Intracellular metabolite levels during the acidogenic-solventogenic transition.
The metabolism of C. acetobutylicum is characterized by an initial acidogenic phase (exponential growth), followed by solvent production and acid re-assimilation. The goal of the present study was to characterize the metabolic changes associated with the transition from exponential growth to solventogenesis. The system under study consisted of pH-controlled (pH 4.7) batch cultures grown on chemically defined media containing glucose as the sole carbon source and ammonia as the sole nitrogen source. As shown in , as the culture exits the exponential growth phase, the rate of acid production decreases and solvent production begins. A few hours later, the concentration of acids in the media starts to decrease due to their re-assimilation and conversion into solvents. Cell growth continues during solventogenesis, albeit at a slower rate than during the exponential phase. The observed fermentation profile is typical for C. acetobutylicum batch cultures.
Fig. 1. Progressive metabolome remodeling during the acidogenic-solventogenic transition. (A) Fermentation profile and growth curve obtained in a pH 4.7 controlled batch culture. (B) Metabolome dynamics during the acidogenic-solventogenic transition. Unsupervised (more ...)
Samples for the analysis of intracellular and extracellular metabolites were taken over a 28-h period (14 time points). The first four data points (4 to 8 h) correspond to mid- to late-acidogenic phase, the remaining time points correspond to the transitional and solventogenic phases (9 to 32 h). Based on its fermentation profile, the solventogenic phase can be divided into two stages: an early solventogenic phase (11 to 16 h) in which mixed acid-solvent production is observed and a mid-solventogenic phase (22 to 32 h) where acid re-assimilation takes place and the rate of solvent production increases. Late solventogenesis, where acid re-assimilation stops and solvent production decreases, was beyond the scope of the present study and was not examined. Samples were analyzed by a set of liquid LC-MS methods which together enable the analysis of ~200 known metabolites. Quantitative data (relative or absolute concentrations) throughout the time course was obtained for 114 intracellular metabolites and 50 extracellular metabolites. Absolute concentrations were obtained for 79 of the intracellular metabolites, and relative concentrations were obtained for 35 intracellular metabolites.
Our analysis revealed an orderly progression of changes in the levels of intracellular metabolites throughout the time course ( and see Fig. S1 and S2 in the supplemental material). Rather than being restricted to metabolites in the acid-solvent-producing pathways, these alterations encompassed all analyzed pathways. The unsupervised hierarchical clustering analysis shown in highlights the major trends in the data. The time point samples cluster into three major groups corresponding to the exponential (acid production), early solventogenic (mixed acid/solvent production), and mid-solventogenic (acid reutilization with solvent production) phases. This clustering, based solely on the measured metabolite levels, is fully consistent with the actual order of sample collection. While metabolites belonging to the same class (such as amino acids) or pathway (such as glycolysis) do not always cluster together, some general trends are evident: most amino acids, upper glycolytic intermediates, and pentose phosphate pathway compounds are at high levels in acidogenesis and lower levels in solventogenesis; butyryl-phosphate and acetyl-phosphate are at high levels during acidogenesis and early solventogenesis but at low levels in late solventogenesis; ATP, NADH, and butyryl-CoA peak in early solventogenesis; and the levels of low-energy forms of nucleotides (including diphosphates, monophosphates, nucleosides, and bases) increase late in the time course.
Changes in the composition of the core C. acetobutylicum metabolome.
The absolute intracellular concentrations of 79 metabolites were determined in acidogenic and solventogenic cells (). The intracellular metabolome was dominated, on a molar basis, by a small number of abundant metabolite classes: amino acids (~55% during both acidogenesis and solventogenesis), central carbon intermediates (~27% during acidogenesis, 17% in solventogenesis), and nucleotides (5% during acidogenesis, 10% in solventogenesis). Similar to E. coli
), during exponential growth the most abundant metabolites were glutamate and fructose-1,6-bisphosphate.
Intracellular metabolite concentrations in acidogenesis and solventogenesisa
The transition from exponential growth into solventogenesis has significant effects on the overall metabolome composition (). We observed a decrease in the total concentration of measured metabolites, from ~150 mM during the mid-exponential phase to ~70 mM during mid-solventogenesis. This reflects in large part decreases in concentration of a few highly abundant metabolites such as glutamate, which is both a key nitrogen donor and a counter ion for potassium. Glutamate is replaced in part by aliphatic amino acids, transiently by valine, and subsequently by alanine. Fructose-1,6-bisphosphate, the second most abundant metabolite during exponential growth, also declines steadily during the transition to solventogenesis.
Fig. 2. Changes in metabolome composition, energy charge, and NAD(P)H/NAD(P)+ ratios during acidogenic-solventogenic transition. (A) Changes in overall metabolome composition. The graph shows the molar abundance of 79 different metabolites, most of them combined (more ...)
In the nucleotide class, the ratio of high-energy triphosphates to lower-energy bisphosphates and monophosphates decreases in solventogenesis (). In C. acetobutylicum, a major route of ATP generation is acid production. Conversely, a main route for acid re-assimilation utilizes ATP. Therefore, the switch from acid production to acid re-assimilation may cause the observed decrease in energy charge.
The ratio of redox cofactors is also affected. Surprisingly, both the NADH/NAD+
the and NADPH/NADP+
ratios decrease during solventogenesis (). This may initially appear counterintuitive since high NAD(P)H/NAD(P)+
ratios should favor solvent production. However, the decrease in these ratios may result from rapid consumption of NADH and NADPH by solvent-producing pathways. It has been shown that high NADH/NAD ratios reduce the activity of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) in C. acetobutylicum
). A potential benefit of maintaining a low NADH/NAD ratio during solventogenesis would be to maintain glycolytic flux; if this ratio were to be increased in order to drive solventogenesis, it could potentially have an adverse effect on glycolytic flux because of the inhibition of GAPDH.
In addition to the typical fermentation products (i.e., acetate, butyrate, butanol, ethanol, and acetone), we detected and quantified 50 extracellular metabolites ( and see Fig. S3 in the supplemental material). During exponential growth, modest amounts of malate, fumarate, succinate, and pyruvate are secreted into the media (). As the cells exit exponential phase, malate and fumarate are rapidly consumed. Succinate excretion continues during early solventogenesis but slows markedly during mid-solventogenesis.
Fig. 3. Extracellular metabolites during the solventogenic transition. (A) Unsupervised hierarchical cluttering analysis of extracellular metabolite concentration data. Rows represent specific metabolites. Columns correspond to the fermentation time points shown (more ...)
At the onset of solvent production, a large number of typically intracellular metabolites, including phosphorylated compounds, start to accumulate in the media (). The total concentration of extracellular metabolites increases from ~1 mM during acidogenesis to ~3 mM during solventogenesis (). One possible explanation is increased membrane permeability during solventogenesis caused by the accumulation of extracellular butanol, which may have a chaotropic effect that disrupts the integrity of the cellular membrane. In C. acetobutylicum
, butanol has been shown to affect membrane fluidity and composition, generally increasing the ratio of saturated to unsaturated fatty acids tails in membrane phospholipids (3
). However, some metabolites with large intracellular concentrations, such as glutamate or glutamine, are not detected extracellularly at significant levels during solventogenesis. It is possible that active uptake mechanisms exist for potentially useful metabolites such as this (glutamate and glutamine are valuable carbon and nitrogen sources) that prevent them from accumulating extracellularly despite some leakage caused by increased membrane permeability during solventogenesis.
Metabolic impact of pH, acids, and solvents.
As the cells transition from acidogenesis into solventogenesis, they are subjected to several changes in their environment: acid accumulation, decreasing pH, increasing cell density, and finally solvent accumulation. Here we examined the metabolic consequences of maintaining the fermentation culture at pH 6.0, of adding exogenous acid, or of adding exogenous butanol. As expected based on prior literature (33
), maintaining the fermentation at pH 6 increased initial cell growth and allowed only a very short period of solvent production as the cells exit exponential growth (see Fig. S4A in the supplemental material). There was no acid reuptake or overt solventogenic phase. Consistent with these observations, the metabolome pattern during acidogenesis and up to the onset of solventogenesis was largely unaffected but then diverged markedly from the lower pH culture (see Fig. S4B in the supplemental material).
To test the effects of acid accumulation, together with a drop in pH, we added exogenous butyrate (30 mM) and acetate (30 mM) to exponentially growing cells (OD600 = 0.6, initial pH 6.0) while simultaneously decreasing the pH to 4.7. This treatment did not, on a short time scale (2 h), trigger solvent production and did not reproduce the overall metabolic changes observed during acidogenesis or early solventogenesis (see Fig. S6A in the supplemental material). However, the addition of acids caused a large increase in the levels of butyryl-CoA, butyryl-phosphate, and acetyl-phosphate and a modest transient increase in acetyl-CoA levels (see Fig. S6B in the supplemental material). This suggests that the extracellular accumulation of acids during acidogenesis and early solventogenesis may be responsible for the increase in the levels of some of these metabolites during that time. The inability of this treatment to initiate the transition into solventogenesis suggests that other factors, perhaps related to high cell densities, are also required to produce the full scope of metabolic changes leading to effective solventogenesis.
To test whether butanol is responsible for the increased membrane permeability or metabolome changes during solventogenesis, we treated exponentially growing cells (OD600
= 0.6, pH controlled at 6.0) with butanol (70 mM). We observed a decrease in growth rate of ca. 30% after butanol addition. There was not, however, substantial release of intracellular metabolites into the media (see Fig. S7A in the supplemental material). Although butanol did not have a discernible effect on membrane permeability, it had significant effects on the intracellular concentrations of several metabolites (see Fig. S7B in the supplemental material). These included a rapid increase in valine, leucine/isoleucine, and glutamine that have a slight similarity to the changes occurring at the time of the solventogenic transition. In B. subtilis
, changes in membrane fluidity can be accomplished by the synthesis of branched-chain fatty acids from the branched amino acids valine, leucine, and isoleucine (32
). It has been established that C. acetobutylicum
adjusts membrane fluidity in response to butanol stress (3
) and that branched-chain amino acid biosynthetic genes are upregulated at the onset of solvent production (1
). Therefore, it is likely that the increased production of branched-chain amino acids that we observe at the onset of solventogenesis, and during butanol challenge, enable synthesis of branched-chain fatty acids and adjustments to membrane fluidity. However, with the exception of the few branched-chain amino acids just discussed, the addition of butanol failed to reproduce the intracellular concentration changes associated with the transition into solventogenesis (see Fig. S7 in the supplemental material).
The results of the experiments in this section indicate that no single factor (low pH, acid or solvent accumulation) is responsible for the organized, progressive metabolome evolution during the solventogenic transition; instead, multiple factors are required to drive this progression.
Probing intracellular fluxes.
In addition to metabolite concentrations, in vivo reaction rates are among the most important characteristics of a biochemical network. The widespread changes in metabolite concentrations during the acidogenic-solventogenic transition suggested that core metabolic fluxes are affected. Although metabolic fluxes are not directly observable, they can be measured based on nutrient uptake and waste excretion rates and on the assimilation of stable isotope tracers into intracellular metabolites.
During solventogenesis the specific rate of glucose consumption decreases >3-fold, going from ~12 mM h−1 grams cell dry weight−1 (gCDW−1) during the mid-exponential phase to ~3.5 mM h−1 gCDW−1 during the mid-solventogenic phase. About 60 to 70% of the consumed glucose is excreted as acids during acidogenesis. During solventogenesis, as much as 85% of the glucose consumed is converted into solvents while additional solvent production eventually occurs from acid reutilization. The overall conversion yield of glucose to biomass is higher during acidogenesis (~31 gCDW mol−1) than during solventogenesis (~17 gCDW mol−1).
Since glucose consumption decreases during solventogenesis, overall intracellular metabolic flux must also decrease. However, we do not expect all metabolic pathways to be equally (proportionally) affected. To investigate the remodeling of central metabolic fluxes during solventogenesis, we monitored the dynamics of incorporation of labeled glucose into intracellular metabolites during acidogenesis and solventogenesis. Universally 13C-labeled glucose was added into mid-exponential or mid-solventogenic cultures in a matching amount to the nonlabeled glucose remaining in the media to reach a 1:1 ratio between 13C-labeled and nonlabeled glucose. Samples were taken at short time intervals (e.g., 1, 2, 3, 5, 7, 10, 15, 30, and 60 min) after [13C]glucose addition. The rate of incorporation of 13C-labeled glucose into downstream metabolites was monitored using LC-MS.
Consistent with the persistent uptake of glucose throughout the fermentation, the rate of incorporation of 13C-labeled glucose into glycolytic intermediates (e.g., hexose-phosphates, fructose-1,6-bisphosphate, dihydroxyacetone-phosphate, and phosphoenolpyruvate) and pentose phosphate pathway metabolites (e.g., ribose-5-phosphate) is fast in both acidogenic and solventogenic cells (). By 1 min, metabolites in these pathways reach isotopic steady state (no further change in their 13C-labeled fraction). The steady-state labeling fraction of these metabolites is consistent with the ~50% labeling of extracellular glucose. In contrast, the labeling rate in most amino acids is considerably slower during solventogenesis compared to acidogenesis. In the reductive TCA cycle, malate and fumarate display nearly nondetectable labeling during solventogenesis. However, metabolites in the oxidative TCA cycle, such as citrate, still become labeled.
Fig. 4. Isotopic tracer experiments reveals differences in pathway activity in acidogenic versus solventogenic cultures. The rate of incorporation of 13C-labeled glucose into downstream metabolites reveals differences in pathway activity in acidogenic versus (more ...)
Several acidogenic/solventogenic metabolites display interesting labeling patterns. Although the labeling rate of acetyl-CoA is comparable to glycolytic intermediates, it does not reach the expected 50% 13
C-labeled fraction in either acidogenic (only 35%) or solventogenic (only ~25%) cells (). The 13
C-labeled fraction in acetyl-phosphate is even lower: ~10% in acidogenic and ~5% in solventogenic cells. These labeling patterns are expected in solventogenic cells where nonlabeled acetate is re-assimilated from the media and converted into acetyl-phosphate and acetyl-CoA, but not in aci-dogenic cells with high rates of acetate production. The lower-than-expected 13
C-labeled fraction of acetyl-CoA and acetyl-phosphate suggests that, even during acidogenesis, there is a rapid interchange between the acetyl-CoA pool and the extracellular acetate pool, likely achieved through the rapid reversibility of the enzymes phosphotransacetylase (conversion of acetyl-CoA to acetyl-phosphate) and acetate kinase (conversion of acetyl-phosphate to acetate). Consistent with this hypothesis, when we added 13
C-labeled acetate to acidogenic cultures, we observed rapid labeling of intracellular acetyl-phosphate, followed by partial labeling of acetyl-CoA and downstream metabolites such as butyryl-CoA (see Fig. S8A in the supplemental material). In addition, using a filter culture methodology (2
), acidogenic cells can be rapidly switched to fresh media containing 100% 13
C-labeled glucose with no extracellular acetate. In this situation, rapid and complete labeling of acetyl-CoA and acetyl-phosphate are observed (butyryl-CoA and butyryl-phosphate are also rapidly and completely labeled), indicating that partial labeling of acetyl-CoA and acetyl-phosphate in batch fermentation experiments is due to extracellular unlabeled acetate (see Fig. S8B in the supplemental material). Similarly to acetyl-CoA and acetyl-phosphate, butyryl-CoA and butyryl-phosphate do not reach their expected 13
C-labeled fractions in acidogenic cells. This also suggests a rapid interchange also between butyryl-CoA and extracellular nonlabeled butyrate. Finally, the hypothesis of a rapid interchange between extracellular acids and their intracellular CoA derivatives also implies that, as the fractions of extracellular 13
C-labeled acetate and butyrate slowly increase in the media after the addition of [13
C]glucose, the 13
C-labeled fractions of acetyl-CoA, butyryl-CoA, acetyl-phosphate, and butyryl-phosphate will also slowly increase beyond the initial pseudo-steady-state values discussed above. Figure S8C in the supplemental material shows this phenomenon at longer time scales (1 h).
It has been previously established that C. acetobutylicum
has a bifurcated TCA cycle: oxaloacetate flows to succinate both through citrate/α-ketoglutarate (oxidative branch) and via malate/fumarate (reductive branch) (2
). We argued that such a bifurcated TCA cycle could play a role in cellular redox balance (2
). Here we find that during solventogenesis the electron-consuming reductive TCA cycle is shut down, since there is no detectable labeling of malate or fumarate from 13
C-labeled glucose (). However, the NADH producing oxidative (clockwise) right side of the TCA cycle remains active. This reorganization may represent a deliberate effort to channel reducing power toward solvent production. As shown before, malate and fumarate are excreted during acidogenesis and re-assimilated during the transition into solventogenesis (). To determine the fate of the re-assimilated fumarate, we added 13
C-labeled fumarate during the time of the acidogenic-solventogenic transition. The labeled fumarate was rapidly converted to malate and metabolized via the TCA cycle into glutamate or aspartate (see Fig. S9 in the supplemental material). This rapid re-assimilation of fumarate and malate during the transition may be helpful in kick-starting or sustaining early solvent production by providing additional reducing power.
It is possible that the switch into solventogenic metabolism involves not only changes in metabolic fluxes but the utilization of alternative pathways in core metabolism. In particular, the oxidative pentose pathway or the Entner-Doudoroff pathway for glucose catabolism could be advantageous for the production of NADPH. Previously, we ruled out the activity of these pathways during acidogenesis using additional labeled substrates such as [1,2-13
). Similar experiments conducted here show that during solventogenesis these pathways are also not active. In addition, the correspondence between the amino acid labeling patterns of acidogenic and solventogenic cells indicates that the routes for amino acid biosynthesis also remain unchanged (see Table S3 in the supplemental material).
Quantitative flux analysis during acidogenesis and solventogenesis.
We used kinetic flux profiling (KFP) (42
) to obtain a quantitative estimate of metabolic fluxes from the kinetics of incorporation of 13
C-labeled glucose into downstream metabolites during acidogenesis and solventogenesis. We constructed an ordinary differential equation (ODE) model for the metabolic network shown in Fig. S10 in the supplemental material that describes the isotope labeling kinetics of metabolites following the addition of universally labeled [13
C]glucose. Given the model equations, we used a nonlinear global search algorithm (genetic algorithm) to identify sets of fluxes that can quantitatively reproduce the experimental labeling dynamics (16
). The inputs to the model included the metabolite labeling kinetics, intracellular metabolite concentrations, glucose uptake rate, and acid/solvent production rates.
shows a map of the identified median flux values (normalized to glucose consumption) during acidogenesis () and their fold changes during solventogenesis (). Although glycolytic flux during solventogenesis decreases proportionally to the specific rate of glucose consumption, we observed a disproportionally large decrease in the synthesis of nearly all amino acids. After normalizing to glucose uptake, the average decrease of carbon flux into amino acids during solventogenesis is ~14-fold. This large and generalized decrease in amino acid biosynthesis redirects carbon toward acetyl-CoA and the acidogenic/solventogenic pathways (). Amino acid biosynthesis requires large amounts of ATP and NADPH. Therefore, the decrease in amino acid biosynthesis during solventogenesis means that limited energy and reducing power can be instead directed toward solvent production. Indeed, the fact that the energy charge and the NADPH/NADP+ ratio decrease during solventogenesis suggests that ATP and/or NADPH may be limiting resources during solventogenesis. As mentioned before, we observed a spike in the concentration of some amino acids (serine and valine) during the transition and early solventogenesis; at this time, however, the energy charge and NADPH/NADP ratio are still relatively high.
Fig. 5. Alterations in metabolic fluxes during solventogenesis. We used kinetic flux profiling (KFP) to obtain a quantitative estimate of metabolic fluxes from the kinetics of incorporation of 13C-labeled glucose into downstream metabolites during acidogenesis (more ...)
Our results suggest that the regulation of metabolic fluxes surrounding pyruvate is a hallmark of the transition into solventogenic metabolism. The various alterations that we observe have the net effect of redirecting flux from pyruvate into acetyl-CoA. Our model estimates a >10-fold decrease of carbon flux into oxaloacetate from pyruvate (catalyzed by pyruvate carboxylase), which appears to be of particular importance since this is the second largest outflux from pyruvate during acidogenesis (). Since oxaloacetate is required as a precursor for the biosynthesis of 11 different amino acids, reduced flux through pyruvate carboxylase may be a major contributing factor in the generalized decrease of amino acid biosynthesis during solventogenesis.
Correlation with transcriptomic studies.
Two previous studies have investigated the transcriptional alterations associated with the acidogenic-solventogenic transition in C. acetobutylicum
). The study by Alsaker and Papoutsakis reported the changes in gene expression that occur from mid-exponential phase to early solventogenic phase in a pH-controlled batch culture using experimental conditions similar to the ones used in the present study (1
). These researchers observed that, in addition to the expected upregulation of solvent-producing genes, the expression of core metabolism genes (i.e., glycolysis, TCA cycle, and amino acid biosynthesis) was significantly affected during the solventogenic transition. We compared the reported gene expression data to our metabolomic data ().
Fig. 6. Correlation between metabolomic data and previous transcriptomic studies. Selected average expression profiles for various enzymes and metabolic pathways are shown on the left. The concentration profiles of related metabolites are shown on the right. (more ...)
Although the expression of glycolytic genes changes only slightly, glucose-specific phosphotransferase system (PTS) genes were greatly downregulated at the onset of solventogenesis. The expression profile of glucose PTS genes (cac0570
, and cac3427
) correlates with the drop in the levels of the upper glycolysis intermediates hexose-phosphate (combined pools of glucose-6-phosphate and fructose-6-phosphate) and fructose-1,6-bisphosphate (). The PTS system is the main route for glucose uptake in C. acetobutylicum
), and its downregulation and the concomitant low levels of upper glycolysis intermediates may contribute to the decreased glucose uptake and glycolytic flux during solventogenesis. Moreover, the PTS system consumes phosphoenolpyruvate, providing one of the major drains from the bottom of glycolysis. Decreased PTS activity is accordingly a potential contributor also to the accumulation of lower glycolytic intermediates during solventogenesis. The other major drain from phosphoenolpyruvate is pyruvate kinase. Given that fructose-1,6-bisphosphate is an allosteric activator of pyruvate kinase in many organisms, its lower levels, also potentially secondary to lower PTS activity, may augment these effects.
We found that the reductive TCA cycle shut downs during solventogenesis and that fumarate levels peak before the onset of solventogenesis. These observations correlate with the decreased expression during the transition of fumarate hydratase (cac3090 and cac3091), which interconverts malate and fumarate. This may be a contributing factor, in addition to lower NAD(P)H levels, to the decreased activity of the reductive TCA.
The majority of ribosomal genes are greatly downregulated at the onset of solventogenesis (1
); this implies decreased protein biosynthesis, which is consistent with our observation of the generalized decrease in amino acid biosynthesis. The expression patterns of amino acid biosynthetic genes, however, are more complex (1
). Serine, valine, leucine, and isoleucine biosynthetic pathway genes are sharply upregulated during the transition, and their metabolite levels also peak during that time (). The expression levels of other amino acid biosynthetic genes, such as those involved in glutamate, phenylalanine, arginine, and histidine biosynthesis, also increase during solventogenesis (1
); however, the associated pathway fluxes and end product concentrations are both markedly decreased. Similarly, the expression of the pyruvate carboxylase gene (cac2660
) actually increases during solventogenesis, despite the greatly decreased pyruvate carboxylase flux. The reason for high expression of pyruvate carboxylase and certain amino acid biosynthetic genes, whose associated pathways are not metabolically active, is currently unknown. These contradictory findings, however, illustrate the importance of directly measuring metabolic activity rather than merely inferring it from gene expression studies.