverts carbon that could otherwise be used to synthesize biomass or valuable compounds. Acetate therefore decreases the productivity of bioprocesses. Since acetate is only produced at high glucose uptake rates (Nanchen et al, 2006;Renilla et al, 2012;Valgepea et al, 2010), a common approach to reduce acetate overflow is to limit glucose uptake by process engineering (e.g. glucose-limited chemostat or fed-batch culture) or by metabolic engineering (e.g. deletion of glucose transporters) (Bernal et al, 2016;De Mey et al, 2007;Eiteman & Altman, 2006;Fuentes et al, 2013;Kutscha & Pflügl, 2020;Luli & Strohl, 1990;Peebo et al, 2014). The utilization of acetate in combination with glycolytic substrates could alleviate growth inhibition and allow acetate to act as both a renewable carbon source and a relevant (co-)substrate for bioproducts that are derived from acetyl-CoA (Krivoruchko et al, 2015;Kutscha & Pflügl, 2020;Peebo et al., 2014;Wei et al, 2013). However, despite decades of attempts to reduce its accumulation and improve its utilization, acetate is still considered a major problem in biotechnology (Bernal et al., 2016;De Mey et al., 2007;Luli & Strohl, 1990;Pinhal et al., 2019).
Recently, we found that acetate can also act as a co-substrate for glycolytic carbon sources such as glucose, fucose and gluconate (Enjalbert et al., 2017), and is therefore more than just a waste by-product. The switch between acetate production and consumption was found to be determined thermodynamically by the acetate concentration itself. Acetate is also a global regulator of E. coli metabolism and physiology (Millard et al, 2021). These observations suggest an alternative rationale for acetate toxicity, namely that the inhibition of growth and glucose uptake may result from coordinated downregulation of the expression of most central metabolic genes (glucose uptake systems, glycolysis, TCA cycle) and ribosomal genes. To date, the role of acetate as a co-substrate and regulator of glycolytic metabolism has only been investigated under high glycolytic flux, and little is known about the response of E. coli to acetate at low glycolytic fluxes, conditions that are nevertheless commonly experienced by this bacterium in its natural, industrial and laboratory environments (Cummings & Englyst, 1987;Cummings et al, 2001;de Graaf et al, 2010;Fabich et al, 2008;Jones et al, 2007;Kleman & Strohl, 1994;Macfarlane et al, 1992;Shen et al, 2011;Wolfe, 2005).
In this study, we took a systems biology perspective to clarify the interplay between glycolytic and acetate metabolism in E. coli. We used a kinetic model of E. coli metabolism to generate hypotheses on the relationships between glycolytic, acetate and growth fluxes. We validated model predictions experimentally, explored the underlying biochemical and regulatory mechanisms on glucose, and extended our findings to other glycolytic nutrients.

Model-driven hypotheses on the interplay between glucose and acetate metabolism in
Escherichia coli. We explored the functional relationship between glycolytic and acetate metabolism in Escherichia coli using a coarse-grained kinetic model coupling glucose and acetate metabolism with growth ( Figure 1A) (Millard et al., 2021). The model included two compartments, six compounds, and six reactions representing (i) glucose uptake and its conversion into acetyl-CoA by glycolysis, (ii) acetyl-CoA utilization in the TCA cycle and anabolic growth pathways, and (iii) the reversible conversion of acetyl-CoA into acetate via phosphotransacetylase (Pta), acetate kinase (AckA), and acetate exchange between the cell and its environment. Acetate utilization via acetylCoA synthetase (Acs), which is not active when glucose is present (Enjalbert et al., 2017), was not included. The model also included two inhibitory terms representing on the one hand inhibition of glucose uptake and glycolysis, and on the other inhibition of the TCA cycle and growth, in line with the observed inhibition of catabolic and anabolic processes (Millard et al., 2021). This model has been extensively validated and shown to have good predictive capabilities (Millard et al., 2021). We used it to simulate the response of glycolytic, acetate and growth fluxes to different degrees of glycolytic limitation by progressively reducing the activity (i.e. Vmax) of the glycolytic pathway -and thereby the glycolytic flux -from 100 % to 20 % of its initial value. Since glycolytic and acetate fluxes are both determined by the concentration of acetate (Enjalbert et al., 2017;Millard et al., 2021;Pinhal et al., 2019), these simulations were carried out over a broad range of acetate concentrations (from 0.1 to 100 mM).
The results obtained ( Figure 1B-D) highlight the strong bidirectional interplay between glycolytic and acetate metabolism, which eventually modulate growth. Glucose uptake and acetate flux both respond nonlinearly to perturbations ( Figure 1B-C). According to the model, increasing the acetate concentration inhibits glycolysis ( Figure 1B) and reverses the acetate flux ( Figure 1C) over the full range of glycolytic fluxes, as previously reported at the highest glucose uptake rate (Millard et al., 2021). The acetate concentration at which the flux reversal occurs appears to depend on the glycolytic flux ( Figure 1C). Reducing the glycolytic flux lowers acetate production when acetate is a by-product of glucose metabolism, but seems to enhance acetate utilization when acetate is a co-substrate of glucose ( Figure 1C). Finally, the glycolytic flux and the acetate concentration both have a strong, nonlinear effect on the growth rate ( Figure 1D). At maximal glycolytic activity, the predictions of the model are consistent with the established deleterious effect of acetate on microbial growth ( Figure 1D). However, despite the dual inhibition of catabolism and anabolism by acetate included in the model, it also predicts that at low glycolytic flux, increasing the acetate concentration should boost E. coli growth ( Figure 1D). A global sensitivity analysis showed that the predictions of the model are robust to parameter uncertainty. We present these model-driven hypotheses in detail (including the sensitivity analysis results) in the following sections along with experimental results collected for validation. Note that none of the data obtained in this study were used to build or calibrate the model and therefore represent independent tests of its predictions and inferred hypotheses. A Representation (in Systems Biology Graphical Notation format, http://sbgn.org) of the glucose and acetate metabolism of Escherichia coli, as implemented in the kinetic model used in the study (reproduced from (Millard et al., 2021)).
B-D Glucose uptake flux (B), acetate flux (C) and growth rate (D) simulated over a broad range of glycolytic activity levels (from 20 to 100 % of the initial Vmax) and acetate concentrations (from 0.1 to 100 mM).
Enhanced acetate utilization at low glycolytic flux buffers carbon uptake. We investigated the glycolytic control of E. coli metabolism by confronting the model predictions with experimental data (Figures 1 and 2). The simulations show that, at low acetate concentrations (0 mM in Figure 2A-D), acetate production is predicted to increase with the glycolytic flux, in line with the relationship observed in glucose-limited chemostat experiments (Nanchen et al., 2006;Renilla et al., 2012;Valgepea et al., 2010) and in a set of mutants with different levels of glucose uptake (Fuentes et al., 2013). In contrast, when the acetate concentration is high (30 mM in Figure 2A-D), such that it is co-consumed with glucose, acetate utilization is expected to increase when the glycolytic flux decreases. This response implies, as suggested previously (Millard et al., 2021) but never verified experimentally, that the control exerted by glycolysis on the acetate flux is non-monotonic.
To test this hypothesis, we measured growth, glucose uptake, and acetate fluxes in E. coli K-12 MG1655 grown on 15 mM glucose and increased the concentration of the glycolytic inhibitor methyl α -D-glucopyranoside (αMG, an analogue of glucose that is taken up and phosphorylated but not metabolized by glycolysis) (Bren et al, 2016;Negrete et al, 2010) ( Figure 2E-H). Experiments were carried out without acetate or with 30 mM acetate for comparison with the predictions shown in Figure 2A-D. Results confirm that α MG inhibits glucose uptake under both conditions ( Figure 2E). In the absence of acetate, the acetate (production) flux decreased when the α MG concentration was increased ( Figure 2F), as expected based on the positive control exerted by glycolysis on the acetate flux. In contrast, in the presence of 30 mM acetate, increasing the α MG concentration increased the acetate uptake flux ( Figure 2F). These results are consistent with the predictions of the model ( Figure  2A-D) and demonstrate that glycolysis controls the acetate flux in a non-monotonic manner. The control exerted by glycolysis on acetate metabolism thus depends on the concentration of acetate itself.
The growth and total carbon uptake rates decreased when the α MG concentration was increased in the absence of acetate ( Figure 2G-H), but were more stable in the presence of acetate despite a ~2-fold reduction in glucose uptake (between 5.7 and 3.3 mmol·g DW −1 ·h −1 with 30 mM acetate; Figure 2E). Indeed, both the total carbon uptake and the growth rates were less sensitive to glycolytic perturbations in the presence of acetate than in the absence of acetate ( Figure 2I-J), suggesting that acetate compensates for the drop in glucose uptake. Consistently, a strong, nonlinear relationship between glucose uptake and acetate flux was observed in all experiments ( Figure 2K), which was reflected in the contribution of acetate to the total carbon uptake of E. coli ( Figure 2L). At the maximal glucose uptake rate, acetate production represented a 10 % loss in carbon. However, when glucose uptake was below 7 mmol·g DW −1 ·h −1 , acetate was co-utilized with glucose and contributed positively to the carbon balance (up to 50 % at a glucose uptake rate of 3.3 mmol·g DW −1 ·h −1 ; Figure 2L). These results highlight the coupling between glycolytic and acetate fluxes in E. coli, with acetate buffering carbon uptake at low glycolytic flux, thereby improving the robustness of E. coli to glycolytic perturbations.

Figure 2. Glycolytic control of E. coli metabolism.
A-D Predicted glucose uptake flux (A), acetate flux (B), growth rate (C), and total carbon uptake flux (D) of E. coli on glucose (15 mM) as a function of the glycolytic activity (20-100 % of the maximal activity), at low (0 mM) and high (30 mM) acetate concentrations.
E-H Experimental glucose uptake flux (E), acetate flux (F), and growth rate (G) of E. coli grown on glucose (15 mM) with different concentrations of α MG (0-100 mM), at low (0 mM) and high (30 mM) acetate concentrations. These data were used to quantify the total carbon uptake flux as a function of the α MG concentration (H). Each data point represents an independent biological replicate, and lines represent the best polynomial fits.
I, J Total carbon uptake (I) and growth rate (J) as a function of the glucose uptake flux, at low (0 mM) and high (30 mM) acetate concentrations. Each data point represents an independent biological replicate, and lines represent the best linear fits. K, L Relationships between the acetate flux and glucose uptake (K), and between the contribution of acetate to carbon uptake and glucose uptake (L). Each data point represents an independent biological replicate, and lines represent the best polynomial fits.
The glycolytic flux determines the acetate concentration threshold at which the acetate flux reverses. As already reported at maximal glucose uptake rates (Enjalbert et al., 2017;Millard et al., 2021), the acetate concentration threshold at which the acetate flux switches between production and consumption is about 10 mM. However, this threshold was predicted to depend on the glycolytic flux ( Figures 1C and 3A-B). According to the model, the switch should occur at an acetate concentration of 3.8 mM when the glycolytic activity is 60 % of its maximum value, and at a concentration of just 0.6 mM when the glycolytic activity is reduced to 20 % of the maximum. This suggests that the glycolytic flux determines the role of acetate as a co-substrate or by-product of glycolytic metabolism.
We tested this hypothesis by measuring acetate fluxes in E. coli grown on glucose (15 mM) plus acetate (from 0.1 to 100 mM) at different concentrations of α MG (0, 20, 40 or 100 mM, Figure 3C). As previously observed at full glycolytic activity (i.e., 0 mM α MG) (Beck et al., 2022;Enjalbert et al., 2017;Millard et al., 2021;Pinhal et al., 2019), acetate production decreased when the acetate concentration was increased ( Figure 3C), and the acetate flux reversed at a concentration of ~6 mM ( Figure 3C-D), resulting in co-consumption of glucose and acetate. This threshold concentration gradually decreased when the glycolytic flux was reduced (from 2 mM acetate at 20 mM α MG to below 0.1 mM acetate at 40 or 100 mM α MG). Here again, the experimental profiles are qualitatively similar to the simulated profiles, albeit with quantitative differences. These results demonstrate that the acetate concentration threshold at which the acetate flux reverses depends on the glycolytic flux. It therefore follows that when the glycolytic flux is limited, acetate is utilized as a co-substrate at lower acetate concentrations. These results also confirm that inhibiting glycolysis enhances acetate utilization over the full range of acetate concentrations tested here, further validating the unusual control patterns identified above, and generalizing their functional relevance for acetate concentrations spanning three orders of magnitude.  (Enjalbert et al., 2017). Each data point represents an independent biological replicate, and the lines represent best fits using a logistic function. D Experimental acetate concentration threshold at which the acetate flux reverses as a function of the α MG concentration, determined from the best fits shown in panel (C).
Acetate enhances E. coli growth at low glycolytic flux. Acetate is considered toxic for microbial cells, but the model predicts this might not always be the case. At full glycolytic activity, the model predicts that maximal growth should occur at low acetate concentrations, with the growth rate remaining above 95 % of the maximum only at acetate concentrations below 12 mM ( Figures 1D and 4A). In agreement with the reported toxicity of acetate, the growth rate then monotonously decreases for acetate concentrations above 12 mM. However, when the glycolytic flux is reduced, maximal growth is predicted to occur at high acetate concentrations ( Figures 1D and 4A). For instance, when the glycolytic activity is reduced to 20 %, the growth rate is expected to be maximal at an acetate concentration of 28 mM, and should remain above 95 % of the maximum over a broad range of acetate concentrations (between 12 and 86 mM). According to the model therefore, acetate should enhance E. coli growth at low glycolytic flux.
To test this prediction, we measured the growth rate of E. coli on glucose (15 mM) with acetate (0.1 to 100 mM) in the presence of 100 mM α MG to inhibit glycolysis ( Figure 4B). The data measured in the absence of α MG are shown for comparison. At maximal glycolytic flux, the growth rate decreased monotonically with the acetate concentration, but this relationship became non-monotonic when glycolysis was inhibited, which is qualitatively consistent with model predictions. In the presence of α MG, the growth rate was lower in the absence of acetate than at high acetate concentrations (0.28 ± 0.02 h −1 with acetate < 0.5 mM versus 0.34 ± 0.01 h −1 with 60 mM acetate, p-value = 8.10 -3 ). As predicted by the model therefore, acetate appears to be beneficial rather than toxic to E. coli when the glycolytic flux is reduced. The growth rate was slightly lower at 100 mM acetate (0.30 ± 0.03 h −1 ) than at 60 mM acetate, but the similar growth rates observed across the concentration range demonstrate that acetate is not toxic when the glycolytic flux is low, even at high concentrations.
As an orthogonal, alternative strategy to the use of a glycolytic inhibitor, we tested the impact of acetate on the growth of mutant strains with low glycolytic flux, namely the E. coli K-12 BW25113 Δ pgi and Δ pfkA strains, where the glycolytic flux is 24 and 29 % that of the wildtype strain, respectively (Long & Antoniewicz, 2019). Their growth rates were measured on 15 mM glucose, with or without 10 mM acetate and the effect of acetate was quantified as the relative change in growth rate. The wild-type isogenic strain was grown in the absence and presence of α MG (100 mM) as a control. As expected, acetate reduced the growth rate of the wild-type strain in the absence of α MG (−6 ± 1 %, p-value = 4.10 -5 ) but increased its growth in the presence of 100 mM α MG (+9 ± 2 %, p-value = 3.10 -5 ) ( Figure 4C). The beneficial effect of acetate on growth was also observed in the Δ pgi and Δ pfkA strains ( Figure 4D), whose growth rate strongly increased despite the absence of α MG (Δpgi : +60 ± 19 %, p-value = 3.10 -2 ; Δ pfkA : +117 ± 21 %, p-value = 1.10 -2 ). These results confirm that the beneficial effect of acetate is not due to the presence of α MG but is a consequence of the decrease in glycolytic flux.
Overall, all results are in line with the predictions of the model and demonstrate that acetate enhances E. coli growth when the glycolytic flux is reduced by either chemical or genetic perturbations.  (Millard et al., 2021) that the apparent "toxicity" of acetate is due to its inhibition of most metabolic genes -including those for the glucose phosphotransferase system (PTS), glycolysis and TCA cycle. Acetate also increases the expression of genes related to stress responses and downregulates genes related to transport, energy biosynthesis and mobility for instance. To determine if acetate has a similar regulatory effect at low glycolytic flux, we compared the transcriptomes of E. coli grown on glucose (15 mM) plus acetate (0, 10, 50 or 100 mM) in the presence of α MG (100 mM), with those measured in the absence of α MG at the same acetate concentrations (Millard et al., 2021).
We observed that contrary to what occurs at maximal glycolytic flux (0 mM α MG) acetate did not trigger a stress response at low glycolytic flux (100 mM α MG) ( Figure 5A). Conversely, acetate increased gene expression for the respiratory chain and some biosynthetic processes (ribonucleotide monophosphate biosynthesis, amino acid production). Other hallmarks of the effect of acetate at maximal glycolytic flux such as the inhibition of mobility-related gene expression were also not observed at low glycolytic flux. The activation of glutamate catabolism, the repression of transport related genes, and the repression of carbohydrate transport and catabolism were among the few similarities between the responses to acetate at high and low glycolytic flux.
At the metabolic level, acetate-driven inhibition of the expression of central metabolic genes (PTS, TCA and glycolytic enzymes) and growth machinery (ribosomal proteins) was not observed at low glycolytic flux ( Figure 5B). In fact, expression of most of these genes (47/65) was slightly higher in the presence of 50 mM acetate and 100 mM α MG, whereas at maximal glycolytic flux it was lower at these acetate concentrations for a similar proportion (46/65). When the glycolytic flux was inhibited, ptsG expression, which controls glucose uptake (Rohwer et al, 2000), was 70 % higher at 10 and 50 mM acetate. The expression of most glycolytic genes was 15 to 110 % higher at 50 mM acetate, while the expression of TCA cycle genes remained stable. Regarding acetate metabolism, the expression of pta and ackA was slightly higher at 10 and 50 mM acetate, while alternative acetate utilization via Acs was inhibited by acetate (5-fold reduction in expression at 100 mM). Finally, expression of genes encoding ribosomal proteins (rpl, rpm and rps operons) was increased by 57 % on average at 50 mM acetate and 48 % on average at 100 mM acetate (Dataset S1).
These results indicate that the stress response and lower metabolic activity triggered by acetate at high glycolytic flux do not occur at low glycolytic flux. In this situation, acetate enhances the expression of ribosomal and central metabolic genes, in line with its positive effect on growth. This is the opposite of what is observed at high glycolytic flux, demonstrating that the regulatory role of acetate is largely determined by the glycolytic flux. Acetate-enhanced growth is supported by the Pta-AckA pathway. According to the model ( Figure 1A), the reversible Pta-AckA pathway (acetate consumption or production) explains by itself how acetate promotes E. coli growth, without the need for Acs (acetate consumption only). To confirm this hypothesis, we measured the growth rates of E. coli K-12 BW25113 wild-type and of Δ pta and Δ acs isogenic mutant strains cultivated on 15 mM glucose plus 100 mM α MG, with or without 10 mM acetate. The effect of acetate was quantified as the relative change in growth rate induced. Acetate had a similar, beneficial effect on the growth of the wild-type strain (+9 ± 2 %, p-value = 3.10 -5 ) and of the Δ acs strain (+8 ± 1 %, p-value = 1.10 -3 ), but not for the Δ pta strain, whose growth rate decreased slightly but significantly (−6 ± 1 %, p-value = 7.10 -3 ). These results confirm that the beneficial effect of acetate on E. coli growth at low glycolytic flux is mediated by the Pta-AckA pathway, with no involvement of Acs in this phenomenon. Acetate boosts growth of E. coli on glycerol and galactose. Assuming that the beneficial impact of acetate on E. coli growth stems from the tight interplay of glycolytic and acetate fluxes, this should be independent of the nature of the carbon source, implying that acetate should boost E. coli growth on glycolytic substrates with low glycolytic flux, such as glycerol and galactose (Gerosa et al, 2015). And indeed, wild-type E. coli growth was enhanced on both glycerol (+7 ± 1 %, p-value = 1.10 -2 ) and galactose (+52 ± 1 %, p-value = 2.10 -4 ) in the presence of 10 mM acetate (Figure 7). Similar levels of enhancement were observed for the Δ acs strain (+8 ± 1 % on glycerol, p-value = 3.10 -4 ; +53 ± 5 % on galactose, p-value = 3.10 -3 ), but not for the Δ pta strain (−8 ± 2 % on glycerol, p-value = 1.10 -2 ; +2 ± 12 % on galactose, pvalue = 0.70). These results demonstrate that acetate promotes E. coli growth on galactose and glycerol. As observed for glucose, the growth enhancement is mediated exclusively by the Pta-AckA pathway, with no involvement of Acs. These results extend and generalize our findings to glycolytic carbon sources other than glucose and support our findings in more physiological conditions, i.e. without an inhibitor or deletion of metabolic genes.

Discussion
Following a systems biology approach, we used a kinetic model of E. coli metabolism to generate a comprehensive map of the functional relationship between glycolytic and acetate metabolisms and predict their combined impact on E. coli growth. In agreement with the predictions of the model, experimental results revealed a strong interplay between acetate and glycolytic fluxes in E. coli, with wide-ranging physiological consequences for the cells. Despite its established toxicity for microbial growth, we found that acetate improves the robustness of E. coli to glycolytic perturbations, and even boosts growth when the glycolytic flux is low.
Previous studies of the glycolytic control of acetate metabolism were carried out in the absence of acetate. In this situation, as observed here, glycolysis as a positive effect on acetate production. However, our results also reveal that in the presence of acetate, glycolysis negatively affects acetate utilization. Glycolytic control of the acetate flux is thus determined by the concentration of acetate itself. Because of this non-monotonous control pattern, reduced glycolytic flux increases acetate uptake and lowers the acetate concentration threshold at which acetate is co-consumed with glucose, with the overall effect of increasing acetate utilization by E. coli. This demonstrates that the metabolic status of acetate as a byproduct or co-substrate of glycolytic carbon sources is determined both by the concentration of acetate (Enjalbert et al., 2017;Millard et al., 2021) and the glycolytic flux, and thus by the availability of glycolytic nutrients. This mechanism is exclusively mediated by the Pta-AckA pathway and its role and activity are determined by the nutritional conditions experienced by E. coli. The Pta-AckA pathway can therefore be seen as a metabolic valve that integrates information on the nutritional environment of the cells to rapidly adjust the acetate flux depending on the availability of glycolytic substrates and acetate. Our results indicate that these properties, which emerge simply from the distribution of control throughout the network, buffers the total carbon uptake flux and makes E. coli more robust to changes in the nutritional environment.
When the glycolytic flux is low, enhanced acetate utilization promotes E. coli growth, demonstrating that acetate can be beneficial to E. coli, even at high concentrations. Whereas 60 mM acetate has been shown to inhibit growth by ~35 % at maximal glucose uptake rates (Enjalbert et al., 2017;Luli & Strohl, 1990;Pinhal et al., 2019), the same concentration of acetate had a positive effect on growth at low glucose uptake rates (+21 % in presence of 100 mM α MG). We confirmed the beneficial effect of acetate on E. coli growth using three orthogonal strategies to reduce the glycolytic flux: i) chemical inhibition of glucose uptake with α MG, ii) mutant strains deleted for glycolytic genes, and ii) glycolytic substrates with a naturally low glycolytic flux (glycerol and galactose). Since the impact of acetate on growth is partly determined by the glycolytic flux, this metabolic by-product is not toxic per se. Importantly, these results also weaken several hypotheses that have been proposed to explain the apparent "toxicity" of acetate for microbial growth. Indeed, perturbations of the proton gradient, of the anion composition of the cell, or of methionine biosynthesis, would have a monotonic impact on growth. The slight but global activating effect of acetate on the transcriptome at low glycolytic flux, which correlates with the growth response, also contrasts with its inhibitory effect at high glycolytic flux. The results of this study indicate that mechanisms that only explain growth inhibition at high glycolytic flux are inadequate. Likewise, the recent suggestion that gluconeogenic substrates such as acetate enhance E. coli growth on glycolytic substrates (as observed for mixtures of oxaloacetate with glycerol, xylose or fucose) (Okano et al, 2020) also fails to capture the non-monotonic growth response of E. coli to acetate. The regulatory effect of acetate on E. coli metabolism, which depends both on the acetate concentration and on the glycolytic flux, is therefore more complex than previously suggested (Millard et al., 2021) and should be investigated further.
Previous reports have shown that the "toxicity" of acetate for E. coli growth depends on the pH (Lawford & Rousseau, 1993;Orr et al, 2019;Pinhal et al., 2019). In our experiments, the pH was initially set to 7.0 but was not adjusted during growth. Nevertheless, the high buffer capacity of the M9 cultivation medium ensured that the pH remained stable in the exponential growth phase where growth rates and extracellular fluxes were measured, with changes of less than 0.3 pH unit in all experiments ( Figure S1). Therefore, the metabolic and physiological responses of E. coli to acetate highlighted in this study are not due to changes in pH. Another potential issue is that the effect of acetate might also depend on the concentration of salts (Bae & Lee, 2017;Lee et al, 2010). The effect of acetate added as sodium salt may differ from its effect when added as acetic acid. However, repeat experiments performed with acetic acid instead of sodium acetate showed that the beneficial effect of acetate on E. coli growth on glucose, galactose and glycerol was maintained ( Figure  S2).
The measured responses of E. coli to a broad range of perturbations of the glycolytic and acetate metabolisms were qualitatively consistent with model predictions, though quantitative differences were observed. These differences may stem from the data used to construct and calibrate the model, which were obtained exclusively at maximal glycolytic flux. Therefore, although it was not initially developed to address the questions of this study, the kinetic model used here successfully predicts the observed interplay between glycolytic and acetate metabolisms, as well as the beneficial effect of acetate on E. coli growth at low glycolytic flux. Since the model does not account for the activation by acetate of the expression of glucose uptake, glycolytic, and ribosomal genes, this suggests that transcriptional activation by acetate may not be crucial for the flux response of E. coli. The metabolic network may indeed directly sense and integrate the availability of glucose and acetate to coordinate glucose and acetate fluxes, as already reported for the coordination other metabolic processes (Doucette et al, 2011;Millard et al, 2017;Yuan et al, 2009). In keeping with this hypothesis, our results support the concept of acetate flux being thermodynamically controlled by acetate itself, as detailed in (Enjalbert et al., 2017), and we show that the thermodynamically-driven flux reversal is also controlled by the glycolytic flux. The reduction in glucose-derived acetyl-CoA is counterbalanced by an increased contribution of acetate, which sustains growth. Metabolic regulation is thus (at least partly) responsible for the interplay between acetate and glycolytic flux. The precise regulatory program driven by acetate remains to be identified.
Acetate concentrations in the gut range from 30 to 100 mM and glycolytic nutrients, such as glycerol and galactose (Fabich et al., 2008;Matamouros et al, 2018;Snider et al, 2009), are generally scarce. Given the established inhibition of microbial growth by acetate, these high acetate levels were considered detrimental to E. coli colonization. Our results indicate that acetate, far from being toxic, may in fact be beneficial to E. coli in this environment. Acetate enhances E. coli growth when glycolytic substrates are limiting and may therefore facilitate its colonization and survival in the gut. Moreover, acetate stabilizes growth when the glycolytic flux is perturbed by temporary nutrient shortages, which are frequent in the gut. We suggest that the underlying regulatory mechanism may therefore be optimal for E. coli in its natural environment. Our results also call for the role of acetate in the gut to be reconsidered, since it represents a valuable and abundant resource that may benefit E. coli and other microorganisms.
In biotechnology, developing efficient strategies to overcome the negative effects of acetate requires understanding what controls the acetate flux and how in turn acetate controls microbial metabolism. By addressing these two needs, our findings should immediately help design bioprocesses with reduced acetate overflow and accelerate the development of mixed feed processes in which acetate is utilized alongside glucose or other glycolytic substrates. Current strategies are mainly based on increasing Acs activity to (re)use acetate present in the feedstock, provided as a co-substrate or produced by the Pta-AckA pathway. Given the Pta-AckA pathway produces one ATP and Acs consumes two ATPs, strategies based on activating the Pta-AckA-Acs cycle are highly energy-consuming (Valgepea et al., 2010). Strategies involving the Pta-AckA pathway alone are inherently less costly for the cell. Our results show that the Pta-AckA pathway enables efficient co-utilization of acetate and glycolytic substrates (glucose, glycerol, galactose) in E. coli by itself, with a positive effect on growth. The comprehensive understanding of the relationship between glycolytic and acetate metabolisms provided by this study may also represents a valuable guide to increase the productivity of valuable products synthesized from acetate-derived AcCoA (Becker & Wittmann, 2015;Zhang et al, 2015).
E. coli is not the only microorganism that co-consumes glucose with fermentation byproducts. Saccharomyces cerevisiae can co-consume ethanol and glucose (Raamsdonk et al, 2001;Xiao et al, 2022), and mammalian cells co-consume lactate with glucose (Faubert et al, 2017;Hui et al, 2020;Hui et al, 2017;Rabinowitz & Enerbäck, 2020). Our findings may thus be generalizable to other microorganisms. Glycolytic control of fermentative pathways has indeed been observed, albeit poorly understood, in virtually all microorganisms. Our results explain why co-consumption of ethanol with glucose in yeast is observed at low glucose uptake (Raamsdonk et al., 2001). A similar relationship between glycolytic and lactate fluxes has also been reported in vitro in mammalian cells (Mulukutla et al, 2012), with the switch between lactate production and utilization being determined by the glycolytic flux. It has been suggested that lactate rather than glucose is the primary circulating energy source in the human body (Faubert et al., 2017;Rabinowitz & Enerbäck, 2020). Our study demonstrates that similarly to lactate, acetate is not a deleterious waste product but a valuable nutrient that can be shuttled between cells. The reversible Pta-AckA pathway enables uncoupling of glycolysis from the TCA cycle, indicating that this evolutionarily conserved design principle of eukaryotic metabolism (Xiao et al., 2022) is also present in bacteria. As a nutrient and global regulator of microbial metabolism, acetate plays a major role in cellular, microbial community and host-level carbon and energy homeostasis.

Materials and Methods
Strains. Escherichia coli K-12 MG1655 was chosen as the model wild-type strain. For experiments involving mutant strains, E. coli K-12 BW25113 was chosen as the reference wild-type strain, and its isogenic Δ pta, Δ pgi and Δ pfkA mutants were constructed from the KEIO collection (Baba et al, 2006) by removing the kanamycin cassette (Datsenko & Wanner, 2000). The Δ acs strain was constructed by CRISPR-Cas9 genome editing using the method and plasmids described in Wei et al. (Jiang et al, 2015). pta deletion was confirmed by genome sequencing and acs, pgi and pfkA deletions by PCR (with primers listed in Table S1).
Growth conditions. E. coli was grown in M9 minimal media (Millard et al, 2014) supplemented with 15 mM glucose, 15 mM galactose, or 30 mM glycerol. Sodium acetate (prepared as a 1.2 M solution at pH 7.0 by addition of potassium hydroxide), acetic acid (prepared as a 1.2 M solution at pH 7.0 by addition of ammonium hydroxide) and methyl α -Dglucopyranoside (αMG, prepared as a 1.2 M solution) were added up to the required concentrations. The cells were grown in shake flasks at 37 °C and 200 rpm, in 50 mL of medium. Growth was monitored by measuring the optical density (OD) at 600 nm using a Genesys 6 spectrophotometer (Thermo, USA). Biomass concentrations were determined using a conversion factor of 0.37 g DW /L/OD unit (Revelles et al, 2013).
Transcriptomics experiments. Cells were grown in M9 medium with 15 mM glucose, 100 mM α MG, and 0, 10, 50, or 100 mM sodium acetate. Sample preparations and transcriptomics analyses were carried out as described previously (Millard et al., 2021). Three independent replicates were analyzed for each condition. The transcriptomics data can be downloaded from the ArrayExpress database (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-11717. Transcriptomics data in the absence of glycolytic inhibition (ArrayExpress database accession number E-MTAB-9086 (Millard et al., 2021)) were measured under the exact same conditions but without α MG.

Sensitivity analyses.
We used a Monte-Carlo approach (Saa & Nielsen, 2017) to estimate the robustness of model predictions. Following the procedure detailed in (Millard et al., 2021), 500 simulated sets of calibration data were generated with noise added according to experimental standard deviations. For each of these artificially noisy data sets, we estimated model parameters and ran all simulations from the optimal parameters values. We calculated the mean value and standard deviation of each flux from the distribution of values obtained for the 500 sets of parameters.

Statistical analyses.
Comparisons between strains or conditions were performed using Student's t-tests with two-tailed distributions. The R notebook used to perform the calculations is provided at https://github.com/MetaSys-LISBP/glucose_acetate_interplay.

Data Availability
The data used to make the figures can be found in the Source Data files. The data, model and computer code produced in this study are available in the following databases: •