Escherichia coli has not only been the prime organism for developing new molecular biology methods but also for producing recombinant proteins, low molecular weight compounds etc. in industrial biotechnology for decades due to its low cost manufacturing and end-product purification and its ability to reach high cell densities grown aerobically [1, 2]. However, a major problem exists with aerobic E. coli cultivation on glucose at high growth rates-formation and accumulation of considerable amounts of acetic acid i.e. overflow metabolism. In addition to being detrimental for target product synthesis, accumulated acetate inhibits growth and diverts valuable carbon from biomass formation [3, 4].

The acetate synthesis and utilization pathways [5] can be seen in Figure 1: acetate can be synthesized by phosphotransacetylase (PTA)/acetate kinase (ACKA) and by pyruvate oxidase (POXB). Acetic acid can be metabolized to acetyl-CoA either by the PTA-ACKA pathway or by acetyl-CoA synthetase (ACS) through an intermediate acetyl-AMP. The high affinity (K_m of 200 μM for acetic acid) ACS scavenges acetate at low concentrations whereas the low affinity PTA-ACKA pathway (K_m of 7-10 mM) is activated in the presence of high acetate concentrations [6].

The phenomenon of overflow metabolism has been studied widely over the years and it is commonly believed to be caused by an imbalance between the fluxes of glucose uptake and those for energy production and biosynthesis [7, 8]. Several explanations such as the saturation of catalytic activities in the tricarboxylic acid (TCA) cycle [9, 10] and respiratory chain [7, 11, 12], energy generation [5, 13] or the necessity for coenzyme A replenishment [14] have been proposed. In addition to bioprocess level approaches [1, 15], various genetic modifications of the acetate synthesis pathways extensively reviewed in De Mey et al. [15] have been made to minimize acetic acid production. For instance, it has been shown that deleting the main acetate synthesis route PTA-ACKA results in a strong reduction (up to 80%) of acetate excretion, maximum growth rate (ca 20%) and elevated levels of formate and lactate (ca 30-fold) [4, 16–18], whereas poxB disruption causes reduction in biomass yield (ca 25%) and loss of aerobic growth efficiency of E. coli[19]. The latter indicates that acetate excretion cannot be simply excluded by disrupting its synthesis routes without encountering other unwanted effects. Unfortunately, no clear conclusions could be drawn from batch experiments with an acs knock-out strain [4]. It should be noted that studies with E. coli genetically modified strains engineered to diminish acetate production in batch cultures have not fully succeeded in avoiding acetate accumulation together with increasing target product production yields and rates [15]. Additionally, these studies have not allowed elucidating the mechanism of overflow metabolism unequivocally [4, 20, 21].

Acetate overflow is a growth rate dependent phenomenon, but no study has specifically focused on growth rate dependency of protein and gene expression regulation, intra-and extracellular metabolite levels using also metabolic modeling. Describing the physiology of an organism on several 'omic levels is the basis of systems biology that facilitates better understanding of metabolic regulation [22]. In this study, E. coli metabolism at proteomic, transcriptomic and metabolomic levels was investigated using continuous cultivation methods prior to and after overflow metabolism was switched on. Usually, chemostat cultures are used for steady state metabolism analysis, however, we applied two changestat cultivation techniques: accelerostat (A-stat) and dilution rate stat (D-stat), see Methods section for details [23, 24]. These cultivation methods were used as they provide three advantages over chemostat. Firstly, these changestat cultivation techniques precisely detect metabolically relevant switch points (e.g. start of overflow metabolism, maximum specific growth rate) and enable to monitor the dynamic patterns of several metabolic physiological responses simultaneously which could be left unnoticed using chemostat. Secondly, it is possible to collect vast amount of steady state comparable samples and by doing so, save time. Thirdly, both A-stat and D-stat enable to quantitatively study specific growth rate dependent co-utilization of growth substrates. Latter advantage was applied for investigating acetic acid consumption capability of E. coli at various dilution rates in this study. Combining changestat cultivation methods enables to study metabolism responses of the same genotype at different physiological states in detail without encountering the possible metabolic artifacts accompanied when using genetically modified strains.

Results obtained by studying specific growth rate dependent changes in E. coli proteome, transcriptome and metabolome in continuous cultures together with metabolic modeling allowed us to propose a new theory for acetate overflow: acetate excretion in E. coli is triggered by carbon catabolite repression mediated down-regulation of Acs resulting in decreased assimilation of acetate produced by Pta, and disruption of the PTA-ACS node.

To gain better insights into the regulation of acetate overflow metabolism in E. coli, we studied specific growth rate dependent proteomic, transcriptomic and metabolomic patterns combined with metabolic modeling using advanced continuous cultivation methods, which has not been carried out before. Continuous monitoring of the specific growth rate effect on E. coli metabolism enabled us to precisely detect important metabolic shift points, the most important being the start of acetate overflow at μ = 0.27 ± 0.02 h^-1 (Figure 2), and changing patterns of a number of important metabolites e.g. acetate, cAMP. Quite surprising was the down-regulation of the known acetate synthesis pathways, PTA-ACKA and POXB expression before overflow switch with increasing growth rate (Figure 1). A similar pattern has been seen before in chemostat cultures but without emphasizing the possible physiological consequences [26–28]. A 10-fold repression of the acetic acid utilization enzyme acetyl-CoA synthetase (Acs) expression was observed concurrently with the down-regulation of the PTA-ACKA pathway indicating that acetic acid synthesis might exceed its assimilation (Figure 1). Our two substrate A-stat and D-stat experiments directly proved that acetate consumption capability of E. coli is specific growth rate dependent as acetate consumption started to decrease at μ = 0.25 h^-1 (Figure S2 in Additional file 1) and acetate consumption capability decreased 12-fold around overflow switch growth rate μ = 0.27 ± 0.02 h^-1, respectively (Figure 5). In addition, it was shown that activation of carbon catabolite repression (CCR) and repression of Acs take place simultaneously prior to the start of overflow metabolism (Figure 1 Figure 4 and Figure 5). As a result, it is proposed that acetate overflow metabolism in E. coli is triggered by Acs down-regulation resulting in decreased assimilation of acetic acid produced by Pta, and disruption of the PTA-ACS node.

We showed that Acs was concurrently down-regulated five times more compared to the acetate synthesis pathways (Figure 1). In addition, the TCA cycle flux decrease as shown by change in CO₂ production at overflow switch growth rate indicates that carbon is not metabolized by the TCA cycle after the start of acetate accumulation with pre overflow switch rates (Figure 2 and Additional file 4). The latter is caused because the amount of acetyl-CoA entering the TCA cycle decreases after carbon is lost into excreted acetate. Stronger repression of the acetate consuming Acs in comparison with acetate synthesizing PTA-ACKA together with a decline in TCA cycle flux suggest disruption of acetic acid cycling at the PTA-ACS node (Figure 1). While this node may seem as a futile cycle, the fact is that numerous metabolic tasks involving the intermediate molecules of this cycle-acetyl phosphate (acetyl-P) and acetyl-AMP-are essential for proper E. coli growth (Figure 6). For instance, these molecules play a crucial role in bacterial chemotaxis regulation in which flagellar rotation is controlled by the activation level of the response regulator CheY [29] through either phosphotransfer from CheA [30, 31] or acetyl-P [31, 32], acetylation by acetyl-AMP [33, 34] or co-regulation of both mechanisms [29]. It has been also demonstrated that acetyl-P synthesis is vital for EnvZ-independent regulation of outer membrane porins [35], pathogenesis [36] and regulation of several virulence factors [5]. Furthermore, it has been presented that acetyl-P interacts with phosphate concentration regulators PhoB-PhoR [37] and NRI protein which is part of a complex nitrogen sensing system [38]. Acetyl-P is critical for efficient degradation of unfolded or damaged proteins by ATP-dependent proteases [39]. Altogether, acetyl-P can influence the regulation of almost 100 other genes [40]. Finally, pta and/or ackA mutations were shown to affect repair-deficient E. coli mutants [41] and a pta mutant has been reported to be impaired in its ability to survive during glucose starvation, while the ackA mutant survived as well as the parent strain [42]. It is important to note that the only known pathway in E. coli for acetyl-P synthesis is the PTA-ACKA [5, 31]. Taking all the previous into account, we conclude that acetyl-P as well as acetyl-AMP are essential for cellular growth of E. coli, and as acetic acid formation is the result of their dephosphorylation, acetic acid should be synthesized and consumed simultaneously during growth to maintain proper balance between metabolites of the PTA-ACS node. This is in agreement with Shin et al. [28] who proposed that wild-type E. coli constitutively synthesizes acetate even when growing on non-acetogenic carbon source succinate or at low growth rates in carbon limited cultures. It also has to be mentioned that acetic acid is a by-product in the synthesis of cysteine, methionine and arginine, covering around 0.4 mmol/g DCW (Additional file 4). Based on our experimental and literature data, production and re-assimilation of acetate might be over 1 mmol/g DCW at μ = 0.2 h^-1 (Text S2 in Additional file 1) which further supports the hypothesis of the necessity for constant acetic acid synthesis.

A similar regulation for overflow metabolism of acetate was posed for Saccharomyces cerevisiae by Postma and co-workers: they postulated that acetate accumulation is the result of insufficient acetyl-CoA synthetase activity for the complete functioning of the pyruvate dehydrogenase bypass because of glucose repression of ACS at high growth rates [43]. The hypothesis proposed here is also consistent with the observation that an acs mutant of E. coli accumulated acetate in chemostat cultures at dilution rate (D) 0.22 h^-1 whereas acetate overflow was started in wild-type at a higher D = 0.35 h^-1[28]. Furthermore, it has been shown that over-expression of acs[44] and constitutively expressed acs together with glyoxylate shunt repressors iclR and fadR mutant resulted in a significant reduction in acetate accumulation in glucose batch fermentations [28]. Adams and co-workers showed that as a result of micro-evolution, E. coli increased acetate consumption capability by over-expressing Acs (not AckA) [45, 46], further supporting the connection between Acs activity and acetate accumulation.

As Acs down-regulation is responsible for triggering overflow metabolism and the resulting accumulation of acetate is detrimental to cellular growth, it bears questioning why E. coli has not evolved towards maintaining sufficient Acs levels for acetate assimilation in all growth conditions. Growth conditions in E. coli native environments are rough as concentrations of utilizable carbon sources including acetate are in the low mg L^-1 range and access to nutrients is troublesome [47]. These harsh conditions force E. coli to make its metabolism ready for scavenging all possible carbon sources including acetate. However, in nutrient rich laboratory conditions, E. coli focuses on anthropic growth [48] and biomass production rate, primarily realized by enhancing readily oxidizable substrate (glucose) uptake kinetics which in turn results in Acs repression through CCR and thus, acetate accumulation [46]. This indicates that an active Acs is not essential for rapid growth for E. coli. It seems that maintaining high expression levels of acetate assimilation components (and also other alternative substrates ones) is energetically not favorable at higher growth rates. Moreover, as the space on cell membrane is limited and as E. coli achieves more rapid growth probably by increasing the number of glucose transport machinery components on the membrane, using area for alternative substrate transport proteins is not beneficial for faster growth. Interestingly, even in one of its natural environments-urinary tract-where a continuous dilution of acetate occurs, it has been shown that metabolizing acetate to acetyl-CoA by Acs is not essential for normal E. coli colonization as PTA-ACKA pathway and maintenance of a proper intracellular acetyl-P concentration are necessary for colonizing murine urinary tract [32].

Based on all the points discussed above, PTA-ACS might function as a futile cycle to provide rapid regulation of acetyl-P concentration in the cell for an active chemotaxis that is vital in natural nutrient-depleted environments, fighting against other organisms (acetate production), pathogenesis, biofilm formation etc. This hypothesis is consistent with the fact that the flagellar assembly and regulation operon (tar-tap-cheRBYZ) was more intensively expressed at lower growth rates (Additional file 2) where residual glucose concentration is smaller.

Concerning Acs down-regulation, it is possible that CCR is responsible for its repression as proposed by Treves et al. [46] showing the link between ACS expression level and acetate accumulation. In our experiments, it was shown that activation of CCR and repression of Acs take place simultaneously prior to the start of overflow metabolism (Figure 1 and Figure 4). As it is well known that CCR is initiated by the presence of glucose in the medium [49, 50], we propose that increasing residual glucose concentration accompanying smooth rise of dilution rate in A-stat triggers Acs down-regulation by CCR. The cAMP-Crp complex is one of the major players in CCR of E. coli as cAMP binding to Crp drastically increases its affinity towards activating the promotors of catabolic enzymes, including Acs [6, 49, 50]. We measured a 1.5-fold decrease in Crp expression with increasing growth rate (Figure 1) that is in agreement with the data in the literature [51]. In addition, when E. coli mutant defective in the gene crp was cultivated in glucose-limited chemostat at a low D = 0.10 h^-1, it accumulated acetate whereas the wild-type did not [52]. Furthermore, it exhibited a 34% higher biomass yield relative to the wild-type-this increase might be explained by reduced ATP wasting in the acetate futile cycle, which can be directed to biomass synthesis. Moreover, Khankal et al. [53] noted that E. coli CRP* mutants that do not require Crp binding to cAMP to activate the expression of catabolic genes showed lowered glucose effect on xylose consumption, 3.6 times higher acs expression levels and secreted substantially less acetate in xylitol producing batch fermentations. The connection between cAMP concentration and acetic acid consumption capability, together with the two-phase acetate accumulation profile observed in A-stat and D-stat cultures (Figure 2 and Figure 5) suggests a correlation between increasing residual glucose concentration mediated cAMP-Crp repression and acetate accumulation. Thus, cAMP-Crp dependent regulation of Acs transcribing acs-yjcH-actP operon might be a reason for acetate excretion, as also proposed by Veit et al. [10]. Our hypothesis of the CCR mediated acetate overflow metabolism is as well in agreement with the fact that rising glucose lowers the intracellular Crp level through the autoregulatory loop of the crp gene [54]. However, other mechanisms can also be involved in Acs down-regulation, for example by Cra (Figure 1). Indeed, Sarkar and colleagues have shown that glucose uptake and acetate production rates increased with a decrease of acetate consumption in an E. coli cra mutant [55].

What could be the biological relevance of the disruption of the PTA-ACS node? Firstly, decline of the ATP-spending PTA-ACS cycle throughput with increasing growth rate points to possible lower ATP spilling (our model calculations). Secondly, disruption of the PTA-ACS node decreases the energy needed for expression of this cycle's components. As the disruption of PTA-ACS cycle is CCR-mediated, repression of other alternative substrate transport and utilization enzymes by CCR enables to save additional energy. This could all lead to the decrease of ATP production as was indicated by the diminishing TCA cycle fluxes (Figure 2). Hence, it is plausible that cells repress (by CCR) the expression levels of alternative substrate utilization components (including Acs) for making space on the cell membrane for more preferred substrate (glucose) utilization and ATP producing components to achieve faster growth (see above).

Finally, it was demonstrated that highly reproducible A-stat data are well comparable to chemostat at the level of major growth characteristics and transcriptome, hence quasi steady state data from A-stat can be considered steady state representative (Table 1; Figure S1 and Figure S3 in Additional file 1). Furthermore, as shown also by Postma et al. for S. cerevisiae[43], chemostat is not fully suitable for characterization of dilution rate dependent metabolic transitions, whereas A-stat should be considered an appropriate tool for this. A-stat is especially well suited for the studies of the details of transient metabolism processes. Dynamic behavior of acetate, cAMP etc. with increasing specific growth rate (Figure 2 Figure 3 and Figure S1 in Additional file 1) and change in acetic acid consumption capability in the two-substrate A-stat (Figure S2 in Additional file 1) could be cited as good examples of the latter.

This study is an excellent example of how a systems biology approach using highly reproducible advanced cultivation methods coupled with multiple 'omics analysis and metabolic modeling allowed to propose a new possible regulation mechanism for overflow metabolism in E. coli: acetate overflow is triggered by carbon catabolite repression mediated Acs down-regulation resulting in decreased assimilation of acetate produced by Pta, and disruption of the PTA-ACS node. The practical implications derived from this could lead to better engineering of E. coli in overcoming several metabolic obstacles, increasing production yields etc.