Development of efficient processes for converting renewable lignocellulosic biomass into biofuels requires microorganisms that produce target biofuels such as ethanol at high yields, titers, and productivities (Blanch et al., 2008). It is a challenging task to engineer microorganisms to perform well in the presence of compounds derived from biomass hydrolyzates, which typically contain mixtures of hexoses and pentoses together with inhibitors such as organic acids (e.g., ferulic acid, acetic acid), furan derivatives (e.g., furfural, hydroxymethylfurfural), and phenolic compounds. In addition, biofuels, such as ethanol and butanol, produced by these microorganisms are themselves inhibitory to cell growth and product formation (Palmqvist and Hahn-Hägerdal, 2000). Determining the mechanisms of such inhibition can provide useful strategies for developing microorganisms with reduced levels of inhibition.
Several microorganisms have been employed for converting sugars into biofuels such as ethanol. Both Saccharomyces cerevisiae and Zymmomonas mobilis are native ethanologens that produce ethanol at high yields and titers. However, their inability to ferment pentose sugars limits their application in fermenting lignocellulosic biomass hydrolyzates, although much research has been directed toward overcoming this barrier (Hahn-Hagerdal et al., 2007; Ho et al., 1998; Zhang et al., 1995). Even though Escherichia coli is not a native ethanologenic microorganism, Ingram and coworkers (Ohta et al., 1991) were the first to develop a chromosomally integrated recombinant ethanologenic strain (KO11) that contained ethanol-producing genes from Z. mobilis. Higher producing strains were subsequently developed by others using plasmid-encoded genes (Dien et al., 1998, 2000; Trinh et al., 2008). An advantage that E. coli has over S. cerevisiae and Z. mobilis is that it can natively ferment a wide range of pentose and hexose sugars present in biomass hydrolyzates. However, E. coli is not as ethanol-tolerant as either S. cerevisiae or Z. mobilis.
Since KO11 was first developed, it has been characterized in several different laboratories (Dumsday et al., 1999; Huerta-Beristain et al., 2008; Kim and Lee, 2007; Lau et al., 2008; Okuda et al., 2008; Rao et al., 2007). Typically, from 30 to 50 g/L ethanol accumulates during batch fermentations, which significantly reduces further ethanol production. KO11 is also unable to maintain ethanol production at high yields during long-term continuous fermentation when sugars other than glucose serve as substrates (Dumsday et al., 1999). In the present study, we employed the ethanologenic E. coli strain KO11 as a model recombinant ethanologen to examine the mechanisms of ethanol toxicity in both batch and continuous fermentations.
Materials and Methods
Organisms and Plasmids
E. coli KO11 (ATCC# 55124) was obtained from the American Type Culture Collection. The strain has the pyruvate decarboxylase gene (pdc) and alcohol dehydrogenase gene (adhB) from Z. mobilis integrated into the E. coli chromosome and located behind the pflB promoter. This insertion disrupted the chromosomal gene pflB (Ohta et al., 1991). The vector pLOI297 (ATCC# 68239) that contains the pyruvate decarboxylase gene (pdc) and alcohol dehydrogenase gene (adhB) from Z. mobilis was also used in the study. KO11/pLOI297 contained additional copies of pdc and adhB genes.
Strain KO11 was characterized in Luria–Bertani (LB) medium containing 5 g/L yeast extract, 10 g/L tryptone, 5 g/L NaCl, glucose at various concentrations, and 40 µg/mL chloramphenicol (Cm). When KO11/pLOI297 was characterized, 100 µg/mL ampicillin (Amp) was additionally used for selection. For batch fermentation, unless specified, 100 g/L glucose was used. For continuous fermentation, 20 g/L glucose was used. For bioreactor runs, antifoam 204 (cat# A6426, Sigma–Aldrich, St. Louis, MO) was added before inoculation at a concentration of 100 µL/L.
Batch fermentations were conducted in 3.7 L bioreactors (Bioengineering AG, Wald, Switzerland) with a working volume of 2 L, under anaerobic growth conditions. The operating conditions were 200 rpm, pH 6.5, and 37°C (unless otherwise specified). Anaerobic conditions were maintained by continuously sparging the bioreactor with N2 at a volumetric flow rate of 100 mL/min throughout the fermentation. Anaerobic growth was confirmed by a 0% level of dissolved oxygen indicated during the fermentation. Exponentially grown cell cultures were used for the inoculation with a volume ratio of 1:40, and the OD600 of initial cultures after inoculation was 0.05. Optical density was measured at 600 nm using a spectrophotometer (Model: Genesys 20, Thermo Scientific, Waltham, MA). A correlation between OD600 and dry cell weight (DCW) was used to calculate the biomass concentration during fermentation (1 OD600 = 0.5 g DCW/L). All batch fermentations were performed at least twice in independent experiments.
Continuous fermentation was conducted with a working volume of 1 L under similar growth conditions as in batch fermentation. The feed medium was also sparged with N2. Cells were first grown in batch mode for several hours until an optical density of 0.5 (0.25 g DCW/L) was obtained and then switched to continuous mode. Steady state was established when OD600 did not change after 3–4 residence times. Unless specified, a dilution rate of 0.1 h−1 was set for all chemostat runs. To determine the maximum specific growth rate, the feed medium flow rate was increased to provide a dilution rate of 1 h−1. By observing the time course of OD600, the maximum specific growth rate was obtained (data not shown).
To determine the effect of added ethanol and acetate on cell growth of KO11, growth experiments were conducted in 500 mL custom-made bioreactors by using spinner flasks (cat# 1965-61015, Bellco Biotechnology, Vineland, NJ) with a working volume of 250 mL at 37°C. The medium contained LB with 5 g/L glucose and had initial pH 6.5 adjusted with either H3PO4 or KOH. Nitrogen was sparged into the fermenter at a volumetric flow rate of 60 mL/min 1 h before inoculation and throughout the fermentation to maintain anaerobic growth conditions.
Yield Calculation and Carbon Balance
For batch fermentation, the ethanol yield on glucose was calculated by the formula YETOH/Glc = rETOH/rGlc, where rETOH (g L−1 h−1) and rGlc (g L−1 h−1) represent the ethanol production rate and glucose uptake rate, respectively. For continuous fermentation, the ethanol yield on glucose was determined from YETOH/Glc = (CETOH,o − CETOH,i)/(CGlc,i − CGlc,o), where CETOH and CGlc (g/L) represent the concentrations of ethanol and glucose, respectively, and the indices i and o stand for the inlet and outlet streams, respectively. The yields for other metabolites were calculated in the same manner.
A carbon balance was determined from the carbon-molar (C-mole) productivities of biomass, succinate, lactate, ethanol, formate, and carbon dioxide production. CO2 production rates were estimated based on the stoichiometry and rates of succinate, acetate, and ethanol production. For every C-mole of either acetate or ethanol formed, 0.5 C-mol of CO2 is produced. For every C-mole of succinate formed, 0.25 C-mol of CO2 is consumed. For these calculations the carbon content of biomass was assumed to be 0.5 g of carbon per gram DCW of biomass (Carlson and Srienc, 2004).
Metabolites such as glucose, succinate, lactate, acetate, and formate were quantified by high-performance liquid chromatography as described elsewhere (Trinh et al., 2008).
Pyruvate decarboxylase and alcohol dehydrogenase activity assays were performed on cell pellets collected from 5 mL aliquots. The cell pellets were washed with phosphate-buffered solution (PBS) and stored at −80°C for later analysis. Prior to using them in activity assays the cell pellets were washed again, resuspended in PBS, and the cell suspension was divided equally for activity measurements of the two enzymes.
For pyruvate decarboxlyase activity measurements, the cell pellet was washed twice with a 0.1 M Tris–HCl buffer (pH 7) containing 1 mM TPP, 1 mM MgCl2, 0.1 mM EDTA, and 2 mM mercaptoethanol. The cells were resuspended in this same buffer and lysed using a tip-sonicator. The cell lysate was centrifuged at 4°C at 14,000 rpm for 20 min. The supernatant was collected and heated in a 60°C water bath for 30 min to denature any competing enzymes. This solution was then re-centrifuged for 5 min at 14,000 rpm and the supernatant was collected. In a 4.5 mL cuvette, 2.6 mL of 0.1 M Tris–maleate buffer (pH 6.5), 0.1 mL of 0.3 M pyruvate, 0.2 mL of 1 mM NADH, and 0.6 µL of a 10 mg/mL yeast alcohol dehydrogenase solution were added. To initiate the reaction an aliquot of the cell extract containing 36 mg/mL protein was added. The absorbance was then measured at 340 nm every 1 s for at least 60 s (Hoppner and Doelle, 1983).
For alcohol dehydrogenase activity measurements, the cell pellet was washed twice with 15 mM potassium phosphate buffer (pH 7) containing 1 mM mercaptoethanol, 0.1 mM EDTA, and 100 µL of 100 mM PMSF solution in isopropanol. The cells were resuspended in this same buffer and lysed using a tip-sonicator. The cell lysate was centrifuged at 4°C at 14,000 rpm for 20 min. In a 4.5 mL cuvette, 2 mL of 0.1 M Tris–maleate buffer, 0.2 mL of 6 mM NADH, and 0.2 mL of 0.5 M acetaldehyde were added. To initiate the reaction an aliquot of the cell extract containing 36 mg/mL protein was added. The absorbance was measured at 340 nm every 1 s for at least 60 s (Hoppner and Doelle, 1983). One unit of pyruvate decarboxylase or alcohol dehydrogenase activity is defined as the creation of 1 µmol of NAD+ per minute. The total protein content of the supernatant was measured using a Bradford assay (cat# 500–0006, Biorad, Hercules, CA).
The presence of ethanol in the growth medium causes growth inhibition of the ethanologenic strain KO11. Figure 1A shows the effect of various concentrations of added ethanol on the batch growth of KO11. Cell growth was strongly inhibited at ethanol concentrations >30 g/L where the specific growth rate decreased more than 90% compared to the cell growth without added ethanol. Under anaerobic growth conditions, E. coli generates ATP through substrate-level phosphorylation. One of the main pathways used by E. coli to generate ATP is via acetate production, even though acetate can be toxic to the cells. Figure 1B illustrates the effect of acetate on growth of KO11.
Batch Fermentation of KO11
Figure 2 shows the fermentation characteristics of strain KO11 grown in batch. There are three distinct phases: exponential growth, inhibited growth, and no growth. During the exponential growth phase (0–6 h), KO11 has a specific growth rate of 0.84 h−1. During this phase, fermentative products, primarily ethanol, lactate, acetate, and formate were present at relatively low concentrations. It should be noted that the specific growth rate during exponential growth was lower than that presented in Figure 1A for the case of no added ethanol. The differences in growth rate may have resulted from the different concentrations of glucose and/or fermentative products. During the inhibited growth phase (6–18 h), the specific growth rate of KO11 decreased significantly, from 0.84 to 0.08 h−1. Fermentative products, including ethanol, lactate, acetate, and formate, accumulated during this phase. During the no-growth phase (18–30 h), 30 g/L of glucose remained from the original concentration of 100 g/L.
Fermentative products including lactate, acetate, ethanol, and formate were produced during the growth-inhibited and no-growth phases (Fig. 2). Ethanol was the dominant product followed by lactate, acetate, and formate, respectively. During the growth-inhibited phase, the rates of glucose consumption and fermentative product formation remained constant. As cells entered the no-growth phase, the acetate concentration was 4 g/L and the ethanol concentration was about 26 g/L. A small amount of acetate was produced immediately after cessation of the no-growth phase when glucose was completely consumed. The ethanol production rate decreased and formate consumption commenced. While production rates of other metabolites decreased, the volumetric rate of glucose consumption (g/L/h) did not change and the volumetric rate of lactate production increased. These results suggest that accumulation of fermentative products (primarily acetate and ethanol) inhibited cellular growth and redirected primary carbon metabolism.
Redirection of metabolism was also evident from a carbon balance (Fig. 3). The carbon balance was closed during the exponential growth and growth-inhibited phases but could not be closed during the no-growth phase. This imbalance was likely due to partial redirection of carbon towards maintenance energy, resulting in formation of CO2, which was not directly measured in the present work but was estimated by accounting for the production of the observed fermentative products. During the no-growth phase, the estimated CO2 based on measured fermentation products was insufficient to close the carbon balance, implying that maintenance requirements were significant but not completely accounted for. Growth of KO11 was also examined at a higher glucose concentration of 120 g/L, revealing the same growth characteristics and pattern of fermentative product formation as in the case of 100 g/L. The accumulation of ∼4 g/L acetate and 30 g/L ethanol arrested cell growth and caused reduction of ethanol production.
Regulation of Fermentative Product Formation in KO11
The glucose uptake rate remained unchanged and the lactate production rate increased during the transition from the growth-inhibited phase to the no-growth phase. This suggests that pyruvate decarboxylase and/or alcohol dehydrogenase were limiting in the ethanol production pathway. The glycolytic flux was sufficiently high that it could not be completely directed to ethanol formation from pyruvate (Fig. 4A and B). A portion of this carbon flux is thus redirected toward lactate. To verify this, several experiments were conducted to demonstrate that the glycolytic flux controls the mixed acid fermentation and that the limiting flux through the ethanol-producing pathway is key to the behavior of KO11.
Effect of High Glycolytic Flux on Mixed Acid Fermentation (Fig. 4C and D)
We characterized growth of KO11 in chemostats at dilution rates of 0.1 and 0.3 h−1. Steady-state concentrations of cells and metabolites are shown in Figure 5. An increase in the dilution rate resulted in increased lactate production due to an increase in the glycolytic flux. As the dilution rate was increased from 0.1 to 0.3 h−1, the specific glucose uptake rate increased from 5.63 ± 0.02 to 12.55 ± 0.11 mmol g−1 DCW h−1, and the lactate yield from glucose increased from 0.02 to 0.07 g/g. This trend was also observed during the first 2 h following a switch from batch to continuous cultivation. Because glucose was in excess and acetate and ethanol concentrations were below their inhibitory levels (Fig. 5B), cells were able to grow much more rapidly than the growth rate corresponding to a dilution rate of 0.1 h−1 and hence had correspondingly high glucose uptake and lactate production rates.
Rate-Limiting Steps in Ethanol-Producing Pathways Affect Both Ethanol Yield and Mixed Acid Fermentation (Fig. 4 E and F)
Even though ethanol was the main fermentative product, the high concentrations of lactate and acetate during the growth-inhibited phase indicate that the flux through the ethanol-producing pathway was limiting (Figs. 2 and 3). To confirm this, we increased the gene dosages in the ethanol-producing pathway by introducing both pdcZM and adhBZM into KO11 through the plasmid pLOI297. The resulting strain was designated as KO11/pLOI297. It should be noted that KO11 itself contained both of these two genes integrated into its chromosome, disrupting the pflB gene and positioned behind the pflB promoter. Figure 6 compares the performance of KO11/pLOI297 to the parent strain KO11 under identical growth conditions. During the first 12 h, the profiles of both biomass production and glucose consumption were almost identical for both strains (Fig. 6A). However, the distribution of mixed acid fermentation products was different. KO11/pLOI297 was able to convert glucose into ethanol at a high ethanol yield of 0.49 ± 0.00 g/g, representing 96% of the stoichiometric value with minimal acetate and lactate production. In contrast, KO11 had a lower ethanol yield of 0.39 ± 0.01 g/g due to production of acetate and lactate (Fig. 6B and C). As shown in Figure 7A, the increase in ethanol production of KO11/pLOI297 directly correlated with the increase in the specific activities of Pdc and Adh because the strain carried more copies of the pdcZM and adhZM genes. The specific activities of Pdc and Adh in KO11/pLOI297 were 4- and 6-times higher than those of KO11.
After 12 h, KO11/pLOI297 entered the no-growth phase when the ethanol concentration reached 28 g/L, with low concentrations of lactate and acetate. KO11/pLOI297 accumulated only 1.25 g/L lactate and 0.77 g/L acetate while KO11 produced more lactate (7.48 g/L) and acetate (3.78 g/L) at the beginning of the no-growth phase. However, when KO11/pLOI297 entered the no-growth phase, the ethanol production rate decreased while the lactate production rate increased. Similar to KO11, the carbon balance closed during the growth-inhibited phase but failed to close during the no-growth phase, which may reflect increasing cell maintenance requirements (Fig. 6D). Due to lower concentrations of acetate and lactate, KO11/pLOI297 likely directed less energy toward maintenance in the no-growth phase. Under these conditions KO11/pLOI297 achieved an ethanol yield of 0.36 ± 0.00 g/g while KO11 achieved only 0.25 ± 0.00 g/g.
Ethanol Accumulation Caused Product Inhibition That Resulted in Cessation of Growth and Decreased Ethanol Synthesis
To explain the reduction of ethanol production as KO11 and KO11/pLOI297 entered the no-growth phase, the specific activities of both Adh and Pdc were measured and compared in both the growth-associated and no-growth phases. The measured activities of both Adh and Pdc remained relatively constant in both phases, suggesting that neither of these enzymes was down-regulated at the transcriptional level (Fig. 7A). However, the activities decreased in the presence of ethanol. As shown in Figure 7B, the specific activities of Pdc and Adh in KO11/pLOI297 decreased by 60% and 40%, respectively, in the presence of 30 g/L ethanol. At this ethanol concentration, both KO11 and KO11/pLOI297 transitioned from the growth-inhibited to no-growth phases, implying that the ethanol inhibited Adh and/or Pdc, resulting in decreased ethanol production.
Continuous Fermentation of KO11
To further examine the mechanisms of ethanol toxicity, KO11 was grown in a chemostat under controlled environmental conditions. A glucose concentration of 20 g/L was used to ensure that glucose was limiting and byproduct formation was not high enough to significantly affect cellular growth and ethanol production. Following establishment of steady state, a bolus of ethanol was added to the fermentor to raise its ethanol concentration by 0, 10, or 20 g/L, and the feed medium was adjusted to contain ethanol at 0, 10, or 20 g/L. For example, the 10 g/L ethanol addition (bolus plus medium adjustment) to the vessel resulted in an ethanol concentration of 17 g/L ethanol.
Figure 8A shows the biomass concentration during continuous fermentation. The maximum specific growth rate determined from the increase in biomass during the first 6 h in continuous culture was the same for all chemostat experiments before ethanol addition, and was the same as obtained in batch culture (∼0.8 h−1). With the addition of 10 g/L ethanol the biomass concentration did not change, reaching steady state between 24 and 96 h. However, the apparent steady state was subsequently lost and oscillations in biomass, glucose, and metabolite concentrations occurred. For the case of 20 g/L ethanol addition, biomass concentration immediately decreased and then increased to follow the profile observed in 10 g/L ethanol. Furthermore, similar oscillations again appeared.
Cause of Oscillations Following Ethanol Addition
The oscillation in biomass correlates closely with the oscillations in lactate, acetate, and ethanol (Fig. 8B–D). The accumulation of both ethanol and acetate triggered the oscillation by reducing the rate of cell growth below that equivalent to the dilution rate. Consequently, the cell concentration decreased. The amplitude of acetate oscillation depended on the concentration of ethanol added. Lactate concentration increased sharply and simultaneously with the recovery in biomass concentration (Fig. 8D), primarily due to the availability of glucose and an increase in the glycolytic flux. However, the lactate concentration decreased back to its pre-oscillation value as the cells approached a new steady state.
Effect of Growth Oscillation on Ethanol Yields
Before ethanol treatment, ethanol yields were consistent among different experiments and were in the range of 0.35–0.40 (g ethanol/g glucose). Ethanol and acetate were the main fermentation products. Addition of 10 or 20 g/L of ethanol reduced the ethanol yields, by 46% or 57%, respectively (Fig. 9).
Under anaerobic growth conditions, production of acetate provides ATP for cellular growth and maintenance. Synthesis of other fermentative products such as succinate, lactate, ethanol, and formate contributes to the redox balance by recycling and controlling the synthesis of NADH (Alam and Clark, 1989). The distribution of these fermentative products depends on the E. coli strain. In the present work with E. coli KO11, a high glycolytic flux, evidenced by a high glucose uptake rate, was maintained during the growth-inhibited and no-growth phases of batch fermentation; however, the distribution of fermentative products changed significantly. In the no-growth phase, lactate production increased while ethanol production decreased, acetate production ceased, and formate consumption commenced. Formate consumption in the absence of acetate production indicated that pyruvate was not converted to acetyl CoA. Because acetyl CoA is a precursor for biomass synthesis, the failure to produce acetyl CoA resulted in a cessation of growth during the no-growth phase of batch fermentation.
In E. coli, D-lactate and ethanol share the same precursor, pyruvate (Fig. 4). The shift in metabolism of both KO11 and KO11/pLOI297 from the growth-inhibited phase to the no-growth phase resulted in more D-lactate production and less ethanol biosynthesis. This result indicates that ethanol inhibited its own production, possibly via inhibition of Pdc and/or AdhB, resulting in increased lactate production (Fig. 4B and F). The specific activities of Pdc and Adh in the presence of 30 g/L ethanol indicated significant inhibition (Fig. 7B).
Ethanol synthesis in KO11 competes for carbon with acetate and lactate-producing pathways. The high production of lactate and acetate during the growth-inhibited phase resulted from a bottleneck in the ethanol-producing pathway. This limitation was overcome by increasing gene dosages of pdcZM and adhZM in KO11 through introduction of the plasmid pLOI297. Both acetate and lactate production decreased and most of the carbon was directed toward the ethanol-producing pathway, resulting in an ethanol yield close to the theoretical limit (0.51 g/g). Even though KO11/pLOI297 secreted much less lactate and acetate during the exponential and growth-inhibited phases, the biomass production and glucose consumption rates were the same as for the parent strain KO11 (Fig. 6A–C). This implies that disruption of the lactate-producing pathway could potentially increase the carbon flux toward ethanol synthesis without affecting cell growth. In this case, ethanol could serve to recycle NADH. The removal of the acetate-production pathway could also potentially increase the ethanol yield but might decrease ethanol productivity due to the accompanying loss of ATP formed from acetyl-phosphate during acetate formation.
Although metabolic engineering can be employed to redirect carbon flux to ethanol production, improving ethanol titers and yields still faces several challenges. One major challenge is ethanol toxicity. KO11/pLOI297 was able to direct almost all carbon flux to ethanol production during the growth-inhibited phase. However, the accumulation of ethanol triggered the inhibition of Pdc and Adh, resulting in decreased ethanol production and increased formation of lactate. Therefore, end-product inhibition can be alleviated and ethanol production improved by implementing advanced fermentation techniques such as extractive fermentation (Maiorella et al., 1983) and/or by incorporating solvent pumps or engineering Pdc/Adh that are less subject to ethanol inhibition. A further challenge is the presence of other chemical inhibitors such as organic acids in the fermentation broth. For the case of KO11, even though the acetate concentration was in the range of 3–4 g/L, increased inhibition was observed in the presence of both acetate and ethanol. Moreover, the acetate-producing pathway can be removed from solventogenic microorganisms; however, the broth from biomass hydrolysates derived from lignocellulosics typically contains acetate. Thus, alleviating acetate toxicity would be useful to improve ethanol yields, titers, and productivities. Furthermore, the efficiency of an ethanologenic strain could be further increased by metabolic engineering to remove pathways that compete with ethanol production during the growth-associated phase, as demonstrated in recent reports (Trinh and Srienc, 2009; Trinh et al., 2008).