Adaptation of Escherichia coli to Elevated Sodium Concentrations Increases Cation Tolerance and Enables Greater Lactic Acid Production

Abstract

Adaptive evolution was employed to generate sodium (Na⁺)-tolerant mutants of Escherichia coli MG1655. Four mutants with elevated sodium tolerance, designated ALS1184, ALS1185, ALS1186, and ALS1187, were independently isolated after 73 days of serial transfer in medium containing progressively greater Na⁺ concentrations. The isolates also showed increased tolerance of K⁺, although this cation was not used for selective pressure. None of the adapted mutants showed increased tolerance to the nonionic osmolyte sucrose. Several physiological parameters of E. coli MG1655 and ALS1187, the isolate with the greatest Na⁺ tolerance, were calculated and compared using glucose-limited chemostats. Genome sequencing showed that the ALS1187 isolate contained mutations in five genes, emrR, hfq, kil, rpsG, and sspA, all of which could potentially affect the ability of E. coli to tolerate Na⁺. Two of these genes, hfq and sspA, are known to be involved in global regulatory processes that help cells endure a variety of cellular stresses. Pyruvate formate lyase knockouts were constructed in strains MG1655 and ALS1187 to determine whether increased Na⁺ tolerance afforded increased anaerobic generation of lactate. In fed-batch fermentations, E. coli ALS1187 pflB generated 76.2 g/liter lactate compared to MG1655 pflB, which generated only 56.3 g/liter lactate.

INTRODUCTION

Escherichia coli is an industrially important microorganism that can be cultured to a high cell density on a large scale using simple defined media. Because the central carbon metabolism of E. coli is well understood, the microbe is widely used in metabolic engineering to generate a diverse range of products including amino acids and alcohols (e.g., ethanol and isobutanol) as well as organic acids (e.g., succinate and pyruvate). For bioprocesses using E. coli to generate organic acids, product accumulation reduces the pH and quickly hinders cell growth. In order to maintain the optimal pH for continued product formation, a base such as NaOH or KOH must be added into the system. Unfortunately, addition of a base for pH control causes the accumulation of cations (i.e., Na⁺ or K⁺). For example, in a fed-batch process to accumulate 56 g/liter pyruvate, a Na⁺ concentration of over 0.6 M was also attained (1). Similarly, greater than 0.7 M Na⁺ or K⁺ was necessary to neutralize the medium during succinic acid production using E. coli (2, 3). These high cation concentrations or the resulting osmotic stress could inhibit cell growth and limit further organic acid formation. Indeed, bacterial growth and acid production may ultimately be hindered more by the presence of Na⁺ (or K⁺, etc.) than by the organic acid itself. Unfortunately, distinguishing acid toxicity from cation toxicity at a single, optimum pH is difficult. Organic acid toxicity and the associated tolerance of E. coli have been recently reviewed (4).

Rational genetic alteration of a microorganism for a specific purpose is difficult in situations when an understanding of the relationship between phenotype and genotype is limited or when the relationship is multifaceted. Fortunately, bacteria have a remarkable ability to adapt to environmental stress, and adaptive techniques are useful when beneficial mutations can be encouraged by environmental conditions. For example, adaptation for improved strains has been used to increase substrate consumption (5), to adapt strains to defined medium (6), to perform complicated chemical syntheses (7–9), and to increase tolerance to the inhibitors found in lignocellulosic hydrolysates (10–12). Various approaches of adaptive evolution have been employed, including continuous cultivation with progressively increased feed concentrations (13) and shake flask cultures with prolonged exponential-phase growth (5, 14). Recently, a device to facilitate adaptive evolution was developed (15) and has been used to make E. coli thermophilic (16).

The objective of this study was to determine whether organic acid accumulation would increase by enhancing the sodium chloride tolerance of E. coli. Four independent isolates of E. coli were generated that could tolerate increased Na⁺ concentrations using adaptive evolution through serial transfers in media with progressively increasing NaCl concentrations. Lactic acid was selected as the model organic acid, since E. coli lacking pyruvate formate lyase activity due to a single knockout of the pflB gene accumulates significant quantities of this acid.

MATERIALS AND METHODS

Strain and growth medium.

E. coli MG1655 (F⁻ λ⁻ ilvG rfb-50 rph-1) was used in this study. The defined adaptation (DA) medium contained the following (per liter): 10 g glucose, 1.70 g citric acid, 13.30 g KH₂PO₄, 4.50 g (NH₄)₂HPO₄, 1.2 g MgSO₄·7H₂O, 13 mg Zn(CH₃COO)₂·2H₂O, 1.5 mg CuCl₂·2H₂O, 15 mg MnCl₂·4H₂O, 2.5 mg CoCl₂·6H₂O, 3.0 mg H₃BO₃, 2.5 mg Na₂MoO₄·2H₂O, 100 mg Fe(III) citrate, 4.5 mg thiamine·HCl, and 8.4 mg Na₂(EDTA)·2H₂O. This medium was supplemented with NaCl as described below, and all concentrations are reported as total Na⁺ (the sum of Na⁺ in the medium, Na⁺ for pH adjustment using NaOH, and Na⁺ from added NaCl).

Adaptive evolution.

To adapt the strains for increased Na⁺ tolerance, 10 ml of E. coli MG1655 was cultured in four independent 125-ml shake flasks in DA medium at 37°C and 250 rpm (19 mm pitch). Every 24 h, the optical density (OD) was measured, and 1 ml of the culture was transferred into 9 ml of a fresh medium. If the OD was much greater than that observed in the previous culture transfer, then the NaCl concentration was increased. This process was continued for 73 days, after which time a single colony from each culture was isolated on solid (agar) lysogeny broth (LB) medium containing 0.98 M Na⁺. These four single-colony isolates were cultured in liquid LB, then suspended in LB containing 25% glycerol, and stored at −80°C.

Strain stability and measurement of tolerance.

Strain stability was confirmed by growing an aliquot of the frozen stock for each of the four isolates (and strain MG1655 as a negative control) in DA medium without additional Na⁺, transferring once into the same medium, and then transferring into DA medium with 0.91 M Na⁺. Sodium and potassium ion tolerance was quantified by growing E. coli MG1655 and each isolate in DA medium, then transferring the culture into a series of media containing 0.77 to 1.10 M Na⁺ or K⁺ (primarily as NaCl or KCl, respectively). The maximum specific growth rate for a particular Na⁺ or K⁺ concentration was calculated using 5 to 7 OD measurements over time. Once the nominal concentration of Na⁺ or K⁺ resulting in a growth rate of approximately 0.1 h⁻¹ for each isolate was determined, we measured specific growth rate of cultures grown at that Na⁺ or K⁺ concentration in triplicate and compared the growth rates between isolates using the Student t test with P < 0.05 as the confidence interval. We also measured the growth rates of the four isolates in triplicate at approximately the same nominal Na⁺ concentration using Na₂SO₄. Sucrose tolerance was quantified similarly using 0.75 to 0.90 M sucrose.

Sequencing.

Genomic DNA was prepared from the four isolates (designated ALS1184 to ALS1187) and strain MG1655. One-quarter runs were completed on isolates ALS1186 and ALS1187 and our laboratory MG1655 strain using a 454 Life Sciences genome sequencer. Specifically, 59.3-fold coverage of the genome was achieved for MG1655, our laboratory strain, while 10.5-fold coverage of the genome was achieved for isolate ALS1186 and 21.4-fold coverage of the genome was achieved for isolate ALS1187. The genome sequences were compared to the published sequence for strain MG1655 (17) (GenBank accession number U00096), and all identified differences between isolates ALS1186 and ALS1187 and our laboratory strain were verified by PCR amplification of a 750-bp region containing the suspected mutation and conventional sequencing. Conventional sequencing was also used to determine whether the five mutations identified in ALS1186 and ALS1187 were present or absent in ALS1184 and ALS1185.

Lactate production.

The pflB knockouts MG1655 pflB and ALS1187 pflB were constructed using P1vir lysate prepared from strain NZN111 (18). The pflB::Cam deletion was transduced into strain MG1655 and isolate ALS1187, and chloramphenicol-resistant transductant colonies were selected.

For lactate production, E. coli MG1655 pflB or ALS1187 pflB was first grown in a 250-ml shake flask containing 50 ml DA medium. After 12 h, the contents were used to inoculate the 2.5-liter bioreactor (Bioflo 310; New Brunswick Scientific Co., New Brunswick, NJ, USA) containing 1.0 liter DA medium but with 20 g/liter glucose. During an initial aerobic phase of about 6 h, agitation was maintained at 400 rpm, and air and O₂ were mixed as necessary at a total flow rate of 1.0 liter/min to maintain the dissolved oxygen (DO) above 40% of saturation until the OD reached 8.0. During a second, anaerobic phase, agitation was maintained at 200 rpm, and a 9:1 mixture of N₂ and CO₂ was sparged at 0.4 liter/min. During this phase, the glucose concentration was maintained at 2 to 4 g/liter by automatically feeding a 600-g/liter glucose solution in response to the measurement of an online glucose analyzer (YSI 2700 select biochemistry analyzer; YSI Life Sciences Inc., Yellow Springs, OH, USA). An experiment was terminated when the glucose solution was not automatically fed for 3 h. For both aerobic and anaerobic phases, the pH was maintained at 7.0 using 30% (wt/vol) NaOH, and the temperature was kept at 37°C. The fermentations were run in duplicate.

Chemostats.

Continuous fermentations of 900-ml volume at several dilution rates were operated as carbon-limited chemostats and initiated in batch mode in a 2.5-liter bioreactor (Bioflo 310). The influent medium contained DA medium but with 5.0 g/liter glucose and either low Na⁺ (0.18 M) or high Na⁺ (0.68 M) concentration. A steady-state condition was assumed after four residence times at which time the oxygen and CO₂ concentrations in the effluent gas remained unchanged. For cell (dry weight) measurement, three 25.0-ml samples were centrifuged (3,300 × g, 10 min), the pellets were washed by vortex mixing with 30 ml of 0.9% saline solution, and the resulting solutions were then centrifuged again. After the washing step using deionized (DI) water was repeated twice, the cell pellets were dried at 60°C for 24 h and weighed. All fermentations were conducted at 37°C with an air flow rate of 0.5 liter/min, agitated at 400 rpm, and a pH of 7.0 by adjusting feed pH and by additionally using 30% (wt/vol) NaOH. The DO during each process remained above 40% saturation. The maintenance coefficients and true biomass yield coefficients were calculated from the slopes and intercepts of plots of dilution rate versus dilution rate/observed yield (19).

Analytical methods.

The optical density at 600 nm (OD₆₀₀) (UV-650 spectrophotometer; Beckman Instruments, San Jose, CA, USA) was used to monitor cell growth, and this value was correlated to cell mass (dry weight). The concentrations of oxygen and CO₂ in the off-gas were measured by using a gas analyzer (Innova 1313 gas monitor; Lumasense Technologies, Ballerup, Denmark), and these values used to calculate the specific rates of oxygen uptake and CO₂ evolution (q_O2 and q_CO2, respectively, in millimoles of gas per gram [dry weight] of cells per hour). The concentrations of soluble organic compounds were determined by high-performance liquid chromatography (20).

RESULTS

Adaptive evolution of E. coli MG1655.

In order to select for E. coli mutants with increased Na⁺ tolerance, four independent cultures were exposed in parallel to increasing concentrations of NaCl as described in Materials and Methods, and four isolates (ALS1184 to ALS1187) were obtained from E. coli MG1655 after 73 days of transfer. The adaptation of isolate ALS1187 is shown in Fig. 1, and the other three isolates had similar adaptation patterns (data not shown). After each isolate was thawed from frozen stock, the specific growth rates of strain MG1655 and the isolates were compared as a function of Na⁺ concentration (Fig. 2). At low Na⁺ concentration, the growth rates of MG1655 and the four isolates were statistically indistinguishable. When the Na⁺ concentration exceeded 0.4 M, the specific growth rates for the adapted strains were similar to each other but greater than the wild-type strain MG1655. All of the strains exhibited a roughly linear decrease in the maximum specific growth rate as the concentration of Na⁺ increased, though this relationship between Na⁺ concentration and growth rate was shifted 0.10 to 0.15 h⁻¹ for the four adapted isolates compared to MG1655. The highest Na⁺ concentration examined that supported growth of all four isolates was 0.98 M, while the wild-type MG1655 strain was unable to grow at a Na⁺ concentration of 0.91 M.

FIG 1.

Evolution progress of E. coli ALS1187. The OD was measured after 24 h of growth (●) in a shake flask culture and then exposed to increasing total concentration of Na⁺ (indicated by the gray background). The Na⁺ concentration is shown in moles per liter.

FIG 2.

Maximum specific growth rate (μ_MAX) of the adapted strains ALS1184 (▲), ALS1185 (▼), ALS1186 (◇), ALS1187 (●), and wild-type MG1655 (□) as a function of Na⁺ concentration.

The specific growth rates of E. coli MG1655 and the isolates were also compared over a range of K⁺ concentrations using primarily KCl (Fig. 3). While strain MG1655 and the four isolates had greater K⁺ tolerance than Na⁺ tolerance, the isolates grew significantly faster than MG1655 at K⁺ concentrations above 0.5 M. Similar to growth rates observed for high Na⁺ concentrations, all four isolates attained growth rates of about 0.10 h⁻¹ in a medium containing 1.10 M K⁺, while MG1655 could not grow at this K⁺ concentration.

FIG 3.

Maximum specific growth rate (μ_MAX) of the adapted strains ALS1184 (▲), ALS1185 (▼), ALS1186 (◇), ALS1187 (●), and wild-type MG1655 (□) as a function of K⁺ concentration.

We next determined carefully whether a statistically significant difference existed between the growth rate of the four isolates on either Na⁺ or K⁺. Table 1 shows the specific growth rates of the four isolates in DA medium containing either 0.91 M NaCl or 1.10 M KCl, concentrations at which the parent strain MG1655 could not grow. In 0.91 M NaCl, isolate ALS1187 was observed to have the greatest specific growth rate (P < 0.05), isolates ALS1185 and ALS1186 were statistically indistinguishable, and isolate ALS1184 attained the lowest specific growth rate. In 1.10 M KCl, the only significant difference was that ALS1184 and ALS1185 showed greater specific growth rates than ALS1187. Although some differences were observed in the growth rates of the four isolates under these high-salt conditions, these differences are relatively small compared to the great differences observed between any of these isolates and the parent strain MG1655, which could not grow at the highest Na⁺ and K⁺ concentrations.

TABLE 1.

Comparison of specific growth rates of four E. coli isolates

Isolate	Specific growth rate (h−1) (mean ± SD) of isolate ina:
	0.91 M Na+ (NaCl)	1.10 M K+ (KCl)	0.88 M Na+ (Na2SO4)
ALS1184	0.129 ± 0.002 c	0.109 ± 0.009 a	0.144 ± 0.004 ab
ALS1185	0.156 ± 0.003 b	0.123 ± 0.008 a	0.139 ± 0.006 b
ALS1186	0.160 ± 0.004 b	0.100 ± 0.013 ab	0.142 ± 0.001 ab
ALS1187	0.178 ± 0.000 a	0.084 ± 0.003 b	0.149 ± 0.006 a

^aComparison of specific growth rates of four E. coli isolates in medium containing glucose and 0.91 M NaCl, 1.10 M KCl, or 0.44 M Na₂SO₄ (0.88 M Na⁺). Under these conditions, the wild-type MG1655 strain showed no growth. In each column, statistically significant differences in specific growth rates are denoted by different letters (P < 0.05), with the greatest values assigned the letter “a”, etc.

We also examined whether tolerance was attributable to total ion concentration (e.g., Na⁺ plus Cl⁻), ionic strength, or specifically to cation concentration. This comparison was accomplished by growing the four isolates on DA medium containing 0.44 M Na₂SO₄, which compared to 0.91 M NaCl had only 3% lower Na⁺, 27% lower total ion concentration and 45% greater ionic strength. The observed mean growth rate of the isolates in 0.44 M Na₂SO₄ was only 7% greater than the mean growth rate of the isolates in 0.91 M NaCl (Table 1). This small difference in growth rates between similar NaCl and Na₂SO₄ medium compositions on a Na⁺ basis supports the conclusion that the Na⁺ concentration (and not total ion concentration or the ionic strength) is the primary determinant of the specific growth rate under the conditions of the experiments (Table 1).

We finally examined the growth of E. coli MG1655 and the four adapted strains in media containing 0.75 to 0.90 M sucrose. No difference in growth rate (P > 0.05) was observed between any of the four isolates and strain MG1655, with the growth rate of each measured to be 0.08 h⁻¹ at 0.87 M sucrose, while no growth was observed at 0.90 M sucrose (data not shown).

Sequencing of adapted strains ALS1186 and ALS1187.

The genomic sequences of adapted strains ALS1186 and ALS1187 were determined as described in Materials and Methods. Both ALS1186 and ALS1187 contained mutations affecting five genes, emrR, hfq, kil, rpsG, and sspA. Conventional sequencing was then completed to determine whether adapted strains ALS1184 and ALS1185 contained any of the mutations that had occurred in ALS1186 and ALS1187. The mutational changes that occurred between wild-type MG1655 and ALS1186 and ALS1187 are listed in Table 2. The five genes that were mutated, emrR, hfq, kil, rpsG, and sspA, could all potentially affect the ability of E. coli to tolerate increased Na⁺ and K⁺ concentrations. Four of the mutations that occurred were single-base changes, while the fifth was a 21-bp deletion. Table 3 indicates the conservation of the mutations that were observed between the four adapted strains. In contrast to ALS1187 and ALS1186, no mutations were observed in the hfq and kil genes in ALS1184 and ALS1185.

TABLE 2.

Mutational changes detected in the ALS1186 and ALS1187 evolved strains versus the wild-type MG1655 strain^a

Mutation position(s) (base no.)b	Gene affected	Coded protein	Mutationc	Position affected (change) or effectc,d
1416202	kil	Part of the defective Rac prophage	G to T	22 (P→T)
2808774	emrR	Regulatory protein that controls the EmrAB multidrug resistance pump	T-to-C mutation in the emrR promoter
3375245–3375266	sspA	Stringent starvation protein	DEL [CTCACGATCCACCAGGGTCGG]	DEL [60(P) 61(T) 62(L) 63(V) 64(D) 65(R) 66(E)]
3471636	rpsG	30S ribosomal subunit S7 protein	C to T	156 (W→TGA stop codon)
4398400	hfq	Global regulator	T to G	30 (I→M)

^aALS1186 proved to be an unstable strain. The kil and hfq mutations, while present when the strain is grown selectively in DA medium, are lost when ALS1186 is grown nonselectively in LB medium. We also observed that ALS1186 takes significantly longer to acclimate to DA medium than ALS1187 does.

^bPosition(s) in the E. coli MG1655 genome.

^cDEL, deletion. The sequence or amino acids deleted are shown in the brackets.

^dThe number(s) indicate the amino acid change(s) in the coded protein due to the mutation.

TABLE 3.

Conservation of mutational changes that occurred in the adapted strains ALS1184, ALS1185, ALS1186, and ALS1187 versus the wild-type strain MG1655

Gene	Mutational change in strain:
	ALS1184	ALS1185	ALS1186	ALS1187
kil	Absent	Absent	Presentb	Present
emrR	Present	Presenta	Present	Present
sspA	Present	Present	Present	Present
rpsG	Present	Present	Present	Present
hfq	Absent	Absent	Presentb	Present

^aThe mutation in the emrR promoter region in strain ALS1185 is a G-to-A change at bp 2808778. The mutation in the emrR promoter region in strains ALS1184, ALS1186, and ALS1187 is a T-to-C change at bp 2808774.

^bThe hfq and kil mutations in strain ALS1186 are rapidly lost when ALS1186 is grown nonselectively in LB medium.

Although strain ALS1186 was isolated from a single colony like the other evolved strains, sequencing confirmed an anomaly with this strain. The kil and hfq mutations, while present when the strain was grown selectively in DA medium, were rapidly lost when ALS1186 was grown nonselectively in LB media. Consistent with this observation, we also noticed that ALS1186 required a longer time to acclimate to DA medium than ALS1187 did. The fact that ALS1186 behaved differently than ALS1187 suggests the presence of another mutation in one of the two strains which was missed by genomic sequencing. Collectively, the genomic sequencing results nevertheless lead to the conclusion that the mutations in emrR, hfq, kil, rpsG, and sspA can affect the ability of E. coli to tolerate increased Na⁺ and K⁺ concentrations, since all five mutations were observed in strains ALS1186 and ALS1187 and emrR, rpsG, and sspA mutations were observed in strains ALS1184 and ALS1185.

Steady-state growth.

Although the four isolates were very similar in their Na⁺ and K⁺ tolerance, we selected isolate ALS1187 for additional studies because of its greater growth rate in 0.91 M NaCl (Table 1). First, carbon-limited chemostats were conducted at two different Na⁺ concentrations (0.18 M and 0.68 M) to compare the steady-state carbon-limited metabolism of strains MG1655 and ALS1187. The results showed the expected linear relationship between the specific glucose consumption rate (q_S) and the specific growth rate (Fig. 4). However, the results demonstrated a difference in the glucose consumption rates between high and low Na⁺ concentration and between strain MG1655 and isolate ALS1187.

FIG 4.

Specific glucose consumption rate as a function of specific growth rates (i.e., dilution rate) for E. coli MG1655 and ALS1187 at low and high Na⁺ concentrations in steady-state continuous culture. The specific glucose consumption rate (q_S) of E. coli MG1655 at 0.18 M Na⁺ (○) and at 0.68 M Na⁺ (△) and of E. coli ALS1187 at 0.18 M Na⁺ (●) and at 0.68 M Na⁺ (▲) is shown in grams of glucose per gram (dry weight) of cells per h.

For E. coli MG1655, growth in medium having 0.68 M Na⁺ consistently required a 2- to 2.5-fold-greater glucose consumption rate than the identical growth rate in the same medium having 0.18 M Na⁺. For isolate ALS1187, growth in medium having 0.68 M Na⁺ required a similarly greater glucose consumption rate than growth in medium having 0.18 M Na⁺. However, to support any given specific growth rate at either 0.18 M or 0.68 M Na⁺, ALS1187 required a 5 to 40% greater glucose consumption rate than MG1655. The difference between glucose consumption rates of MG1655 and ALS1187 was greatest at 0.18 M Na⁺ and at the greater growth rates. Since steady-state glucose consumption rate is directly related to energy production (and demand), these results suggest that some characteristic of the adapted isolate ALS1187 led to a greater ATP requirement than MG1655 during growth.

The maintenance coefficient is the non-growth-related substrate consumption and because its value is at least partly due to cellular requirements for osmoregulation (21), we speculated that a high Na⁺ concentration would correlate with high maintenance and that an adapted strain would exhibit lower maintenance. Similarly, the true biomass yield (i.e., after subtracting for maintenance) is a measure of the maximal biomass that can be generated from the carbon source, and we anticipated that in the presence of elevated Na⁺, the biomass yield should be reduced. Using a linear regression of specific glucose uptake rate versus specific growth rate (Fig. 4), values for the maintenance coefficient (m_S) and for the biomass yield (Y_X/S) were obtained (Table 4; see Table S1 in the supplemental material). For both strains, Na⁺ concentration strongly affected maintenance and the adapted strain ALS1187 and strain MG1655 showed similar maintenance coefficients at 0.18 M Na⁺. However, at 0.68 M Na⁺, the maintenance coefficient for ALS1187 was 30% lower than the maintenance coefficient for MG1655, and both values were over twice as great at the higher Na⁺ concentration. Furthermore, for growth of either strain at 0.68 M Na⁺, the true biomass yield decreased by about 50% compared to the value observed at 0.18 M Na⁺.

TABLE 4.

Non-growth-related energy metabolism and the biomass yield on glucose for wild-type MG1655 and adapted strain ALS1187 at low and high Na⁺ concentrations

Strain	Na+ (M)	mS (g/g · h)a	YX/S (g/g)b
MG1655	0.18	0.031	0.591
	0.68	0.111	0.287
ALS1187	0.18	0.033	0.445
	0.68	0.078	0.233

^aThe maintenance coefficient (m_S) is shown in grams of glucose/grams (dry weight) of cells · hour and is a measure of non-growth-related energy metabolism.

^bThe biomass yield on glucose (Y_X/S) is shown in grams (dry weight) of cells/gram of glucose.

The rates of oxygen consumption and CO₂ evolution indicate the metabolic rates, and we were interested to learn whether E. coli ALS1187 and MG1655 showed different patterns of carbon metabolism. The specific oxygen uptake rate (q_O2) generally correlated proportionately with q_S, with no difference observed between strains MG1655 and ALS1187 or at different Na⁺ concentrations (Fig. 5A). The range of glucose consumption rates examined in this study was rather low, and below the threshold glucose consumption rate for overflow metabolism reported previously in strain MG1655 (22). Hence, the values for q_O2 observed in this study are below the maximum plateau q_O2 for wild-type E. coli. The specific carbon dioxide evolution rate (q_CO2) also increased proportionately with q_S but did not depend on the strain or on the concentration of Na⁺ (Fig. 5B). Note that in Fig. 5A and B, the independent variable is specific glucose consumption rate, not the specific growth rate. Because q_S itself is a function of growth rates (Fig. 4), the values for q_CO2 and q_O2 did vary with growth rate and for each strain. Consistent with lower biomass yields and maintenance coefficients, ALS1187 always showed greater CO₂ evolution at any growth rate. For example, at the highest growth rate studied (0.14 h⁻¹) with 0.18 M Na⁺, MG1655 attained a q_CO2 of 4.6 mmol/g · h, while ALS1187 attained a q_CO2 of 9.6 mmol/g · h. However, the differences observed in q_CO2 at any given growth rate can be attributed to differences in q_S between strains or Na⁺ concentrations. In other words, growth at the higher Na⁺ concentration caused a greater generation of CO₂ solely because at an elevated Na⁺ level, the glucose consumption rate was greater. Collectively, these results confirm that ALS1187 showed a greater rate of carbon metabolism than MG1655 did.

FIG 5.

Specific rates (q, in millimoles of gas per gram [dry weight] of cells per hour) as a function of specific glucose consumption rate (q_S) for E. coli MG1655 and ALS1187 at low and high Na⁺ concentrations in steady-state continuous culture. E. coli MG1655 at 0.18 M Na⁺ (○) and at 0.68 M Na⁺ (△) and E. coli ALS1187 at 0.18 M Na⁺ (●) and at 0.68 M Na⁺ (▲) are shown. (A) Specific oxygen uptake rate (q_O2); (B) specific carbon dioxide production rate (q_CO2).

Lactate production.

We hypothesized that one potential benefit of increased salt tolerance in E. coli would be an ability to accumulate a higher concentration of an organic acid product while maintaining the same controlled pH. E. coli with a knockout in the pflB gene encoding pyruvate formate lyase readily accumulates lactic acid (1, 23). In order to determine whether the adapted isolate could accumulate a higher concentration of lactate, two strains, E. coli MG1655 pflB and ALS1187 pflB were compared in a two-phase aerobic-anaerobic process. In both cases during the second lactate production phase, the glucose concentration was automatically maintained at 2 to 4 g/liter, and each process was terminated when the online glucose monitor demanded no additional glucose. Lactate generation also called for the automatic addition of NaOH to maintain the pH, and therefore, Na⁺ accumulated during the processes. For strain ALS1187 pflB, the lactate concentration reached 76.2 g/liter, 35% greater than the 56.3 g/liter concentration achieved for MG1655 pflB (Fig. 6). Moreover, for strain ALS1187 pflB, the final Na⁺ concentration was calculated to be 0.97 M, whereas for strain MG1655 pflB, the Na⁺ concentration reached 0.88 M. The fermentation time using ALS1187 pflB was prolonged 8 h compared to the fermentation using MG1655 pflB. Based on the total glucose consumed, ALS1187 attained a lactate yield of 0.77 g/g, while MG1655 attained a lactate yield of 0.64 g/g. A duplicate set of fed-batch fermentations with MG1655 pflB and ALS1187 pflB yielded essentially identical results. This study demonstrates that a strain adapted for increased cation tolerance, after some additional metabolic engineering to promote the production of a specific organic acid, can accumulate an elevated concentration of the acid product.

FIG 6.

Lactate concentration and Na⁺ concentration in fed-batch fermentations using E. coli MG1655 pflB (white symbols) and ALS1187 pflB (black symbols). In each case, cells were grown aerobically for 6 h, and then anaerobic conditions commenced during which lactate accumulated. The lactate concentration (triangles) and Na⁺ concentration (circles) were monitored until glucose was no longer consumed.

DISCUSSION

In order to enhance the Na⁺ tolerance of E. coli, adaptive evolution was conducted with the wild-type strain MG1655 in media containing increasing amounts of NaCl. Four independent isolates, ALS1184 to ALS1187, adapted from E. coli MG1655 showed enhanced maximum specific growth rates at elevated Na⁺ and K⁺ concentrations but showed no difference in growth rates at elevated sucrose concentrations. Though the four strains were similar, the derivative with the highest Na⁺ tolerance, ALS1187, was further characterized for steady-state growth parameters and, after deletion of the pflB gene, for the improved ability to accumulate the acid product lactate in fed-batch fermentations. The results indicate that the adaptation of MG1655 enhanced Na⁺ tolerance compared to the wild type and that the adaptation significantly improved lactate production.

E. coli MG1655 and ALS1187 showed markedly different glucose consumption rates and maintenance coefficients under steady-state conditions. The maintenance coefficient is the non-growth-related substrate consumption (24), caused by (i) shifts in metabolic pathways, (ii) proofreading of protein and turnover of mRNA, (iii) cell death and lysis, and (iv) osmoregulation (21, 25, 26). In this study, strains MG1655 and ALS1187 exhibited similar maintenance coefficients at a low Na⁺ concentration (0.18 M) but the coefficients were significantly different at a high Na⁺ concentration (0.68 M). Specifically, the maintenance coefficient for ALS1187 was 30% less than the maintenance coefficient for MG1655 at this elevated Na⁺ concentration. A lower maintenance implies that ALS1187 is much more efficient at nongrowth metabolism under conditions of high Na⁺ concentration than wild-type MG1655. Sachidanandham et al. (27) similarly reported that in a hyperosmotic medium, a Bacillus strain isolated from the desert showed a lower maintenance than wild-type Bacillus thuringiensis. Of course, the maintenance coefficient describes metabolism in the absence of growth and may not be directly predictive of an elevated maximal growth rate or the ultimate tolerance of a strain to Na⁺ or K⁺. Because maintenance is a measure of the cell's energy expenditure to survive in its environment (21), it comes as no surprise that as a consequence of improved salt tolerance, the adapted strains have a reduced maintenance in the presence of high Na⁺ concentrations.

An important result of enhanced Na⁺ tolerance in E. coli ALS1187 was the ability in fed-batch fermentation to attain a significantly greater final concentration of a model organic acid, lactic acid. Although not optimized for lactate production, E. coli with just a single knockout of the pflB gene encoding pyruvate formate-lyase (PFL) will accumulate significant lactate in an anaerobic phase which has followed aerobic growth (1, 23). PFL is required for growth of E. coli under anaerobic conditions (28) but is inactive under aerobic conditions (29). Thus, not only would one expect that the pflB knockout should have no effect on aerobic growth in the presence of high Na⁺ or K⁺ concentrations, but in the present study, the cells did not encounter a high Na⁺ concentration until lactate had accumulated in a nongrowth anaerobic phase. While strain ALS1187 tolerated only 10 to 15% more Na⁺ (or K⁺) than strain MG1655 did as measured by the increased growth rates under aerobic conditions (Fig. 2 and 3), strain ALS1187 pflB was able to accumulate 35% more lactate than strain MG1655 pflB did. At neutral pH, lactic acid exists dissociated as (sodium) lactate, and Na⁺ ion must be added in tandem with the formation of lactic acid to maintain the pH. Increased Na⁺ tolerance was able to prolong the fermentation time despite the very limited growth of the pflB mutants during lactate formation. Moreover, the Na⁺ concentrations were very similar for the two strains until the final 12 h of the processes, even though the lactate concentration was greater in the ALS1187 process essentially throughout the nongrowth phase. Thus, we do not directly attribute the elevated lactate formation to improved growth rate in the presence of cations (“tolerance”) but rather to fundamentally different behavior of ALS1187 in a high Na⁺ environment. Liu et al. (13) similarly isolated an osmotolerant Torulopsis glabrata which resulted in enhanced pyruvate production. Despite nongrowth conditions, lactate formation ceased when the Na⁺ concentration reached about 0.04 to 0.05 M below the Na⁺ concentration limit we observed for growth for either MG1655 or ALS1187 (Fig. 2).

A surprising result of our studies is that adaptive evolution for Na⁺ tolerance using NaCl simultaneously conferred significantly increased tolerance to K⁺. However, the adaptive evolution did not confer increased tolerance to sucrose. Moreover, we observed only a small difference (7%) in growth rates in media containing Na₂SO₄ compared to NaCl at about the same Na⁺ concentration of 0.9 M, despite larger differences in ionic strength (45%) and total ion concentration (27%). Thus, while we cannot rule out an additional “anion effect,” the mutations appear to be specific to providing tolerance to increased cation concentration, rather than increased anion concentration, osmotolerance, or increased overall ionic strength.

E. coli has vastly different mechanisms for regulating intracellular Na⁺ and K⁺. E. coli has two Na⁺/H⁺ antiporter membrane proteins, NhaA and NhaB, and both play important roles in osmotic homeostasis (30, 31). These antiporters are driven by the electrochemical proton gradient (i.e., H⁺_out > H⁺_in) generated by the primary proton pumps. In E. coli, NhaA encoded by nhaA (previously designated ant) is the archetypal Na⁺/H⁺ antiporter and is required for survival in the presence of 700 mM Na⁺ at a pH of 6.8 (31). Because NhaA and NhaB are not involved in K⁺ efflux in E. coli (32, 33), any mutation in the nhaA and nhaB genes resulting from the present adaptive evolution strategy would be expected to affect only Na⁺ tolerance, not K⁺ tolerance. Wu et al. (34) showed that overexpression of both nhaA and the regulator nhaR allowed increased lactate formation in E. coli, although this combination had no effect on cell growth rate.

Intracellular K⁺ concentrations are regulated through numerous K⁺ uptake and efflux systems. K⁺ uptake is mediated by Kdp and TrkD/Kup systems (35–37), while K⁺ efflux is mediated by KefB and KefC (38), Kch channels (39), MscK channels (40, 41), as well as MdfA and ChaA transporters (32, 33, 42). MdfA and ChaA normally become significant only at alkaline pH and are able to extrude K⁺ under inwardly or outwardly directed K⁺ concentration gradients (32, 33). The K⁺ ion is typically accumulated under conditions of low osmolarity.

Though none of the five genes in E. coli ALS1187 shown by sequencing to contain a mutation (emrR, hfq, kil, rpsG, and sspA) are known to be directly involved in either Na⁺ or K⁺ transport, each of these mutations could impact the ability of E. coli to tolerate Na⁺ or K⁺. The mutation in sspA was a 21-base deletion, while the mutations that affected emrR, hfq, kil, and rpsG were all single-base substitutions. Interestingly, we observed that strain ALS1186 contained the identical mutations as ALS1187, but ALS1186 rapidly lost the hfq and kil mutations when the strain was grown in rich medium, while the ALS1184 and ALS1185 strains lacked the hfq and kil mutations entirely.

The emrR gene codes for the EmrR regulatory protein which controls the EmrAB multidrug resistance (MDR) pump. EmrR functions by binding to the promoter region of the emrRAB operon and preventing transcription of itself as well as emrA and emrB (43, 44). The T- to-C mutation that occurred at base 2808774 of strain ALS1187 is within the promoter region of emrRAB and thus could potentially affect the binding of EmrR. The EmrAB MDR pump belongs to a family that includes the AcrAB and EmrE MDR pumps (45) and is responsible for the extrusion of the antibiotics nalidixic acid and thiolactomycin and other antimicrobial agents, such as salicylate and 2,4-dinitrophenol (DNP) (43). The better-characterized AcrAB MDR pump is activated by a number of global stresses, including Na⁺ (46, 47). Thus, MDR pumps could be involved in the extrusion of other compounds in addition to antimicrobial agents.

The hfq gene codes for the Hfq protein, a well-known global regulator that can affect gene expression either positively or negatively (for a recent review, see reference 48). The disruption of Hfq causes a number of pleiotropic changes, including osmosensitivity. Numerous functions of Hfq have been elucidated. For example, Hfq can facilitate the interaction of regulatory small noncoding RNAs (sRNAs) with their target mRNA as well as affect mRNA degradation by binding to poly(A) mRNA tails. Hfq also impacts the well-known global regulatory proteins Rho and RpoS. Hfq causes transcriptional antitermination of Rho terminated mRNAs, and it is essential for the posttranscriptional regulation of the RpoS sigma factor, which is involved in a number of stress responses, including osmosensitivity.

Hfq is a Sm-like (LSm) protein that forms a homohexameric ring containing two surfaces that interact with RNA. AU-rich RNA sequences and sRNAs bind to the proximal face of Hfq, while A-rich RNA binds to the distal face of Hfq. Additionally, both the proximal and distal faces of Hfq are involved in the regulation of RpoS (49). The T-to-G mutation that occurred at base 4398400 of E. coli ALS1187 changes amino acid number 30 of Hfq from an isoleucine to a methionine and would impact the distal face of Hfq. Link et al. showed that the amino acids at positions 25, 26, 30, and 32 on the distal face of Hfq were the key amino acids for the binding of poly(A)-RNA and that if amino acid 30 was changed to an alanine or aspartic acid, the ability of Hfq to bind poly(A)-RNA was dramatically reduced (50). Interestingly, E. coli adapted to grow under glucose-limited conditions, another well-established stress, contained a single mutation that changed amino acid 25 of the Hfq distal face from a tyrosine to an aspartic acid, which provided several potential benefits to the cell, including increased stability in the glucose uptake gene ptsG and elevated outer membrane permeability (51). Since osmotic stress is known to reduce glucose transport (52), it is possible that the isoleucine-to-methionine change in ALS1187 at amino acid 30 enhances the cell's tolerance to increased Na⁺ in a manner analogous to this tyrosine-to-aspartic acid change at amino acid 25 which improves the ability of E. coli to grow under conditions of glucose limitation.

The kil gene codes for the Kil protein, which is part of the defective Rac prophage. Kil has been shown to inhibit FtsZ, an essential cellular division protein (53). FtsZ is a tubulin-like cytoskeletal protein that initiates the divisome complex, which coordinates the synthesis and division of the septum peptidoglycan at the beginning of cell division (for a recent review, see reference 54). It is well-known that the peptidoglycan sacculus is a key factor in maintaining the osmotic stability of E. coli. The G-to-T mutation that occurred at base 1416202 of E. coli ALS1187 changes amino acid number 22 of Kil from a proline to a threonine and thus could dramatically alter the functionality of the Kil protein. It has been postulated that prophages such as Rac are induced during cell stress (55), and the aberrant Kil protein in strain ALS1187 may prevent the inhibition of FtsZ that would normally occur during excess Na⁺.

The rpsG gene codes for the 30S ribosomal subunit S7 protein, which plays a critical role in ribosome assembly and function. In addition to anchoring the 16S rRNA, the S7 protein acts as a translational feedback repressor, regulating its own synthesis as well as that of other ribosomal proteins (56). The E. coli K-12 S7 protein contains a 23-amino-acid extension at the carboxyl terminus that is not found in most other members of the family Enterobacteriaceae (57), including E. coli B, C, and W strains. This extension causes the E. coli K-12 S7 protein to be tagged for degradation by the transfer-messenger RNA (tmRNA) system that rescues stalled ribosomes. The C-to-T mutation that occurred at base 3471636 of E. coli ALS1187 changes amino acid 156 of S7 from a tryptophan to a TGA stop codon, and thus, the S7 protein in ALS1187 lacks the last 24 amino acids and is virtually identical to the S7 protein found in E. coli B, C, and W strains. This mutation could confer a significant advantage to the ALS1187 ribosomes, and ribosomes are known to play an important role in the cell's response to various stresses (58).

The sspA gene codes for the stringent starvation protein SspA, an RNA polymerase-associated protein that is important for the cell's ability to survive stationary phase and enables the cell to respond to environmental stresses. SspA appears to act by inhibiting the nucleoid-associated protein HN-S, a global regulator, which functions primarily as a repressor (59). HN-S represses several genes involved in responding to environmental stresses, including low pH and high osmolarity (60). Crystallographic analysis has shown that SspA assumes the characteristic fold of glutathione S-transferase (GST), although SspA lacks GST activity and does not bind glutathione (61). The deletion of bases 3375245 to 3375266 in strain ALS1187 results in the deletion of amino acids 60 to 66 from the SspA protein. These amino acids encompass the β3 beta sheet, which is a highly conserved region among all SspA homologues and is normally involved in glutathione binding in GST. The deletion of the β3 beta sheet would be expected to dramatically alter the functionality of the SspA protein.

Surprisingly, two identical mutations (rpsG and sspA) were found in each of the four independent isolates that underwent the adaptive evolution process. This result is similar to the observations of others who used adaptive evolution but obtained identical mutations in independent isolates (62). Although the two mutations were not found in the original MG1655 strain used to begin the adaptation, it is possible that infrequent variants that carried these mutations were selected from the original population during the early stage of the process, with other mutations then subsequently appearing. Nevertheless, each of the five mutations in strain ALS1187 occurred in genes that could impact global cellular functions, while no mutation occurred in genes normally associated with Na⁺ transport (e.g., nhaA, nhaB). This result is consistent with our observation that ALS1187 also showed a marked increase in K⁺ tolerance despite the fact that the adaptive evolution selection process only involved increasing Na⁺ concentrations. Furthermore, the observation of 35% more lactate production under identical conditions by E. coli ALS1187 pflB compared to MG1655 pflB demonstrates not only the importance of salt tolerance in limiting organic acid accumulation in E. coli but also the importance of genes that impact global regulation on the ultimate industrial utility of E. coli in organic acid production.