Evolution of the adhE Gene Product ofEscherichia coli from a Functional Reductase to a Dehydrogenase

The multifunctional AdhE protein ofEscherichia coli (encoded by the adhE gene) physiologically catalyzes the sequential reduction of acetyl-CoA to acetaldehyde and then to ethanol under fermentative conditions. The NH2-terminal region of the AdhE protein is highly homologous to aldehyde:NAD+ oxidoreductases, whereas the COOH-terminal region is homologous to a family of Fe2+-dependent ethanol:NAD+oxidoreductases. This fusion protein also functions as a pyruvate formate lyase deactivase. E. coli cannot grow aerobically on ethanol as the sole carbon and energy source because of inadequate rate of adhE transcription and the vulnerability of the AdhE protein to metal-catalyzed oxidation. In this study, we characterized 16 independent two-step mutants with acquired and improved aerobic growth ability on ethanol. The AdhE proteins in these mutants catalyzed the sequential oxidation of ethanol to acetaldehyde and to acetyl-CoA. All first stage mutants grew on ethanol with a doubling time of about 240 min. Sequence analysis of a randomly chosen mutant revealed an Ala-267 → Thr substitution in the acetaldehyde:NAD+ oxidoreductase domain of AdhE. All second stage mutants grew on ethanol with a doubling time of about 90 min, and all of them produced an AdhEA267T/E568K. Purified AdhEA267T and AdhEA267T/E568K showed highly elevated acetaldehyde dehydrogenase activities. It therefore appears that when AdhE catalyzes the two sequential reactions in the counter-physiological direction, acetaldehyde dehydrogenation is the rate-limiting step. Both mutant proteins were more thermosensitive than the wild-type protein, but AdhEA267T/E568K was more thermal stable than AdhEA267T. Since both mutant enzymes exhibited similar kinetic properties, the second mutation probably conferred an increased growth rate on ethanol by stabilizing AdhEA267T.

AdhE appears to be the evolutionary product of a gene fusion. The NH 2 -terminal region of this protein is highly homologous to the family of aldehyde:NAD ϩ oxidoreductases, whereas the COOH-terminal region is homologous to the family of Fe 2ϩ -dependent alcohol:NAD ϩ oxidoreductases (3)(4)(5). Despite the fact that both AdhE-catalyzed reactions are reversible, E. coli fails to grow on ethanol as a sole carbon and energy source apparently for two main reasons. First, the adhE gene is insufficiently expressed under aerobic conditions (6 -8). Second, the catalytic half-life of the AdhE protein is shortened during aerobic metabolism by a metal-catalyzed oxidation (MCO) 1 cycle. In this disabling process, the amino acid chains of AdhE are thought to be covalently attacked by the highly reactive hydroxyl radicals locally generated by the Fe 2ϩ bound to the active site of the alcohol:NAD ϩ oxidoreductase domain. In fact, AdhE has been identified as one of the major targets of protein oxidation in E. coli (9,10).
The case of the adhE gene product and its role in general fermentation is analogous to that of the fucO gene product and its role in specific L-fucose fermentation (11). Unlike AdhE, FucO is not a fusion protein but a simple enzyme that belongs to the family of Fe 2ϩ -dependent alcohol:NAD(P) ϩ oxidoreductases and catalyzes physiologically the reduction of L-lactaldehyde to L-1,2-propanediol (11)(12)(13). Like AdhE, FucO also fails to serve as a dehydrogenase for aerobic growth, because the gene is inadequately expressed, and the enzyme is highly sus-ceptible to MCO during aerobic metabolism. We have previously characterized E. coli mutants that acquired aerobic growth ability on L-1,2-propanediol by recruiting FucO to serve as a dehydrogenase (Ref. 14 and references therein). Two kinds of mutations contributed to such an ability. First, an IS5 insertion occurred in the fucO promoter that resulted in high constitutive expression of the fucAO operon (15). Second, a missense mutation occurred that conferred resistance of FucO to MCO (9,14,16).
A mutant that grew on ethanol as sole carbon and energy source was previously reported (6). However, the nature of the mutation(s) responsible for the ethanol ϩ phenotype was not definitively determined. Here we report the isolation of 16 series of independent ethanol ϩ mutants and the characterization of the genetic changes at the molecular level. Our results showed that the evolution of AdhE as a dehydrogenase followed a strategy that differs from that of FucO.

Bacterial Strains, Growth Conditions, and Preparation of Cell Extracts-
The relevant characteristics and sources of bacterial strains, plasmids, and phages used in this study are given in Table I. Luria Bertani (LB) medium containing 0.1 M MOPS and 0.2% glucose was adjusted to pH 7.4 (LB-glucose medium). Minimal medium was prepared as described previously (10) and was supplemented with 0.2% glucose or 2% ethanol as carbon and energy source. Solid media contained 1.5% Bacto-agar (Difco). Culture absorbance (A 600 ) was determined in a DU640 Beckman spectrophotometer. Aerobic 10-ml cultures were grown at 37°C in 250-ml Erlenmeyer flask shaken at 250 rpm. Anaerobic cultures were grown at 37°C in 100-ml flasks filled to the brim. For anaerobic growth on solid media, a Gas-Pak system was used. When appropriate, antibiotics (Sigma) were added at the following concentrations: 200 g ampicillin/ml and 10 g tetracycline/ml.
For enzyme assays of cell extracts, cultures during mid-exponential phase of growth were harvested and prepared as described previously (17). The cell pellets were suspended in four times their wet weight in 50 mM Tris-HCl (pH 8.5), 50 mM KCl, 5 mM dithiothreitol, 2 mM NAD, and protease inhibitors 1 mM phenylmethylsulfonyl fluoride, 0.1 mM N-tosyl-L-phenylalanine chloromethyl ketone, and 1 M pepstatin A. The suspended cells were disrupted by sonication. Samples were centrifuged at 12,000 ϫ g for 30 min at 4°C, and the supernatant was used for assays.
Enzyme Assays-To determine the level of AdhE activities, the cells were grown on glucose or ethanol minimal mineral medium. Ethanol dehydrogenase activity was assayed in 1.5 M ethanol, 2.5 mM NAD, and 300 mM potassium carbonate at pH 10.0. Acetaldehyde dehydrogenase activity was assayed in 100 mM acetaldehyde (10 mM for purified enzymes), 2.5 mM CoASH, 2.5 mM NAD, and 300 mM potassium carbonate at pH 10.0 (18). The rates of both dehydrogenase reactions were monitored by the formation of NADH at 340 nm (19). Acetyl-CoA reductase activity was assayed in 1 mM acetyl-CoA, 0.25 mM NADH, and 50 mM Tris-HCl (pH 7.5). Acetaldehyde reductase activity was assayed in 100 mM acetaldehyde (10 mM for purified enzymes), 0.25 mM NADH, and 50 mM Tris-HCl (pH 7.5) (18). The rate of both reductase reactions was monitored by the disappearance of NADH at 340 nm (19). In all of the assays 1 activity unit corresponds to 1 mol of substrate converted per min. ␤-Galactosidase activity was assayed as described previously by Miller (20).
Purification of AdhE Wild-type and Mutant Proteins-6 liters of anaerobic cultures were grown overnight at 37°C in minimal medium containing 0.2% glucose. Cultures were centrifuged and disrupted by two 1-min cycles of sonication with 1-min intervals of resting on ice. Cell-free extracts were obtained by centrifugation at 15,000 ϫ g for 30 min at 4°C. Ammonium sulfate was slowly added to the extracts being stirred and chilled on ice until 20.6% saturation was reached. The mixture was left chilled for 30 min and then centrifuged for 15 min at 4°C at 15,000 ϫ g. The supernatant fraction was recovered, and ammonium sulfate was added until 30% saturation while being stirred and chilled on ice. After 30 min of equilibration, the mixture was again centrifuged, as described above. This time the pellet was recovered and resuspended in 2.5 ml of 50 mM Tris-HCl (pH 8.5) (approximately 30 -35 mg of protein/ml). The sample was applied to an Ultrogel AcA44 gel filtration column (IBF Biotechnics, Paris, France) pre-equilibrated with 100 mM KCl, 50 mM MOPS-KOH at pH 7.65. Elution with the same buffer was carried out at a flow rate of 1.4 ml/min. Fractions showing significant AdhE activity were pooled and diluted (1:1 v/v) with 50 mM Tris-HCl (pH 8.5). The sample was then loaded into a DEAE-15HR column (Waters Associates, Milford, MA) equilibrated with 50 mM Tris-HCl (pH 8.5) and 50 mM NaCl. After a washing step of 30 min with the same buffer, elution was carried out by a linear gradient of 50 -200 mM NaCl at a flow rate of 5 ml/min for 20 min. Peak AdhE elution occurred at approximately 15 min. Purity of the samples was examined by SDS-polyacrylamide gel electrophoresis (21). Protein concentration was determined by the Bradford method (22), using bovine serum albumin as standard.
Thermal Stability Assays of AdhE-Purified wild-type and mutant AdhE proteins (0.14 mg/ml) were incubated at 37°C. Samples were withdrawn at different time intervals for assay of ethanol dehydrogenase activity.
Oxidative Inactivation of AdhE-Purified wild-type and mutant AdhE proteins (0.14 mg/ml) were incubated at 15°C either in the presence of 1 mM NADH or 2 mM ascorbate plus 50 M FeCl 3 (23).  (20). Standard methods were used for restriction endonuclease digestion and ligation of DNA (24,25). Plasmid DNA was isolated by using the QIAprep system, and the DNA fragments were isolated from agarose gels with the QIA quick kit (Qiagen). Bacteria were transformed with plasmid DNA electroporation (24) with an E. coli Pulser (Bio-Rad). Polymerase chain reaction amplifications were carried out in a Minicycler (MJ Research), using Pfu DNA polymerase from Stratagene (La Jolla, CA). Oligonucleotides were custom-synthesized (DNA Integrated Technologies). Sequence determination of adhE alleles was carried out by amplifying four different fragments (A, B, C, and D) of the adhE gene using the following primers: , and D3 (5Ј-GAAGGGGCCGTTTAT-GTTG-3Ј). Each polymerase chain reaction fragment was subjected to automated DNA sequencing at the CORE facility at Harvard Medical School.
Insertion of Wild-type and Mutant adhE Alleles into Host Chromosomes via Vectors-The plasmids pJMADH1, pJMADH4, pJMADH5, and pJMADH6 were digested with restriction enzymes BamHI and EcoRI to yield 3.8-kb fragments that contain a full-length adhE operon. These fragments were then ligated using T4 DNA ligase (Promega, Madison, WI) with wild-type DNA cut with the same enzymes. The ligation mixtures were assembled into complete phage particles by using the Gigapack Gold Lambda packaging extract (Stratagene, La Jolla. CA). Phage particles were used to transduce strain ECL3999 (adhE::kan). Single copy insertions of the adhE operon were confirmed by Southern blots probed with adhE fragments at both ends of the adhE sequence.

Selection of Mutants with Acquired Aerobic Growth Ability on Ethanol as a Sole Carbon and Energy
Source-Since direct selection on minimal ethanol medium failed to yield the desired mutants, we used a two-step approach. About 100 cells of the merodiploid strain ECL4000 (adhE ϩ ⌽[adhE-lacZ]) were spread on each of 100 MacConkey base agar (Difco) plates containing 2% ethanol. The plates were then sealed with parafilm to retard evaporation and then incubated for 24 h at 37°C. Up to this time, all the colonies were semi-transparent and colorless. After 6 days of incubation under sealed conditions, red papillae appeared on many colonies. On 5 control plates without ethanol, no red papillae appeared. Among about 10 4 colonies screened on the MacConkey-ethanol agar, 1425 colonies presented red papillae. At least one papilla from each of these 1425 colonies was streaked on the same kind of agar for purification. A single red colony from each streak was then tested for growth ability on agar containing mineral medium and 2% ethanol. For reasons undetermined, only 31 red clones were found to have an ethanol ϩ phenotype. Sixteen independent ethanol ϩ mutants were adopted for further study. The doubling time of all these mutants in liquid mineral medium containing 2% ethanol was about 240 min at 37°C. A phage P1 lysate was then prepared from each of the mutants to transduce strain ECL4000 and selected for aerobic growth on ethanol. One ethanol ϩ transductant (first stage mutants) from each transduction experiment was subjected to an additional 100 generations of selection in the same ethanol liquid medium. A sample of the cells from each line of selection was then plated again on ethanol agar, and one colony that clearly exhibited an increased growth rate was isolated. A phage P1 lysate was then prepared from each of the 16 clones with improved growth rate on ethanol to transduce strain ECL4000, and the transductants were selected for aerobic growth on ethanol. All backcrosses gave rise to the large colony size phenotype. When each of these back-crossed strains (second stage mutants) were grown at 37°C in liquid mineral medium containing 2% ethanol, all showed a doubling time of about 90 min. These results indicate that the first and second mutations are transductionally linked. A first stage mutant, JE46, and a second stage mutant, JE52, were picked as representatives for both genetic and biochemical characterizations (Table II).
Indication of cis Mutations by Enzymatic Analysis of Merodiploid Evolvants-For a preliminary cis/trans test for the two mutations, we took advantage of the adhE ⌽(adhE-lacZ) merodiploid background. A cis mutation in adhE should primarily increase the ethanol dehydrogenase activity level, whereas a trans-positive regulatory mutation should elevate the expression of both adhE and ⌽(adhE-lacZ).
When wild-type and mutant cells were grown aerobically on glucose, ethanol dehydrogenase activities were found to be 5.3-fold elevated in strain JE46 and 7.6-fold elevated in strain JE52 when compared with the wild-type level. The levels of the dehydrogenase activity were even more elevated when the mutant cells were grown aerobically on ethanol (possibly because of substrate stabilization). Curiously, the ␤-galactosidase activity levels were found to be 40% lower in strain JE46 and 46% lower in strain JE52 when compared with the wild-type level (Table II). In any event, the increase in ethanol dehydrogenase level without concomitant increase in ␤-galactosidase activity level would suggest that both mutations acted in cis. According to Leonardo and co-workers (8), the change of ␤-galactosidase activity levels in a direction opposite to that of ethanol dehydrogenase is best explained as follows. A cis mutation was responsible for elevating the dehydrogenase activity levels in the mutants JE46 and JE52. The increase in this activity (consumption of NADH coupled with the reduction of acetyl-CoA) during aerobic growth on glucose would raise the cellular redox potential. Such a state would in turn cause a decrease in the synthesis of ␤-galactosidase under the direction of the adhE promoter. The adhE promoters of Aerobacter aerogenes (now classified as Klebsiella pneumoniae) and E. coli are thought to be activated by low cellular redox states, as reflected by high NADH/NAD ratios (7,28,29). Further Evidence in Support of Mutations cis to adhE by Transduction Analysis-To confirm that the mutations in JE46 and JE52 were cis to the adhE locus situated at min 27.8 (31), we prepared a P1 vir lysate from each of the mutants to transduce strain CAG12169 that bears zci-506::Tn10 insertion at min 28 (31). If the mutation(s) responsible for the increase of adhE expression is indeed cis, then all of the transductants selected for the ability to grow on ethanol should lose the closely linked transposon Tn10 that confers tetracycline resistance. When 2000 ethanol ϩ transductants, obtained from each of the two P1 vir lysates, were analyzed for the drug resistance, all were found to be Tet s , affirming that the mutation(s) was in cis. There are several ways by which a mutation cis to adhE can increase the ethanol dehydrogenase activity level. For example, a regulatory mutation could increase the rate of transcription or translation of the gene, or a structural gene mutation may occur that enhances the catalytic activity or the half-life of the protein.
Sequence Determination of Mutant adhE Alleles-To locate the mutations, we sequenced each entire mutant gene from 1 kb upstream of the ATG codon to the end of the open reading frame (see "Experimental Procedures"). As expected, the promoter region of strains JE46 and JE52 was identical to that of the wild-type sequence previously reported (32). On the other hand, the coding region of the adhE gene in strain JE46 showed an A to G transition at position 799 from the A of the initiation codon ATG. This change converts amino acid residue 267 from Ala to Thr and is located inside the acetaldehyde dehydrogenase domain (Fig. 1). The coding region of the adhE gene in strain JE52 showed two base changes as follows: the same substitution found in strain JE46, plus an A to G transition at position 1702 that converts the amino acid residue 568 from Glu to Lys, which is located inside the ethanol dehydrogenase domain. Strikingly, when the 15 other independently isolated second stage mutants were also subjected to DNA sequence analysis, the same pair of mutations present in strain JE52 was found. It might be noted that the presence of the mutations considerably downstream of the sequence specifying the RBS site also makes translational control an unlikely mechanism for the increased ethanol dehydrogenase activity level in the mutant strains.
Reconstitution of adhE A267T and adhE A267T/E568K Mutant Alleles-To demonstrate that the mutations identified in strains JE46 (adhE A267T ) and JE52 (adhE A267T/E568K ) are sufficient to account for their phenotypes, we prepared phage bearing wild-type or each of the two mutant alleles (see "Experimental Procedures") to lysogenize strain ECL3999 (adhE::kan). The control lysogen ECL4063 (adhE::kan att::adhE ϩ ) showed aerobic and anaerobic levels of AdhE activity indistinguishable from those of the wild-type strains MC4100 and ECL4000. When strain ECL4064 (adhE::kan att::adhE A267T ) was tested, the activity levels were similar to those of strain JE46, and the cells grew on ethanol as a sole carbon and energy source at a rate similar to that of cells of JE46. Similarly, strain ECL4066 (adhE A267T/E568K ) showed the same phenotypes as strain JE52 (data not shown).
Identity of the Double Mutations in Strain JE52 and the Previously Isolated Strain DC272-Starting from an acetate auxotroph (aceF10, defective in dihydrolipoyltransacetylase component E2p), DC48, Clark and Cronan (6) isolated a nitrosoguanidine-induced mutant, DC81, that acquired the ability to use ethanol as a substitute of acetate. Strain DC81 also grew on ethanol as sole carbon and energy source. When grown aerobically on glucose and acetate, DC81 exhibited an ethanol dehydrogenase activity level more than 20-fold higher than that of DC48. It was suggested on the basis of genetic analysis that the mutation in DC81 altered a transcriptional site of the adh gene at min 27 (6), later referred to as the promoter constitutive mutation adhC (27). Strain DC272, re-examined in this study, was an ace ϩ transductant of strain DC81 bearing that adhC mutation (28). Since all of the 16-second stage mutants possessed wild-type promoters and strains JE52 and DC272 grew at the same rate on ethanol at 37°C (about 90-min FIG. 1. AdhE amino acid sequence and putative binding sites. The acetaldehyde dehydrogenase and ethanol dehydrogenase domains are connected by a proposed linker (bold italicized letters) indicated by the arrow (33). The NAD-binding site is located on the basis of the GXGXXG motif. The iron-binding site (9) is shown with its conserved residues underlined. The substituted amino acid residues (T and K) in the mutant proteins are encircled.
doubling time), we wondered what kind of mutation(s) took place in strain DC272 bearing the adhC locus (6,28). The adhC allele comprising a 1-kb stretch upstream of the initiation codon and the entire open reading frame of adhE was therefore sequenced. To our surprise, the adhE allele of strain DC272 sustained exactly the same 2 base pair changes that occurred in JE52. No base change in the promoter region was found. To confirm that these two mutations were solely responsible for the ethanol ϩ phenotype of strain DC272, we transduced the adhC allele into strain ECL4000 (the isogenic parent of JE52) to yield strain ECL4060. When strains JE52 and ECL4060 were grown in parallel aerobically on ethanol or anaerobically on glucose, the ethanol dehydrogenase activity levels were indistinguishable in the paired extracts (data not shown). We therefore conclude that the locus adhC is no other than adhE A267T/E568K allele (31).
In the case of genetic mobilization of FucO, amino acid substitutions that conferred MCO resistance improved the aerobic growth ability on propanediol (14). To test the sensitivity of AdhE proteins to MCO damage, the oxidative inactivations of the three proteins were compared under two different conditions that promote the Fenton reaction. In the first experiment, the proteins were incubated in a solution containing Fe 3ϩ as the metal ion and ascorbate as the reducing agent (Fig. 4A). In the second experiment, the inactivation was catalyzed by the enzyme bound Fe 2ϩ as the metal ion and added NADH as the reducing agent (Fig. 4B). Under both conditions, the mutant proteins were more sensitive than the wild-type protein. Significantly, AdhE A267T/E568K was less sensitive than AdhE A267T . The fact that the mutant protein with double amino acid substitutions was both more thermal stable and MCO-resistant than the mutant protein with a single mutation would suggest that the second E568K substitution stabilized the architectural integrity of the enzyme rather than improved the catalytic activity. These results are in contrast to those obtained with the FucO mutant proteins. In that case, FucO I7L was more thermal stable but less MCO-resistant than FucO I7L/L8V , thus indicating that the increased fitness of FucO for aerobic function was entirely dependent on MCO resistance.
Kinetic Analysis of Purified AdhE, AdhE A267T , and AdhE A267T/E568K Proteins-Since the mutants were selected for more rapid utilization of ethanol, we compared the purified AdhE proteins for the two substrate-oxidizing reaction rates that the proteins catalyze the specific ethanol and acetaldehyde dehydrogenase activities. The ethanol dehydrogenase specific activities of purified AdhE A267T or AdhE A267T/E568K were similar to each other but only 1.3-fold higher than that of AdhE (Table III). It might be recalled that the ethanol dehydrogenase specific activities of extracts from AdhE A267T or AdhE A267T/ E568K cells grown aerobically on glucose were 5.3-7.6-fold, respectively, higher than that of AdhE cell extracts (Table II). This discrepancy may in part be explained by significant inactivation of the mutant enzymes during the course of purification, as suggested by the stability data shown in Fig. 3. What is critical to note, however, is that the acetaldehyde dehydrogenase specific activities of purified AdhE A267T and AdhE A267T/ E568K were 5-6-fold higher than that of AdhE (Table III). To confirm the occurrence of enzyme inactivation during purification, we then examined acetaldehyde dehydrogenase specific activities of AdhE A267T and AdhE A267T/E568K in extracts of cells grown aerobically on glucose. The specific activities were about 16 -18-fold higher than that of AdhE (data not shown), supporting the notion that partial inactivation of the mutant enzymes occurred during the course of purification. Perhaps it is no coincidence that both mutant proteins contained an amino acid substitution in the acetaldehyde domain (Fig. 1). When the K m values for ethanol and acetaldehyde were examined in the same pair of reactions, the values for the mutant proteins were indistinguishable from each other but were significantly lower than those for the wild-type protein (Table III). DISCUSSION The emergence of the AdhE fusion protein was probably a turning point in the evolution of the fermentative network of an ancestor of E. coli (33). From the perspective of catalysis, the fusion of an acetaldehyde oxidoreductase and an ethanol oxidoreductase probably accelerated the successive reduction of acetyl-CoA to ethanol by bringing the two active sites in close proximity. As a corollary, the steady state level of acetaldehyde, a toxic intermediate, could probably be lowered. It should be mentioned, however, that such a condition might also be achieved by the complexing of the two separate oxidoreductases, as in Clostrium kluyveri (34,35).
Members of the family of aldehyde oxidoreductases have their NAD-binding sites near the COOH-terminal end, whereas members of the family of Fe 2ϩ -dependent alcohol oxidoreductase have their NAD-binding sites near the NH 2 -terminal end. Interestingly, a sequence analysis of AdhE of E. coli revealed only a single NAD-binding motif on the NH 2 -terminal side of the linker (Fig. 1). If indeed there is only a single NAD-binding site, it is possible that evolution of the fusion not only brought the two catalytic sites close together but also made it possible to dispense with the coenzyme-binding site of the parent alcohol oxidoreductase. The sharing of the remaining NAD-binding site could in principle greatly facilitate the sequential catalysis. An added advantage of fusing the two proteins might provide the more elaborate structure of the protein with the potential to acquire other functions, such as the deactivation of pyruvate formate lyase (4). There may well FIG. 2. SDS-polyacrylamide gel electrophoresis analysis of the AdhE proteins. Samples at various stages of purification were electrophoresed in 9% acrylamide. A, cell extract from strain MC41000; B, the preparation after the ammonium sulfate step; C, the preparation after the Ultrogel filtration step; D, the final preparation of AdhE after the liquid chromatography step; E, the final preparation of AdhE A267T following the same procedure described for AdhE; F, the final preparation of AdhE A267T/E568K following the same procedure described for AdhE.
be other accrued functions yet to be discovered. For instance, we do not yet know the biological significance of spirosomes consisting of AdhE molecules (4,32).
The fact that all 17 independent ethanol ϩ mutants characterized in this study exhibited altered AdhE structures would indicate that promoter-up mutations were either extremely rare or deleterious. Indeed, when we plated MC4100 transformant cells bearing a multicopy plasmid (pBR322 derivative) containing the adhE gene under the control of an IPTG-inducible promoter on ethanol-IPTG agar, no growth was observed. Worse yet, when the transformant cells were grown aerobically on glucose in a mineral medium, the addition of IPTG was bactericidal. Thus, an excessive concentration of the AdhE protein appears to be toxic. Even a more moderate increase in the level of the protein seems to be detrimental, since transformant cells bearing the same multicopy plasmid containing the adhE gene under the control of its own promoter failed to grow aerobically on ethanol. Moreover, such cells were growthimpaired on glucose as the sole source of carbon and energy. 2 It is not clear whether the toxicity is related to the AdhE structure or its catalytic activity. Whatever the reason, the deleterious effect of excessive levels of AdhE protein under aerobic conditions may explain why up-regulated promoter mutations were not selected.
The striking increase in the specific acetaldehyde dehydrogenase activity relative to the specific ethanol dehydrogenase activity of AdhE A267T would suggest that the second reaction was rate-limiting when AdhE was selected to catalyze the two sequential reactions in the direction opposite to the physiological one. As a consequence of the mutation, however, the protein appears to be destabilized. Since AdhE A267T/E568K exhib-2 J. Membrillo-Herná ndez, unpublished observations.  ited kinetic properties indistinguishable from those of AdhE A267T but showed increased stability in vitro, it would appear that the Glu-568 3 Lys substitution raised the steady state level of cellular acetaldehyde dehydrogenase activity by partially stabilizing AdhE A267T (however, we cannot rule out the possibility that the Glu-568 3 Lys mutation alone confers the ability to grow on ethanol). An alternative strategy to raise the steady level of cellular acetaldehyde dehydrogenase activity would be to confer resistance of the AdhE protein to MCO damage. However, we failed to isolate such a mutant. Perhaps the Ala-267 3 Thr and Glu-568 3 Lys mutations conferred such large increases in the efficacy of the novel function of AdhE that the MCO-resistant mutations conferring only modest improvements failed to be selected. Finally, it should be mentioned that Ala-267 in the acetaldehyde oxidoreductase domain and Glu-568 in the ethanol oxidoreductase domain are not invariable amino acids within their respective conserved regions (data not shown). An observation that remains to be explored is the basis for the relatively small difference in AdhE activity level of mutant and wild-type cells grown anaerobically on glucose (Table II). An even more intriguing mystery is our failure to select for AdhE A267T directly on minimal ethanol agar. 2 Mobilization of the adhE-encoded oxidoreductase protein and the fucO-encoded oxidoreductase protein for aerobic function provides two different examples for genetic adaptation. In the evolution of FucO, the first step was invariably the activation of the promoter (in 10/10 independent selections), followed by mutations in the coding region that conferred resistance to MCO (14). By contrast, in the evolution of AdhE, the first and second steps were invariably mutations in the coding region. The two biased modes of adaptation may illustrate how preexisting genetic, physiological, and biochemical contexts can predestine the channels for future evolution. The two examples also illustrate the versatility of the bacterial genome: when one evolutionary pathway is blocked, alternative routes are available.