m

r

gg

Metabolic Engineering 4, 182–192 (2002)doi:10.1006/mben.2001.0220, available online at http://www.idealibrary.co

Metabolic Engineering throughEffects on Metabolic Flux Red

Ka-Yiu San,* , ‡ George N. Bennett,† Susana J. BeEmily Horton,† Fred B. Rudolph,† Be

*Department of Bioengineering, ‡Department of Chemical EnInstitute of Biosciences and Bioengineerin

Received October 19, 2001; accepted Novembe

Applications of genetic engineering or metabolic engineering haveincreased in both academic and industrial institutions. Most currentmetabolic engineering studies have focused on enzyme levels and onthe effect of the amplification, addition, or deletion of a particularpathway. Although it is generally known that cofactors play a majorrole in the production of different fermentation products, their rolehas not been thoroughly and systematically studied. It is conceivablethat in cofactor-dependent production systems, cofactor availabilityand the proportion of cofactor in the active form may play an impor-tant role in dictating the overall process yield. Hence, the manipula-tion of these cofactor levels may be crucial in order to furtherincrease production. We have demonstrated that manipulation ofcofactors can be achieved by external and genetic means and thesemanipulations have the potential to be used as an additional tool toachieve desired metabolic goals. We have shown experimentally thatthe NADH/NAD+ ratio can be altered by using carbon sources withdifferent oxidation states. We have shown further that the metabolitedistribution can be influenced by a change in the NADH/NAD+ ratioas mediated by the oxidation state of the carbon source used. Wehave also demonstrated that the total NAD(H/+) levels can beincreased by the overexpression of the pncB gene. The increase in thetotal NAD(H/+) levels can be achieved even in a complex medium,which is commonly used by most industrial processes. Finally,we have shown that manipulation of the CoA pool/flux can be usedto increase the productivity of a model product, isoamyl acetate.© 2002 Elsevier Science (USA)

INTRODUCTION

The production of chemicals by biocatalysts has attainedconsiderable attention. Technological advances for making

181096-7176/02 $35.00© 2002 Elsevier Science (USA)All rights reserved.

genetically engineered strains, improvements in experimen-tal measurement, and theoretical analysis of metabolicfluxes have contributed to the production of high valueadded products (for example, recombinant proteins fordiagnostic and therapeutic uses) and the commercial intro-duction of new processes for the manufacture of certainnovel chemicals (examples using engineered Escherichia coli

on

Cofactor Manipulation and Itsistribution in Escherichia coli

rríos-Rivera,‡ Ravi V. Vadali,* Yea-Tyng Yang,†

na Sariyar,* and Kimathi Blackwood†

ineering, and †Department of Biochemistry and Cell Biology,, Rice University, Houston, Texas 77005

r 26, 2001; published online February 27, 2002

are the production of indigo dye by Genencor and 1,3-pro-panediol by DuPont). The production of lactate, biopoly-mer PHB, and ethanol using metabolic engineering tech-niques is currently being pursued with commercial interest.

Metabolic engineering has the potential to considerablyimprove process productivity by manipulating the through-put of certain pathways. However, despite its powerful andprecise nature, the current status of metabolic engineering isstill hindered by the lack of our full understanding ofcellular metabolism. The aspects of integrated dynamicsand overall control structure are the common obstacles forthe optimal design of pathways to achieve a desired goal.

Applications of genetic engineering or metabolicengineering have increased in both academic and industrialinstitutions and the area has been reviewed (reviews:Stephanopoulos and Vallino, 1990; Bailey, 1990; Carmeronand Chaplen, 1997; Stephanopoulos et al., 1998; Lee andPapousakis, 1999). Most current metabolic engineeringstudies have mainly focused on manipulating enzyme levelsthrough the amplification, addition, or deletion of a par-ticular pathway. However, cofactors play an essential rolein a large number of biochemical reactions and their mani-pulation has the potential to be used as an additional tool toachieve desired metabolic engineering goals. Furthermore,it will also provide an additional means to study cellularmetabolism, in particular the interplay between cofactorlevels/fluxes and metabolic fluxes.

In this project, we focus on two common yet very impor-tant cofactors, nicotinamide adenine dinucleotide (NAD)and acetyl-CoA. NAD functions as a cofactor in over 300oxidation–reduction reactions. The NADH/NAD+ cofac-

2

⁄

tor pair plays a major role in microbial catabolism, in whicha carbon source, such as glucose, is oxidized using NAD+

and producing reducing equivalents in the form of NADH.It is critically important for continued cell growth andproduct formation that this reduced NADH be oxidized toNAD+ and a redox balance be achieved. Under anaerobicgrowth, and in the absence of an alternate oxidizing agent,

FIG. 1. Anaerobic central metabolic pathway of

the regeneration of NAD+ is achieved through fermenta-tion by using NADH to reduce metabolic intermediates(Fig. 1). Some previous efforts to manipulate NADH levelshave included the use of the NADH oxidase system (Lopezde Felipe et al., 1998), the addition of an electron dye carrier(Park and Zeikus, 1999), or the variation of the oxidore-duction potential conditions (Riondet et al., 2000).

Another important cofactor is coenzyme A, CoA, and its

Metabolic Engineering—Cofactor Manipulation

derivative acetyl-CoA. Acetyl-CoA is an essential interme-diate in many energy-yielding metabolic pathways and is asubstrate in enzymatic production of industrially usefulcompounds such as esters and lipid molecules.

The long-term goals of this project are threefold: (1) toimplement these design concepts in a production strain; (2)to characterize their effects on host-cell physiology and

18

E. coli (showing NADH/NAD+and acetyl-CoA).

metabolic patterns; and (3) to test their effects on processproductivity in three different model systems.

This article describes the conceptual designs of variousapproaches that may lead to a change in the intracellularcofactor pools and/or fluxes. We also describe three differ-ent model systems that can be used in future studies to testthe potential advantages of cofactor perturbation onprocess productivity. Furthermore, we present some early

Metabolic Engineering 4, 182–192 (2002)doi:10.1006/mben.2001.0220

positive results based on these design strategies. We showexperimentally that the intracellular NAD(H/+) pool, aswell as its ratio within a host strain, can be perturbed; wefurther show that the host strain may respond to these per-turbations by shifting its metabolic pattern. Finally, wedemonstrate by using two host strains with different CoAlevels that manipulation of the CoA pool/flux can be used

3

to increase the productivity of a model product, isoamylacetate.

MODEL SYSTEMS

Three model systems were chosen to study the feasibilityof improving process productivity through cofactor mani-pulations. These three systems are the biosynthesis of 1,2-propanediol, ester isoamyl acetate, and succinate (Figs.2a–2c). All three systems require either NADH/NAD+ oracetyl-CoA in their synthesis pathways.

Model System for NAD(H/+) Study

The production of 1,2-propanediol (1,2-PD) was used asa model system because its biological synthesis depends onNADH (Fig. 2a). The precursor dihydroxyacetone-P is acommon glycolysis intermediate. It can be converted tomethylglyoxal by the enzyme methylglyoxal synthase(MGS), from which it is converted to D-lactate through theglyoxalase system (Cooper, 1984). This pathway is normallynot very active in E. coli due to the cytotoxicity ofmethylglyoxal (MG) (Ferguson et al., 1998). However, MGcan be converted to 1,2-PD by a series of two reductions. Inthis study, the production of 1,2-PD was achieved byoverexpressing MGS to increase the flux to MG and adehydrogenase (glycerol dehydrogenase). Data on thissystem are not presented here, but the system is undergoingfurther study.

Model System for CoA Study

The biosynthesis of isoamyl acetate was selected as amodel system because it uses acetyl-CoA as a substrate.Isoamyl acetate is not normally formed by wild-type E. coli.However, a genetically engineered E. coli strain bearing analcohol acetyltransferase (ATF)-expressing plasmid grownin the presence of externally added isoamyl alcohol has beenshown to be able to form isoamyl acetate (see Results andDiscussion). Under anaerobic conditions, E. coli formsacetyl-CoA from pyruvate mainly via the pyruvate formatelyase pathway. The newly introduced ester-producingpathway competes directly with the existing ethanol- andacetate-forming pathways for acetyl-CoA (Fig. 2b). It isthus speculated that manipulation of cellular metabolism to

Metabolic Engineering 4, 182–192 (2002)doi:10.1006/mben.2001.0220

increase CoA availability may enhance ester production. Anexample of this strategy is presented.

Model System for NAD(H/+) and CoA Study

The biosynthesis of succinate was also chosen as a modelsystem because of the involvement of both cofactors

18

San et al.

FIG. 2. (a) 1,2-Propanediol biosynthesis pathway (showing involve-ment of NADH/NAD+); MGS, methylglyoxal synthase; GLDH, glyceroldehydrogenase. (b) Ester biosynthesis pathway (including competitivepathways at the acetyl-CoA node); AAT, alcohol acetyltransferase; PFL,pyruvate formate lyase. (c) Succinate biosynthesis pathway (showinginvolvement of NADH/NAD+and acetyl CoA).

4

o

FIG. 3. Schematic showing differences in

NADH/NAD+ and acetyl-CoA. The biosynthesis of suc-cinate from phosphoenolpyruvate (PEP) requires NADH.Two moles of NADH are consumed per mole of succinateformed (Fig. 2c). Note that the enzyme phosphoenolpyru-vate carboxylase activity is activated by acetyl-CoA (Izui etal., 1981; Terada et al., 1991). It is therefore anticipated thatboth NADH/NAD+ and acetyl-CoA may play an impor-tant role in the synthesis rate of succinate. Experimentaldata on the effect of these manipulations on succinate pro-duction are not presented here but the system is currentlyunder study.

NADH/NAD+ MANIPULATION

NADH/NAD+Manipulation Strategies

Two distinct strategies can be used to alter theNAD(H/+) pool and NADH/NAD+. The first approach isbased on the use of carbon sources with different oxidationstates. The other approach is based on genetic manipula-tions of the host cell. It should be noted that these strategies

Metabolic Engineering—Cofactor Manipulation

could be used in conjunction with the more traditionalpathway manipulation approaches to achieve even morepronounced results.

Carbon source. In the first strategy, we used three dif-ferent carbon sources as a simple way to manipulate thecellular NADH/NAD+ ratio. It is noted that the oxidation

18

xidation of glucose, sorbitol, and gluconate.

state of glucose is 0 and those of sorbitol and gluconateare − 1 and+1, respectively. Figure 3 highlights the differ-ences in the oxidation of these carbon sources as they enterthe glycolysis pathway. Sorbitol produces more reducingequivalents in the form of NADH than glucose while glu-conate produces the least amount of NADH since half ofevery gluconate molecule goes directly to pyruvate, skippingthe NADH-producing step in glycolysis (Fig. 3). It is thusexpected that these three different carbon sources will havea significant effect on cellular NADH/NAD+ ratio andsubsequently on the metabolic patterns.

Genetic manipulation (enhancing the pyridine nucleotidesalvage pathway). The second strategy targets the geneticmanipulation of the host cell. Two different approaches canbe used; the first approach aims at increasing the totalNAD(H/+) pool while a second approach focuses onchanging the NADH/NAD+ ratio. It is also conceivablethat a combination of these two approaches may lead toboth an increased NAD(H/+) pool and an increased ratio.E. coli maintains its total NAD(H/+) intracellular pool

by synthesizing NAD through the de novo pathway and the

Metabolic Engineering 4, 182–192 (2002)doi:10.1006/mben.2001.0220

pyridine nucleotide salvage pathway.The salvage pathwayrecycles intracellular NAD breakdown products and pre-formed pyridine compounds from the environment, such asnicotinic acid (NA). The enzyme nicotinic acid phosphori-bosyltransferase (NAPRTase; EC 2.4.2.11), encoded by thepncB gene, catalyzes the formation of nicotinate mono-nucleotide, a direct precursor of NAD, from NA (Fig. 4a).

5

C

FIG. 4. (a) Biosynthesis of nicotinic acid mononucleotide. (b)

This reaction is believed to be the rate-limiting step in theNAD salvage pathway. However, the product of the nadR

Metabolic Engineering 4, 182–192 (2002)doi:10.1006/mben.2001.0220

gene regulates both biosynthetic pathways by negatively-controlling the transcription of the nadA and nadB genes,whose products catalyze the initial steps of the de novo bio-synthetic pathway and the transcription of the pncB gene inthe salvage pathway (Foster et al., 1990).

Based on these findings, we investigated the feasibility ofincreasing the total NAD(H/+) pool by substantially

18

onstruction of an nadR-insensitive NAPRTase expression system.

enhancing the salvaging pathway. This can be achieved byoverexpressing NAPRTase. Direct cloning of the pncB gene

San et al.

on a multicopy plasmid may not be sufficient due to thetranscriptional repression of NadR on pncB (Foster et al.,1990). Further manipulations have to be taken to ensure itsoverexpression: (1) by creating a mutation in the operatorregion where the repressor protein binds (Fig. 4b) or (2) byputting the transcription of the pncB gene under the controlof another promoter, such as the lac promoter system.

6

Results and Discussion

Carbon Source

The feasibility of using carbon sources (glucose, sorbitol,and gluconate) with different oxidation states as a meansto manipulate the intracellular NADH/NAD+ ratio wasinvestigated by performing a series of chemostat experimentsunder anaerobic conditions. In addition, the subsequenteffect on the redistribution of metabolic fluxes was examined.Chemostat mode of operation was chosen because it allowsthe determination of the concentrations of NADH andNAD+ and the metabolic fluxes under steady-state condi-tions, which generally provides a better data quality. Cal-culation of the specific metabolic fluxes at a given growthrate can be obtained by simply measuring the metaboliteconcentrations, cell density, and growth rate (dilution rate).Chemostat experiments have been successfully and routinelyemployed in our laboratories to study the effects of geneticand environmental perturbations on metabolic flux redistri-bution in E. coli (Aristidou et al., 1999; Yang et al., 1999;Berríos-Rivera et al., 2000). It should also be emphasized thatthe chemostat mode allows fixing the specific growth rate foreach carbon source by fixing the dilution rate. As a result, itprovides a better comparison than batch reactor runsbecause any growth rate effect due to the use of differentcarbon sources is eliminated.

Strain and medium. The strain used in this study isGJT001(pACYC184) (Tolentino et al., 1992), a derivativeof MC4100. Luria-Bertani broth (LB) medium supple-mented with 110 mM glucose, sorbitol, or gluconate wasused for the chemostat runs. To reduce the initial lag timethat occurs under anaerobic conditions, 1 g/L NaHCO3was added to the LB medium. The medium was alsosupplemented with 25 mg/L tetracycline and 30 ml/L anti-foam 289 (Sigma).

Bioreactor conditions. The fermentations were carriedout under anaerobic chemostat conditions at a dilution rateof 0.1 h−1 in a 2.5-L bioreactor (New Brunswick Scientific;Bioflo III) initially with 1.3 L of medium during theanaerobic batch stage and maintained at 1.20-L workingvolume for the anaerobic chemostat stage. The pH, tem-perature, and agitation were maintained at 7.0, 32°C, and250 rpm, respectively. A constant flow of nitrogen (10–12

Metabolic Engineering—Cofactor Manipulation

ml/min) was maintained through the fermentor headspaceto establish anaerobic conditions. The continuous culturereached steady state after four to six residence times.Samples were taken during the steady-state phase.

Analytical technique. The procedures for the determi-nation of cell dry weight, metabolite concentrations (HPLC

18

and GC), and the gases CO2 and H2 (GC) have been pre-viously reported (Aristidou et al., 1999; Yang et al., 1999;Berríos-Rivera et al., 2000).

The NADH and NAD+ were extracted by collecting asample from the reactor sample port and immediatelypipetting 1 ml of the sample into two microcentrifuge tubes(one for NAD+ and one for NADH determination). The1 ml samples were centrifuged in a microcentrifuge (BeckmanMicrofuge E) at 15,000 rpm for 1 min ( 12,535g). Thesupernatant was removed and 300 ml of 0.2 M HCl (forNAD+ extraction) or 0.2 M NaOH (for NADH extraction)was added to the pellets to resuspend them. The sampleswere placed in a 50°C water bath for 10 min and then on iceto cool them to 0°C. The extracts were neutralized byadding 300 ml of 0.1 M NaOH (for NAD+ extraction) or300 ml of 0.1 M HCl (for NADH extraction) dropwise whilevortexing. The cellular debris was removed by centrifugingat 15,000 rpm for 5 min ( 12,535g). Supernatants weretransferred to new tubes and stored at − 20°C for no morethan 24 h.

The intracellular NADH and NAD+ concentrations weremeasured by a very sensitive cycling assay (Leonardo et al.,1996; Bernofsky and Swan, 1973). Figure 5 shows aschematic representation of the reactions involved in theassay. The cycling assay was performed using a reagentmixture consisting of equal volumes of 1.0 M Bicine buffer(pH 8.0), absolute ethanol, 40 mM EDTA (pH 8.0), 4.2 mMMTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazoliumbromide; thiazolyl blue), and twice the volume of 16.6 mMPES (phenazine ethosulfate), previously incubated at 30°C.The following volumes were added to 1-ml cuvettes: 50 mlneutralized extract, 0.3 ml water, and 0.6 ml reagentmixture. The reaction was started by adding 50 ml of yeastADHII (500 or 100 U/ml in 0.1 M Bicine (pH 8.0) buffer).The absorbance at 570 nm was recorded for 10 min at 30°C.The assay was calibrated with 0.01–0.05 mM standardsolutions of NADH and NAD+. The slope (DA/min) of thelinear region of the absorbance versus time plot wascorrelated to the concentration of coenzyme (mM) by alinear fit equation. This equation was used to determine theconcentration (mM) of NAD(H/+) in the samples (extracts)from their slope (DA/min). The concentration in micro-moles of NAD(H/+) per gram of dry weight was obtainedby multiplying the concentration of the extract by the cor-responding dilution factor (0.6 ml extract/1 ml culture),

Metabolic Engineering 4, 182–192 (2002)doi:10.1006/mben.2001.0220

dividing by the dry weight (g dry wt/L of culture), andconverting the millimoles to micromoles.

Results

Table 1a shows the steady-state concentrations of variousmetabolites from the carbon source experiments asmeasured by the HPLC. Table 1b shows the calculated

7

FIG. 5. Cyclic assay: NAD+and NADH. ADH, alcohol dehyd

metabolic fluxes in mmol/(g dry wt × h) and NAD(H/+)concentrations in mmol/g dry weight. In these experiments,the NADH/NAD+ ratio increased from 0.75 for glucose to0.94 for sorbitol, which is more reduced and can thereforeproduce more reducing equivalents in the form of NADH.In a similar way, the NADH/NAD+ ratio decreased to 0.51when gluconate, which is more oxidized than glucose, wasused as the carbon source. These changes in the NADH/NAD+ ratio affected the distribution of metabolic fluxes inE. coli as is reflected in the ethanol to acetate ratio. Thisratio increased from 1.00 with glucose to 3.62 with sorbitoland decreased to 0.29 with gluconate. It is evident fromthese results that the cell adjusts its partitioning at the acetyl-CoA node by changing the ethanol- (consumes 2 NADH) to-acetate (consumes no NADH) ratio to achieve a redoxbalance. Therefore, a change in the ethanol-to-acetate ratiocan be used as an indirect indicator of a change in theNADH/NAD+ ratio.

It is interesting to note that the lactate flux, which con-sumes 1 NADH, was affected in a different way by the oxi-dation state of the carbon source. In this case, the lactateflux was highest for gluconate and lowest for sorbitol. Theseresults in combination with the ethanol results indicate thatsince lactate formation (1 NADH) is less NADH demand-ing than ethanol formation (2 NADH), it is preferred as ameans to regenerate NAD+ under more oxidizing condi-tions. On the other hand, when there is an excess of reducingequivalents, ethanol formation is preferred since it providesa faster route to NAD+ regeneration.

These chemostat experiments show that feeding similarcarbon sources with different oxidation states can be used tomanipulate the NADH/NAD+ ratio, which is consistentwith our earlier report (Berríos-Rivera et al., 1999). In

TABLE 1a

Metabolite Concentrations (mM) in an Anaerobic Chemostat(Average of Three Samples) at D=0.1 h−1

Gluconate Glucose Sorbitol

Metabolic Engineering 4, 182–192 (2002)doi:10.1006/mben.2001.0220

Substrate consumed 113.26 ± 6.30 108.17 ± 1.02 122.18 ± 1.68Succinate 7.79 ± 1.38 10.55 ± 1.54 7.82 ± 0.30Lactate 40.29 ± 4.11 8.26 ± 1.51 2.76 ± 0.18Formate 40.37 ± 1.98 48.61 ± 1.76 65.61 ± 0.46Ethanol (Et) 35.62 ± 0.86 83.32 ± 1.22 156.83 ± 2.56Acetate (Ac) 124.86 ± 1.51 83.32 ± 1.32 43.27 ± 0.95Et/Ac 0.29 1.00 3.62

18

rogenase; MTT(H) measured spectrophotometrically at 570 nm.

addition, the oxidation state of the carbon source has adirect effect on the distribution of metabolic fluxes in E.coli, especially in those requiring NADH such as ethanol,lactate, and acetate (by its relationship with ethanol).

The method described here can also be used as a simpleway of determining whether a particular pathway thatrequires NADH is limited by this cofactor. For example, if aproduct requires NADH for its synthesis, feeding a morereduced carbon source such as sorbitol will normallyimprove its yield if the system is cofactor limited. Theopposite is also generally true. This information can helpdetermine which strategy to use when trying to improve theproduction of cofactor-dependent chemicals.

pncBOverexpression

We have examined the effect of overexpressing the pncBgene from Salmonella typhimurium on the total levels ofNAD and the NADH/NAD+ ratio in E. coli. Preliminaryexperiments were performed withE. coli strains DH10B andGJT001 transformed with the plasmids pPNCBIN andpPNCBIM, which contain the native or mutated pncB genein the pBCSK+ backbone, or with pBCSK+ as a control.These six strains were grown anaerobically in glass vialssealed with an open-top cap and a septum. The mediumconsisted of 40 ml LB medium supplemented with 2%glucose, 1 g/L NaHCO3, and 34 mg/L chloramphenicol.After inoculation of 10 ml from an overnight culture, the

TABLE 1b

Anaerobic Chemostat Results D=0.1 h−1

(Equal to Specific Growth Rate at Steady State)

Gluconate(+1) Glucose(0) Sorbitol(−1)

Substrate uptake 6.61 5.10 6.47Succinate 0.45 0.50 0.41Lactate 2.35 0.39 0.15Formate 2.35 2.29 3.48Ethanol 2.08 3.93 8.31Acetate 7.28 3.93 2.29Et/Ac 0.29 1.00 3.62

San et al.

NADH 4.57 4.17 6.45NAD+ 8.94 5.53 6.88NADH/NAD+ 0.51 0.75 0.94Total NAD(H/+) 13.5 9.7 13.3

Note. Metabolic fluxes in mmol/(gDW × h) and NAD(H/+) con-centrations in mmol/g DW. Oxidation states of carbon sources are pre-sented in parentheses. DW, dry weight.

8

L

it

r

TAB

Results of Anaerobic-Tube Experiments w

DH10B DH10B(pBCSK) (pPNCBIN) (p

OD600mm 1.200 1.305NAD+(mmol/g DW) 14.49 22.22NADH(mmol/g DW) 17.57 18.26NADH/NAD+ratio 1.21 0.82Total NAD(H/+) 32.06 40.48% change in total — 26NAD(H/+)

Note. Percentage change in total NAD(H/+) is based on the concentrespectively. DW, dry weight.

tubes were incubated in a reciprocal shaker maintained at250 rpm and 37°C. When the cultures reached an OD600 of0.5 to 1.0, samples were collected with a syringe through theseptum and processed immediately for determination ofNADH and NAD+ levels.

Table 2 summarizes the results of these experiments. Thelevels of NAD+ and NADH are expressed in mmol/g dryweight (DW) after converting the OD600 nm measurements todry weight using the conversion factor 3 OD=1 g DW/L.For the DH10B strain, the total NAD(H/+) levels increasedwith the overexpression of the native pncB gene compared tothe control strain (about 26%) and they increased evenfurther with the overexpression of the mutated gene (about2.16 times), last row of Table 2. Similar results were obtainedfor the GJT001 strain; the total NAD(H/+) levels increasedby 34% with overexpression of the native pncB gene.However, overexpression of the mutated gene led to anincrease of only 16%. The NADH/NAD+ ratios do notexhibit a consistent trend among the different constructs/strains. The lack of an apparent trend may be due to the factthat these experiments were performed in anaerobic tubes,where the growth environment is not controlled. It is wellknown that the NADH/NAD+ ratio is very sensitive togrowth rate as well as other environmental factors. Note thatthe sampling optical densities for the GJT001 strains werelower than those of the DH10B. For this reason it is neces-sary to perform more batch and/or chemostat experiments.Nevertheless, the current results suggest that the totalNAD(H/+) levels can be increased by the overexpression ofthe S. typhimurium pncB gene. These results are consistent

Metabolic Engineering—Cofactor Manipulation

with an earlier study with E. coli pncB gene in a semisyntheticmedium (Wubbolts et al., 1990). The current study alsodemonstrated that the increase in the total NAD(H/+) levelscan be achieved even in a complex medium, which iscommonly used by most industrial processes.

In addition, experiments were performed under anaero-bic chemostat and anaerobic tube conditions to investigate

18

E 2

h Plasmids Overexpressing the pncBGene

DH10B GJT001 GJT001 GJT001PNCBIM) (pBCSK) (pPNCBIN) (pPNCBIM)

0.702 0.57 0.486 0.56131.39 7.89 8.16 11.8137.85 12.27 18.82 11.511.21 1.55 2.31 0.98

69.24 20.16 26.99 23.32116 — 34 16

ation of the control experiments, DH10B(pBCSK) and GJT001(pBCSK),

the effect of overexpressing the pncB gene under the controlof the native, mutated, or lac promoter on the total NADlevels, the NADH/NAD+ ratio, and the distribution ofmetabolites in E. coli. Overexpression of the pncB gene inchemostat experiments increased the total NAD levels,decreased the NADH/NAD+ ratio, and did not signifi-cantly redistribute the metabolic fluxes. However, underanaerobic tube conditions, overexpression of the pncB geneled to a significant shift in the metabolic patterns as evi-denced by a decrease in lactate production and an increaseas high as twofold in the ethanol-to-acetate ratio. Theseresults suggest that under chemostat conditions the totallevel of NAD is not limiting and the metabolic rates arefixed by the system at steady state. On the other hand, undertransient conditions (such as those in batch cultivation) theincrease in the total level of NAD can increase the rateof NADH-dependent pathways (ethanol) and thereforechange the final distribution of metabolites (Berríos-Riveraet al., 2001).

COENZYME A PERTURBATION

CoAPerturbation Strategies

Figure 6a shows the involvement of pantothenate kinase(PanK) in the biosynthesis of CoA. Two important featuresshould be noted: (1) pantothenate kinase is an upstream rate-limiting step in the acetyl-CoA biosynthesis pathway (Rocket al., 2000) and (2) recent results indicated that pantothe-nate kinase is repressed by CoA and acetyl-CoA (Rock etal., 2000). We illustrate the cofactor-manipulating strategy

Metabolic Engineering 4, 182–192 (2002)doi:10.1006/mben.2001.0220

by using a strain with an increased CoA pool presumablyfrom overexpression of pantothenate kinase, an enzymeinvolved in the limiting step in the CoA biosynthesispathway (Vallari and Rock, 1987). Further extension ofthis work would employ a plasmid-based system foroverexpression of CoA biosynthetic enzymes and poten-tially a feedback-insensitive pantothenate kinase.

9

FIG. 6. (a) Schematic showing the involvement of pantothenate kinaseduced ester-forming pathway and the plasmid construct pTAAT carrying the

Results and Discussion

The effect of altered CoA levels/fluxes on metabolic fluxdistribution and ester formation was investigated by usingtwo strains, DV62 and DV79 (Vallari and Rock, 1987), thathave different pantothenate kinase activities. DV79 is aspontaneous revertant of the temperature-sensitive DV62strain. DV79 has a higher CoA pool than DV62 (Vallari and

Metabolic Engineering 4, 182–192 (2002)doi:10.1006/mben.2001.0220

Rock, 1987).The CoA level of two strains, DV62 and DV79, was

examined under both aerobic and anaerobic conditions.These two strains were grown aerobically in 250-ml flasksand anaerobically in 40-ml amber-colored glass vials sealedwith an open-top cap and a septum. The medium consistedof 35-ml LB medium supplemented with 2% glucose and an

19

in the CoA biosynthesis pathway. (b) Schematic showing the newly intro-ATF2 gene from yeast.

appropriate amount of ampicillin. After inoculation with10 ml overnight culture, the flasks or the tubes were incubatedin a reciprocal shaker maintained at 250 rpm and 30°C.Samples were collected and processed immediately when thecultures reached an OD600 of about 0.5 to 1.0. The results ofthese experiments are shown in Table 3. The CoA concentra-tion of DV79 is 149% higher than that of the DV62 strain.

The effect of CoA perturbation on isoamyl acetate pro-

San et al.

duction was also investigated. For this purpose a plasmidencoding ATF2 (Yoshimoto et al., 1999) was constructed,pTAAT, and was transformed into these two strains (Fig.6b). The final metabolite concentrations for both strainsbearing the pTAAT plasmid grown in LB medium withoutand with 10 mM isoamyl alcohol supplementation in 35-mlanaerobic tubes are shown in Tables 4a and 4b, respectively.

0

TABLE 3

Concentration of CoA of DV62 and DV79 in Aerobic Shake Flaskand Anaerobic Tube Cultures

DV62 DV79CoA concentration CoA concentration % change

(nmol/mg) (nmol/mg)

Aerobic 0.49 ± 0.11 1.22 ± 0.14 149Anaerobic 0.51 ± 0.06 1.15 ± 0.22 124

It should be noted that under normal conditions the newlyintroduced pathway would not function since E. coli doesnot produce isoamyl alcohol, i.e., isoamyl acetate would beformed only when isoamyl alcohol is added externally.

The final biomass concentrations for the two strainsbearing the pTAAT plasmid without isoamyl alcoholsupplementation were very similar. Both the final acetateand the final ethanol concentrations for the DV79 strainwere slightly higher than those of the DV62 strain.However, the overall metabolite distributions remainedrelatively unperturbed; most of the changes are within a 5%level (Table 4a). These results suggest that the degree ofchange in the CoA concentration/flux, as mediated by thesetwo strains, has very little effect on the normal host cellmetabolism.

The final biomass concentrations for the two strainsbearing the pTAAT plasmid are still very similar, even whenisoamyl alcohol was added. Measured metabolite concen-trations, such as glucose, lactate, and succinate, remainrelatively unchanged. However, a significant increase,about 33%, in the final isoamyl acetate concentration wasobserved in the DV79 strain. These results suggest that theproductivity of isoamyl acetate may be improved byincreasing the CoA pool/flux.

SUMMARY

We have outlined several strategies that can be used tomanipulate the pools/fluxes of two common yet importantcofactors, NAD(H/+) and acetyl-CoA. We have alsodiscussed three potential model systems that can be adaptedto study the effects of these cofactor manipulations.

In this article, we have shown experimentally that the

Metabolic Engineering—Cofactor Manipulation

NADH/NAD+ ratio can be altered by using carbonsources with different oxidation states. We have furthershown that the metabolite distribution can be influenced bya change in the NADH/NAD+ ratio as mediated by theoxidation state of the carbon source used. We have alsodemonstrated that the total NAD(H/+) levels can beincreased by the overexpression of the pncB gene. The

19

TABLE 4

Final Metabolite Concentrations of DV62(pTAAT) andV79(pTAAT) after 24 h Cultivation

DV62(pTAAT) DV79(pTAAT) % change

(a) Medium without isoamyl alcohol supplement

OD 1.416 ± 0.03 1.393 ± 0.03 − 1.6Glucose consumed (mM) 25.49 ± 6.38 25.64 ± 5.08 0.6Succinate (mM) 40.29 ± 4.39 4.56 ± 0.08 3.9Lactate (mM) 29.42 ± 0.39 28.50 ± 0.36 − 3.1Acetate (mM) 8.79 ± 0.18 9.28 ± 0.15 5.6Ethanol (mM) 10.40 ± 0.23 10.80 ± 0.09 3.8

(b) Medium supplement with 10 mM isoamyl alcohol

OD 1.34 ± 0.03 1.27 ± 0.07 − 5.2Glucose consumed (mM) 32.89 ± 1.55 32.89 ± 0.98 0Succinate (mM) 3.92 ± 0.26 3.92 ± 0.24 0Lactate (mM) 27.48 ± 0.48 27.71 ± 0.27 − 0.8Acetate (mM) 7.92 ± 0.25 8.31 ± 0.26 4.9Ethanol (mM) 8.92 ± 0.77 9.60 ± 0.33 7.6Isoamyl acetate (mM) 0.70 ± 0.04 0.93 ± 0.02 32.8

increase in the total NAD(H/+) levels can be achieved evenin a complex medium, which is commonly used by mostindustrial processes. Finally, we have shown that mani-pulation of the CoA pool/flux can be used to increase theproductivity of isoamyl acetate.

In summary, we have shown in this article that mani-pulation of cofactors can be achieved by genetic means andthese manipulations have the potential to be used as anadditional tool to achieve desired metabolic engineering goals.

ACKNOWLEDGMENTS

This work was supported in part by the National Science Foundation(BES0000303 and BES01030103), by USDA (2002-35505-11638), byRobert Welch Foundation grants (C-1372 and C1268), and by a NationalScience Foundation Graduate Fellowship (S. J. Berríos-Rivera).

REFERENCES

Alam, K. Y., and Clark, D. P. (1989). Anaerobic fermentation balance ofEscherichia coli as observed by in vivo nuclear magnetic resonance spec-troscopy. J. Bacteriol. 171, 6213–6217.

Aristidou, A. A., Bennett, G. N., and San, K.-Y. (1999). Metabolic fluxanalysis of Escherichia coli expressing the Bacillus subtilis acetolactatesynthase in batch and continuous cultures.Biotechnol. Bioeng. 63, 737–749.

Metabolic Engineering 4, 182–192 (2002)doi:10.1006/mben.2001.0220

Bailey, J. E. (1991). Toward a science of metabolic engineering. Science252, 1668–1675.

Bernofsky, C., and Swan, M. (1973). An improved cycling assay for nico-tinamide adenine dinucleotide.Anal. Biochem. 53, 452–458.

Berríos-Rivera, S. J., Lu, B., Yang, Y.-T, San, K.-Y., and Bennett, G. N.(1999). The effect of NADH recycling on the NADH/NAD+ ratio andmetabolic flux redistribution in Escherichia coli. 1999 AIChE AnnualMeeting, Dallas, TX.

1

Berríos-Rivera, S. J., San, K.-Y., and Bennett, G. N. (2001). The effect ofNAPRTase overexpression on the total levels of NAD, the NADH/NAD+ ratio, and the distribution of metabolites in Escherichia coli.Submitted for publication.

Berríos-Rivera, S. J., Yang, Y.-T, San, K.-Y., and Bennett, G. N. (2000).Effect of glucose analog supplementation on metabolic flux distributionin anaerobic chemostat cultures of Escherichia coli. Metab. Eng. 2,149–154.

Cameron, D. C., and Chaplen, F. W. R. (1997). Development in metabolicengineering.Curr. Opin. Biotechnol. 8, 175–180.

Cooper, R. A. (1984). Metabolism of methylglyoxal in microorganisms.Annu. Rev.Microbiol. 38, 49–68.

Ferguson, G. P., Tötemeyer, S., MacLean, M. J., and Booth, I. R. (1998).Methylglyoxal production in bacteria: Suicide or survival? Arch.Microbiol. 170, 209–219.

Foster, J. W., Park, Y. K., Penfound, T., Fenger, T., and Spector, M. P.(1990). Regulation of NAD metabolism in Salmonella typhimurium:Molecular sequence analysis of the bifunctional nadR regulator and thenadA–pnuC operon. J. Bacteriol. 172, 4187–4196.

Izui, K., Taguchi, M., Morikawa, M., and Katsuki, H. (1981). Regulationof Escherichia coli phosphoenolpyruvate carboxylase by multiple effec-tors in vivo. II. Kinetic studies with a reaction system containingphysiological concentrations of ligands. J. Biochem. 90, 1321–1331.

Lee, S. Y., and Papoutsakis, E. T. (1999). ‘‘Metabolic Engineering,’’Dekker, New York.

Leonardo, M. R., Dailly, Y., and Clark, D. P. (1996). Role of NAD inregulating the adhE gene of Escherichia coli. J. Bacteriol. 178,6013–6018.

Lopez de Felipe, Kleerebezem, F. M., de Vos, W. M., and Hugenholz, J.(1998). Cofactor engineering: A novel approach to metabolic engineer-ing in Lactococcus lactis by controlled expression of NADH oxidase.J. Bacteriol. 180, 3804–3808.

Park, D. H., and Zeikus, J. G. (1999). Utilization of electrically reduced

Metabolic Engineering 4, 182–192 (2002)doi:10.1006/mben.2001.0220

Printed in Belgium

19

neutral red by Actinobacillus succinogenes: Physiological functionof neutral red in membrane-driven fumarate reduction and energyconservation. J. Bacteriol. 181, 2403–24109.

Riondet, C., Cachon, R., Wache, T., Alcaraz, G., and Divies, C. (2000).Extracellular oxidoreduction potential modifies carbon and electronflow inEscherichia coli. J. Bacteriol. 182, 620–626.

Rock, C. O., Calder, R. B., Karim, M. A., and Jackowski, S. (2000).Pantothenate kinase regulation of the intracellular concentration ofcoenzyme A. J. Biol. Chem. 275, 1377–1384.

Stephanopoulos, G. N., Aristidou, A. A., and Nielson, J. (1998). ‘‘Meta-bolic Engineering,’’ Academic Press, San Diego.

Stephanopoulos, G. N., and Vallino, J. J. (1990). Network rigidity andmetabolic engineering in metabolite overproduction. Science 252,675–1681.

Terada, K., Murata, T., and Izui, K. (1991). Site-directed mutagenesis ofphosphoenolpyruvate carboxylase from E. coli: The role of His579 in thecatalytic and regulatory functions. J. Biochem. 109, 49–54.

Tolentino, G. J., Meng, S. Y., Bennett, G. N., and San, K.-Y. (1992). A pH-regulated promoter for the expression of recombinant proteins inEscherichia coli. Biotechnol. Lett. 14, 157–162.

Vallari, D. S., and Rock, C. O. (1987). Isolation and characterizationof temperature-sensitive pantothenate kinase (coaA) mutants ofEscherichia coli. J. Bacteriol. 169, 5795–5800.

Wubbolts, M. G., Terpstra, P., Beilen, J. B. V., Kingma, J., Meesters,H. A. R., and Witholt, B. (1990). Variation of cofactor levels inEscherichia coli. J. Biol. Chem. 265, 17665–17672.

Yang, Y.-T, Bennett, G. N., and San, K.-Y. (1999). The interactions of nuoand ack/pta mutations on metabolic flux redistribution in Escherichiacoli. Biotechnol. Bioeng. 65, 291–297.

Yoshimoto, H., Fujiwara, D., Momma, T., Tanaka, K., Sone, H.,Nagasawa, N., and Tamai, Y. (1999). Isolation and characterization ofthe ATF2 gene encoding alcohol acetyltransferase II in the bottomfermenting yeast Saccharomyces pastorianus. Yeast 15, 409–417.

San et al.

2