Regeneration of Adenosine Triphosphatefrom Glycolytic Intermediates forCell-Free Protein Synthesis

Dong-Myung Kim, James R. Swartz

Department of Chemical Engineering, Stanford University, Stanford,California 94305-5025, USA; telephone: 650-723-5398; fax: 650-725-0555;e-mail: swartz@chemeng.stanford.edu

Received 14 September 2000; accepted 28 January 2001

Abstract: A new approach for adenosine triphosphate(ATP) regeneration in a cell-free protein synthesis systemis described. We first show that pyruvate can be used asa secondary energy source to replace or supplement theconventional secondary energy source, phosphoenol py-ruvate (PEP). We also report that glucose-6-phosphate,an earlier intermediate of the glycolytic pathway, can beused for ATP regeneration. These new methods providemore stable maintenance of ATP concentration duringprotein synthesis. Because pyruvate and glucose-6-phosphate are the first and last intermediates of the gly-colytic pathway, respectively, the results also suggestthe possibility of using any glycolytic intermediate, oreven glucose, for ATP regeneration in a cell-free proteinsynthesis system. As a result, the methods describedprovide cell-free protein synthesis with greater flexibilityand cost efficiency. © 2001 John Wiley & Sons, Inc. Biotech-nol Bioeng 74: 309–316, 2001.Keywords: cell-free protein synthesis; chloramphenicolacetyltransferase; phosphoenol pyruvate; pyruvate;acetylphosphate; inorganic phosphate; glycolysis; glu-cose-6-phosphate

INTRODUCTION

Cell-free protein synthesis is an attractive alternative to theconventional technologies for protein production such asbacterial fermentation and cell culture (Jermutus et al.,1998; Kim and Swartz, 1999; Rattan and Christensen, 1990;Stiege and Erdmann, 1995). In contrast to in vivo geneexpression methods, where protein synthesis is carried outin the context of cell physiology and is surrounded by cellwalls and membranes, cell-free protein synthesis provides acompletely open system that allows direct access to thereaction conditions. At the same time, most of the cell func-tions other than protein synthesis need not be maintained inthe cell-free system. Thus, it can be optimized with signifi-cantly wider latitude than for a living organism. For ex-ample, such reaction parameters as pH, redox potential, andionic strength can be measured directly and changed with-out concern for harmful effects on the growth and viability

of cells. In addition, the products of cell-free protein syn-thesis are less likely to affect continued productivity.

Probably the most promising potential of cell-free syn-thesis can be found in its suitability for high throughputexpression systems. The demand for such systems has be-come greater than ever due to the successful progress of thehuman genome project. It would be nearly impossible tokeep pace with the exponentially growing amount of geneticinformation with current expression technologies. In addi-tion to the extensive amount of time and labor required forcloning, transformation, and cell culture, conventional invivo expression systems cannot be easily set up for simul-taneous expression of multiple genes. In contrast, cell-freeprotein synthesis, when conducted as a batch reaction, canbe easily expanded into a multiplexed format. For example,expression of different proteins can be done in a multiwellplate simply by adding different plasmids or polymerasechain reaction (PCR) products to each well containing themixture of cell extract and other reagents required for pro-tein synthesis. Furthermore, biological activities of synthe-sized proteins can be determined directly with the samereaction plate. Cell-free systems also allow the direct isola-tion of products to shorten the time required for preparingpurified proteins (Alimov et al., 2000).

However, all of the potential advantages can be realizedonly when batch, cell-free synthesis provides acceptableand consistent product yields. To date, batch cell-free sys-tems have not produced proteins in significant amounts.This is mainly due to the short duration of the synthesisreaction. In general, protein synthesis stops within 20 min ina batch cell-free system derived fromE. coli (Kim andSwartz, 2000a). We have investigated the factors causingthe cessation of protein synthesis and have shown that rapiddepletion of adenosine triphosphate (ATP) is one of themajor reasons for the early termination of protein synthesis(Kim and Swartz, 1999, 2000a, 2000b).

In most cases, the ATP supply for protein synthesis isaccomplished by using secondary energy sources containinga high-energy phosphate bond. Conventionally, such com-pounds as phosphoenol pyruvate (Pratt, 1984), creatinephosphate (Anderson et al., 1983), and acetylphosphate(Ryabova et al., 1995) have been used for the regeneration

Correspondence to:J. R. SwartzContract grant sponsor: Charles Lee Powell Foundation

© 2001 John Wiley & Sons, Inc.

of ATP in combination with the enzymes pyruvate kinase,creatine kinase, and acetate kinase, respectively. These en-zymes catalyze the transfer of the high-energy phosphatebonds of the secondary energy sources to adenosine diphos-phate (ADP) that is generated from ATP turnover duringprotein synthesis. However, the use of these conventionalsecondary energy sources carries significant drawbacks,such as high cost and substrate instability, due to uncoupled,enzymatic degradation (Kim and Swartz, 2000a). The deg-radation of the secondary energy sources limits the perfor-mance of cell-free protein synthesis systems. In addition,the resulting phosphate accumulation inhibits long-termprotein synthesis. Thus, improved productivity is expectedif new approaches are taken to minimize these limitations.

In this report, we describe the successful use of pyruvateas a secondary energy source in a cell-free protein synthesissystem prepared fromE. coli cell extract. Our previouswork (Kim and Swartz, 1999) describes the use of pyruvatecoupled with the enzyme pyruvate oxidase. However, a fur-ther advantage could be gained if the energy for proteinsynthesis could be provided even in the absence of anyexogenous enzymes and without the need to supply oxygen.In the work described here we found that the addition ofNAD to the E. coli cell-free protein synthesis system en-ables the regeneration of ATP from pyruvate, taking advan-tage of endogenous enzymes present in the cell extract. Theaddition of coenzyme A (CoA) improved protein synthesis.These reactions were also successfully combined with theconventional phosphoenol pyruvate (PEP) system for addi-tional regeneration of ATP utilizing the pyruvate generatedfrom PEP.

More strikingly, it was found that glucose-6-phosphate(G-6-P) could serve as a secondary energy source to pro-duce a higher synthesis yield than with PEP or pyruvate.Because G-6-P and pyruvate are the first and final interme-diates of the glycolytic pathway, respectively, these resultsstrongly support the possibility of using any glycolytic in-termediate to supply the cell-free system with ATP.

MATERIALS AND METHODS

Phosphoenol pyruvate (potassium PEP) and theE. coli totaltRNA mixture were purchased from Roche Molecular Bio-chemicals (Indianapolis, IN).L-[U-14C]-leucine was fromAmersham Pharmacia Biotechnology (Uppsala, Sweden).All other reagents were obtained from Sigma (St. Louis,MO). T7 RNA polymerase was prepared fromE. coli strainBL21 (pAR1219) according to the procedures of Davanlooet al. (1984). Plasmid pK7CAT, containing the bacterialchloramphenicol acetyltransferase (CAT) sequence betweenthe T7 promoter and T7 terminator (Kigawa et al., 1995),was used as a template for protein synthesis. The plasmidwas purified using the Maxi kit from Qiagen (Valencia,CA). S30 extract was prepared fromE. coli K12 (strainA19) as described earlier (Kim et al., 1996; Kim andSwartz, 1999). Prior to the run-off reaction, 20 U of pyru-vate kinase was added per 10 mL of cell extract. The stan-

dard reaction mixture consists of the following components:57 mM HEPES-KOH (pH 7.5); 1.2 mM ATP; 0.85 mM eachof GTP, UTP, and CTP; 1 mM DTT; 0.64 mM cAMP; 200mM potassium glutamate; 80 mM ammonium acetate; 12mM magnesium acetate; 34mg/mL folinic acid; 6.7mg/mLplasmid; 33mg/mL T7 RNA polymerase; 500mM each of20 unlabeled amino acids; 11mM [14C]-leucine, 2% poly-ethylene glycol (PEG) 8000; 32 mM PEP; and 0.24 vol ofS30 extract. In reactions where pyruvate or G-6-P was usedas the ATP regenerating compound, 33 mM of the energysource was added along with 0.33 mM NAD and 0.26 mMCoA. In certain reactions, 2.7 mM sodium oxalate was usedto improve performance. Reactions were run for given timeperiods in 15- to 120-mL reaction volumes at 37°C. Theamount of synthesized protein was estimated from the mea-sured TCA-insoluble radioactivities as described earlier(Kim et al., 1996; Kim and Swartz, 1999) using a liquidscintillation counter (Beckman LS3801). To determine theamount of soluble product, samples were centrifuged at12,000g for 10 min and TCA-precipitable radioactivities inthe supernatants were measured. To estimate the molecularweight of synthesized protein, samples were loaded onto a16% sodium dodecylsulfate–polyacrylamide gel electropho-resis (SDS-PAGE) gel (Invitrogen) with standard molecularweight markers (see Blue, Invitrogen). Resulting gels werestained with Coomassie Brilliant Blue following the stan-dard procedures (Laemmli, 1970). The protein concentra-tions of cell extracts were measured following the proce-dures of Bradford (1976) using a commercial assay reagent(Pierce, Rockford, IL). To measure ATP concentration, di-luted samples were added to an opaque microtiter platecontaining luciferase solution (0.1mg/mL luciferase and125 mM luciferin), and the intensity of luminescence wasmeasured in a plate luminometer (ML 3000, DynatechLaboratories, Chantilly, VA). ATP concentrations in thesamples were determined by comparison to a calibrationcurve obtained with ATP standards. The enzymatic activityof synthesized chloramphenicol acetyltransferase (CAT)was determined by the spectrophotometric procedures de-scribed by Shaw (1975). After diluting a sample 40-foldwith water, 10mL of the diluted sample was added to acuvette containing 1 mL of prewarmed assay mixture (100mM Tris-Cl, pH 7.8; 0.1 mM acetyl-CoA; 0.4 mg/mL 5,58-dithiobis-2-nitrobenzoic acid [DTNB]; 0.1 mM chloram-phenicol) and the rate of increase in absorption at 412 nmwas measured. The change in absorbance units per minutewas divided by 13.6 to give the result in units (1 U of CATacetylates 1mmol of chloramphenicol per minute).

RESULTS

Use of Pyruvate as a Secondary Energy Sourcefor Cell-Free Protein Synthesis

Previously, we reported that pyruvate could serve as a sec-ondary energy source to regenerate ATP during a cell-free

310 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 74, NO. 4, AUGUST 20, 2001

protein synthesis reaction (Kim and Swartz, 1999). In thatsystem the enzyme, pyruvate oxidase, converts pyruvateinto acetylphosphate, which is then used to regenerate ATPwith endogenous acetate kinase. This pyruvate oxidase–dependent system substantially reduces the cost for energysource, leads to a stable maintenance of ATP during proteinsynthesis, and avoids phosphate accumulation.

The conversion of pyruvate into acetylphosphate, how-ever, requires an exogenous enzyme, pyruvate oxidase (E.C.1.2.3.3), fromLactobacillusor Pediococcussp. because theE. coli enzyme (E.C. 1.2.2) cannot catalyze the formation ofacetylphosphate (it converts pyruvate to acetate instead ofacetylphosphate [Shaw-Goldstein et al., 1978]). Thus, in theabsence of exogenous pyruvate oxidase, ATP is not regen-erated and the level of protein synthesis is negligible. Theuse of commercial pyruvate oxidase significantly offsets theeconomic benefits of the new system. Our attempts to ex-press active pyruvate oxidase (E.C. 1.2.3.3) inE. coli havenot been successful so far. More importantly, because thisenzyme requires molecular oxygen for the oxidation of py-ruvate into acetylphosphate, the synthesis reaction cannotbe easily scaled-up in a simple batch configuration due tothe limitation of oxygen transfer. For example, when wesimply increased the reaction volume to 120mL in the samereaction tubes, the final volumetric yield of synthesis de-creased by more than 60% compared to results with thestandard reaction volume (15mL; Kim and Swartz, unpub-lished data). To avoid these limitations, we sought to elimi-nate the requirements for exogenous enzyme and oxygen.For pyruvate to be used for ATP regeneration in the cell-free system, it first needs to be converted to acetylphos-phate. InE. coli cells, pyruvate is not directly converted toacetylphosphate. Instead, two reactions are necessary. First,pyruvate dehydrogenase catalyzes the condensation of CoAand pyruvate to make acetyl-CoA in the presence of NADas a cofactor. NAD is reduced to NADH during this reac-tion. Subsequently, acetyl-CoA is converted to acetylphos-phate by phosphotransacetylase (PTA). On the other hand,pyruvate-formate lyase can also make acetyl-CoA from py-ruvate producing formate as the byproduct. All of the en-zymes required for these reactions were assumed to be inthe cell extract. We thus tested the addition of the cofactors,NAD and CoA, to stimulate the regeneration of ATP frompyruvate in support of cell-free protein synthesis.

Sodium pyruvate (33 mM) was added to reaction mix-tures with or without the cofactors. After a 2-h incubation,significant synthesis was observed (Fig. 1) in the presenceof NAD, with about 20% additional synthesis obtained byalso adding CoA. The final yield of protein synthesis withboth cofactors was about 70% of that from the reaction inwhich PEP was used. The use of pyruvate without the co-factors resulted in low protein synthesis equivalent to thecontrol reaction without any secondary energy source. Thisresult indicates that the conversion of pyruvate to acetyl-CoA is accomplished by pyruvate dehydrogenase ratherthan pyruvate-formate lyase, as the latter enzyme does notrequire cofactors. Because only a catalytic amount of NAD

(0.33 mM) is required, it seems obvious that NAD is alsoregenerated during the synthesis reaction. One possibility isthat the reduced NAD is reoxidized by the conversion ofpyruvate into lactate (Fig. 2). High-performance liquidchromatography (HPLC) analysis confirmed acetate andlactate formation (data not shown). A catalytic amount ofCoA (0.27 mM) was beneficial for protein synthesis, sug-gesting that sufficient CoA is either not present in the initialcell extract or is lost during dialysis.

Utilization of Pyruvate Generated From PEP

Consumption of PEP in the conventional ATP regenerationsystem produces pyruvate as a byproduct. Based on theaforementioned results of pyruvate utilization for ATP re-generation, we tested the addition of NAD and CoA to theconventional system. We reasoned that ATP supply couldbe enhanced through the secondary utilization of the pyru-vate generated from PEP. Figure 3A shows that the presenceof those cofactors does improve ATP supply. The decreasein ATP concentration was substantially slowed by the ad-dition of NAD and CoA.

Recently, we reported that oxalate, a potent inhibitor ofPEP synthetase, enhances ATP concentration in the synthe-sis reactions with pyruvate or PEP (Kim and Swartz,2000b). The effect of oxalate was still observed in the pres-ence of NAD and CoA. The yield of CAT synthesis in-creased upon the addition of 2.7 mM oxalate (Fig. 3B). Inaddition, increasing the initial concentrations of amino acidsfurther stimulated protein synthesis. When the ATP supplywas increased by the addition of the cofactors and oxalate,both rate and duration of protein synthesis were signifi-

Figure 1. Expression of CAT using pyruvate as the secondary energysource. Sodium pyruvate (33 mM), 0.33 mM NAD, and 0.27 mM CoAwere added in the indicated combinations to 15-mL reaction mixtures toregenerate ATP during the synthesis reaction. Reactions were carried outfor 2 h and TCA-insoluble radioactivities were measured. In the controlreaction, 33 mM PEP was used instead of pyruvate and cofactors.

KIM AND SWARTZ: REGENERATION OF ATP FROM GLYCOLYTIC INTERMEDIATES 311

cantly improved by also increasing the initial concentrationsof amino acids from 0.5 to 2 mM (Fig. 3B). Higher con-centrations were not as effective (data not shown). As aresult, the final yield of CAT synthesis after a 1-h incuba-tion averaged 350mg/mL. Furthermore, periodic additionsof fresh amino acids, PEP and magnesium acetate, accord-ing to published procedures (Kim and Swartz, 2000a), al-lowed the synthesis reaction to continue for over 3 h, re-sulting in a final yield of 750mg/mL (Fig. 4). SynthesizedCAT gave a single intense band on a SDS-PAGE gel afterCoomassie Blue staining. Approximately 60% of CAT ex-pressed was soluble and the measured specific activity wasgreater than to the published value of 125 U/mg (Shaw,1977) (Table I).

G-6-P as an Alternative Secondary Energy Source

Encouraged by the observation that pyruvate could be usedfor ATP regeneration, we further investigated if earlier gly-colytic intermediates could be used to regenerate ATP inour cell-free system. The use of G-6-P was examined as itis the first intermediate of the glycolytic pathway. G-6-P hasbeen examined with reticulocyte lysates by several groups(Jackson and Hunt, 1978), showing limited success, but wefound no report of its evaluation in prokaryotic systems.When 33 mM G-6-P was used under the same reactionconditions as in the pyruvate/NAD system, it supported pro-tein synthesis. In addition, although the initial rate waslower than with PEP, protein synthesis continued for over 2h (Fig. 5A). As a result, approximately 30% more CAT wasproduced at the end of incubation. Most likely, this is due tothe remarkably extended maintenance of ATP concentra-tions. The time course of ATP concentration during proteinsynthesis with G-6-P was characterized by a period of rela-tively stable maintenance of ATP level followed by a slowdecrease over the incubation period (Fig. 5B). Unlike thereactions using PEP or pyruvate, the addition of sodiumoxalate did not further improve the maintenance of ATPconcentration. On the contrary, it significantly reducedoverall ATP concentrations (data not shown). Perhaps oneor more of the early glycolytic enzymes are also inhibitedby this compound.

The ATP regeneration with G-6-P indicates that all of theglycolytic enzymes required to convert G-6-P into pyruvate

are active under the present reaction conditions. This pro-vides great flexibility when choosing a secondary energysource for protein synthesis. In theory, any of the glycolyticintermediates between G-6-P and pyruvate can be used forATP regeneration.

The specific activity of synthesized CAT was greater thanthe published data with all of the secondary energy sourcesdescribed here: PEP, pyruvate, PEP/pyruvate, and G-6-P(125 U/mg; Shaw, 1975) (Table I).

DISCUSSION

We have shown that regeneration of ATP during cell-freeprotein synthesis can be accomplished by using alternativeenergy sources, such as pyruvate and G-6-P, in the absenceof exogenous enzymes. With pyruvate as the secondary en-ergy source, a synthesis yield of approximately 70% wasobtained as compared with the reaction using PEP, the con-ventional energy source. This is not surprising because aportion of the pyruvate may be needed to recycle NADH toNAD. However, because of the expense of PEP, the use ofpyruvate significantly improves the economic efficiency ofprotein synthesis. The cost of the G-6-P and NAD is about3% the cost of the K-PEP, and the cost of the pyruvate andNAD is approximately 0.4%. It should be noted, however,that the addition of CoA is significantly more expensive(14% relative to K-PEP) and provides only a 20% increasein protein yield from pyruvate. The CoA addition might beavoided altogether by changing the cell extract growth me-dium from a complex to a defined glucose medium to pos-sibly increase the CoA content of the cell extract or bymodifying the dialysis step during cell extract preparation todecrease CoA loss. The CoA is also unlikely to be neededwhen G-6-P is used because most of the ATP will be gen-erated from the glycolytic reactions and not fromacetylphosphate.

In addition to the postulated mechanisms just indicated,oxidative phosphorylation of ADP is another possible path-way for ATP regeneration in the new system. However,addition of oligomycin, an inhibitor of F1F0ATPase, hasshown no effect on ATP regeneration or protein synthesis.Furthermore, a reaction under anaerobic conditions gave asimilar yield of protein synthesis (data not shown). There-fore, it is very likely that the ATP is regenerated through theaforementioned pathways.

Figure 2. Proposed mechanism of ATP regeneration with pyruvate.

312 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 74, NO. 4, AUGUST 20, 2001

The reactions of ATP regeneration using pyruvate alsocan be used to improve the utilization efficiency of theconventional energy source, PEP. After being used for ATPregeneration or being degraded by phosphatase activities,PEP produces an equimolar amount of pyruvate. If the co-factors NAD and CoA are present in the reaction mixture, atleast half of the newly generated pyruvate is available forATP regeneration (the other half is probably used for theregeneration of NAD). As a result, through the two-stageutilization of the energy source, the overall concentration of

ATP is elevated and prolonged, and the productivity ofprotein synthesis is improved. Further improvements mightbe gained if the nonproductive hydrolysis of PEP could beavoided, but at least the use of the resulting pyruvate allowssome benefit to be salvaged.

In accordance with the previous report, addition of ametabolic inhibitor of PEP synthetase (sodium oxalate) fur-ther enhanced protein synthesis, almost certainly by increas-ing the ATP supply. In addition, with the enhanced ATPlevel, the yield of protein synthesis increased further when

Figure 3. Time course of protein synthesis and ATP concentration. Standard reaction mixtures (120mL) with 33 mM PEP were prepared and incubatedin the presence of 0.33 mM NAD, 0.27 mM CoA, and 2.7 mM sodium oxalate. To measure ATP concentrations (A), 10-mL samples were withdrawn, mixedwith the same volume of 10% TCA solution, and centrifuged for 10 min. Ten microliters of supernatant was used for ATP analysis as described in the“Materials and Methods.” At the given timepoints, 5-mL samples were taken and TCA-insoluble radioactivities were counted to measure protein synthesis(B). (s) Control reaction; (d) with NAD and CoA; (X) with NAD, CoA, and oxalate; (j) with NAD, CoA, oxalate, and 2 mM amino acids. At the endof each reaction, 5-mL samples were taken for a 16% SDS-PAGE gel. The gel was stained with Coomassie Blue following the standard procedures (insetof B). Lanes: M, standard molecular weight markers; C, control reaction without the template plasmid; 1, standard reaction; 2, reaction with 0.33 mM NADand 0.27 mM CoA; 3, reaction with 0.33 mM NAD, 0.27 mM CoA, and 2.7 mM sodium oxalate; 4, reaction with 0.33 mM NAD, 0.27 mM CoA, 2.7 mMsodium oxalate, and 2 mM amino acids.

Figure 4. Supplementation of PEP, amino acids, and magnesium during protein synthesis. A synthesis reaction was carried out in the presence of 2 mMamino acids, 33 mM PEP, 0.33 mM NAD, 0.27 mM CoA, and 2.7 mM sodium oxalate in a 120-mL volume. During the incubation period, the initialconcentrations of PEP, amino acids, and magnesium acetate were added to the reaction every hour. Five microliter samples were taken at the giventimepoints to measure the concentration of ATP (A) and the yield of CAT synthesis (B). The same volumes of water were added to the single batch reaction.(s). Single batch reaction; (d) reaction with the additions. Samples taken at the end of each reaction were analyzed by SDS-PAGE followed by CoomassieBlue staining (inset of B). M, standard molecular weight markers; C, control reaction without template plasmid; B, single batch reaction; FB, reaction withthe additions of PEP, amino acids, and magnesium acetate. Arrow indicates the expressed CAT.

KIM AND SWARTZ: REGENERATION OF ATP FROM GLYCOLYTIC INTERMEDIATES 313

the initial concentrations of amino acids were increased.Finally, fed-batch additions of amino acids and PEP al-lowed the synthesis reaction to continue for over 3 h toproduce a final yield of 750mg/mL. Because this fed-batchprocedure requires only two, hourly additions, it is ame-nable to multiplexed synthesis formats for high-throughputscreening. Further cost reduction and productivity improve-ment might be gained by adding only those amino acids thatare degraded (most likely arginine, cysteine, and tryptophan[Kim and Swartz, 2000a]).

We also showed that G-6-P, the first intermediate of theglycolytic pathway, can be used as the secondary energysource, resulting in a yield higher than with the PEP system.As compared with PEP, the use of G-6-P substantially in-creased the synthesis yield, mainly by prolonging the reac-tion period. This seems to have been due to the enhancedsupply of ATP, which can be explained by the fact thatG-6-P offers a greater potential to regenerate ATP com-pared with PEP or pyruvate. Whereas pyruvate can regen-erate, at best, only the equivalent number of ATP molecules,and PEP only slightly more, three molecules of ATP can be

generated during the oxidation of G-6-P into pyruvate. (Thetwo molecules of pyruvate generated from the glycolyticpathway are probably required to regenerate the two NADmolecules that are reduced by glyceraldehyde-3-phosphatedehydrogenase.) Although the reason for the relatively slowinitial rate remains unanswered, these results strongly implythat we can use any of the glycolytic intermediates as asecondary energy source to support cell-free protein synthe-sis.

In all of the ATP regeneration systems examined, ap-proximately 60% of synthesized CAT was found in thesoluble fraction (see Table I). Surprisingly, this appeared tobe relatively independent of protein yield. Interestingly, thespecific activity of synthesized CAT based on the measuredamount of soluble product was always significantly higherthan the published data (125 U/mg; Shaw, 1975). Possibly,the present procedures underestimate the yield of proteinsynthesis, or perhaps the published specific activity is toolow.

Our eventual goal is to use glucose as the secondaryenergy source. The use of glucose would provide a cell-freesystem that is highly competitive with traditional technolo-gies for protein expression in terms of economic efficiency.Our initial results suggest such a possibility. Even thoughthe yield of synthesis was relatively low (approximately 80mg/mL; Kim and Swartz, unpublished data), the use of glu-cose along with hexokinase did support protein synthesis.However, the results with glucose were inconsistent as wellas low, and we are currently working to improve this at-tractive system. Furthermore, even though ATP is not re-generated by oxidative phosphorylation in our system, wedo not exclude the possibility that the present cell extractcontains active membrane vesicles. If we can use thesevesicles, oxidative phosphorylation would offer an ex-tremely efficient method for ATP supply by mimicking thefunction of living cells.

Figure 5. Expression of CAT using glucose-6-phosphate as the secondary energy source. Glucose-6-phosphate (33 mM), 0.33 mM NAD, and 0.27 mMCoA were added to a 120-mL synthesis reaction. Five- and 10-mL samples were taken to determine protein synthesis and ATP concentration, respectively,and were assayed as in Figure 3. (s) Conventional reaction using PEP; (d) reaction using glucose-6-phosphate as energy source. (A) Time course of CATsynthesis. (B) Time course of ATP concentration.

Table I. CAT yields and specific activities after 3-hr incubations withdifferent ATP regeneration systems.

Secondaryenergy sources

Totalyield ofsynthesis(mg/mL)

Yield ofsolubleproduct(mg/mL) Soluble

Specificactivity(Us/mgsolubleproduct)

PEP 168 ± 12.7 107 ± 10.4 61% 181 ± 21.5Pyruvate 134 ± 8.6 75 ± 15.4 56% 153 ± 15.7PEPa + pyruvate 317 ± 53.0 189 ± 42.7 60% 167 ± 14.9G-6-P 228 ± 12.9 143 ± 4.5 63% 158 ± 26.1

aTo utilize the pyruvate generated from PEP, NAD, CoA, and oxalatewere added to the reaction mixture as described in “Materials and Meth-ods.”

314 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 74, NO. 4, AUGUST 20, 2001

Because each mRNA is used multiple times, we estimatethat the translational demand for ATP dominates. Duringprotein synthesis, the EF-Tu cycle consumes two GTPs(Ehrenberg et al., 1990) and EF-G requires another mol-ecule of GTP for the translocation of ribosome. In addition,because a molecule of ATP is hydrolyzed to AMP duringaminoacylation of tRNAs, we assumed that five moleculesof ATP would be required to add an amino acid residue.Based on these assumptions, we calculated the efficienciesof ATP utilization (Table II). In the standard reaction usingPEP, the final amount of synthesized CAT in a 15-mL re-action was 99.3 pmol, which represents 2.2 × 10−8 mol ofpeptide bonds. Thus, we could assume that 0.11mmol ofATP was used for protein synthesis. Because the totalamount of ATP that could be generated in the same reactionmixture was 0.5mmol, the efficiency of ATP utilization wasestimated to be 22%. On the other hand, we could produce66 pmol of CAT in the reaction using pyruvate, whichequals 1.50 × 10−8 mol of peptide bonds. Because we esti-mated that only half of the pyruvate could be used for ATPregeneration, 0.25mmol of ATP was generated and theutilization efficiency was 30%.

By adding NAD, CoA, and oxalate to the conventionalPEP system, we synthesized 182 pmol of CAT in a 15-mLreaction. Because the estimated total amount of availableATP was 0.75mmol (0.5 mmol from PEP and 0.25mmolfrom the pyruvate produced from PEP), the efficiency ofATP utilization was estimated to be 27%. However, whenthe initial concentrations of amino acids were raised to 2mM, the amount of produced CAT increased to 313 pmoland the efficiency of ATP utilization reached 47%.

In contrast, the efficiency of ATP utilization with G-6-Pwas only 8.8% (note that one molecule of G-6-P can gen-erate three molecules of ATP). Thus, only a small fractionof the potential ATP pool is used for protein synthesis. Thissuggests either that the majority of the regenerated ATP isdegraded by ATPase activities present in the current cellextract or that side reactions are degrading glycolytic inter-mediates. In other words, we can expect to improve proteinsynthesis from G-6-P by identifying and removing suchactivities. For instance, we hope to identify and disrupt the

genes encoding enzymes that catalyze nonproductive deg-radation of ATP.

In addition, as we reported previously (Kim and Swartz,1999), the presence of nonspecific phosphatase activities inthe cell extract may also be degrading the PEP and G-6-P.Even though repeated addition of the secondary energysource maintains the ATP supply, accumulated inorganicphosphate eventually inhibits protein synthesis. This shouldbe less severe with G-6-P than with PEP, but is still aconcern. In both cases, the identification and removal of thephosphatase activities should improve system performance.These activities can be removed by inactivating the corre-sponding chromosomal gene or, if the enzymatic activity isessential for growth, we can genetically mark the enzymewith an affinity tag for removal during cell extract prepa-ration. The same approach is being taken to avoid the prob-lem of amino acid degradation (Kim and Swartz, 2000a).Through such a “genetic optimization” of theE. coli strain,combined with improved ATP regeneration systems, weexpect to develop a highly efficient batch cell-free proteinsynthesis system.

Finally, although this work has focused on the regenera-tion of ATP from ADP, it should be remembered that manyreactions generate AMP and that the regeneration of all thenucleotide triphosphates is required for the coupled, tran-scription/translation system, especially GTP. As we con-tinue to improve the ATP supply, we may encounter otherlimitations in phosphate transfer reactions (e.g., the conver-sion of AMP to ADP by adenylate kinase) that hampersystem performance. Clearly, there are many questions stillto address and many opportunities to improve the system.For example, this work suggests that many, if not most,metabolic pathways can be harnessed to assist cell-free pro-tein synthesis. It is this tremendous potential for metabolicmanipulation that makes the cell-free system so promising.

The authors thank Jim Zawada for developing the ATP assaymethod.

References

Alimov AP, Khmelnitsky AY, Simonenko PN, Spirin AS, Chetverin AB.2000. Cell-free synthesis and affinity isolation of proteins on a nano-mole scale. BioTechniques 28:338–344.

Anderson CW, Straus JW, Dudock BS. 1983. Preparation of a cell-freeprotein-synthesizing system from wheat germ. Meth Enzymol 101:635–644.

Bradford MM. 1976. A rapid and sensitive method for the quantitation ofprotein utilizing the principle of protein-dye binding. Anal Biochem72:248–254.

Davanloo P, Rosenberg AH, Dunn JJ, Studier FW. 1984. Cloning andexpression of the gene for bacteriophage T7 RNA polymerase. ProcNatl Acad Sci USA 81:2035–2039.

Ehrenberg M, Rojas AM, Weiser J, Kurland CG. 1990. How many EF-Tumolecules participate in aminoacyl-tRNA binding and peptide bondformation inEscherichia colitranslation? J Mol Biol 211:739–749.

Jackson RJ, Hunt T. 1978. The use of hexose phosphate to support proteinsynthesis and generate [r-32P]ATP in reticulocyte lysate. FEBS Lett93:235–238.

Jermutus L, Ryabova LA, Pluckthun A. 1998. Recent advances in produc-

Table II. Efficiency of ATP utilization in different ATP regenerationsystems.

MaximumATP

available(mmol/15 mL)

Amount ofsynthesized

CAT(pmol/15mL))

Efficiencyof

ATPutilization

PEP 0.50 99.3 ± 7.1 22.6%Pyruvate 0.25 65.9 ± 17.5 29.9%PEP + pyruvatea 0.75 181.5 ± 60.7 27.5%PEP + pyruvateb 0.75 312.9 ± 44.0 47.4%G-6-P 1.50 116.0 ± 10.1 8.8%

aTo utilize the pyruvate generated from PEP, NAD, CoA, and oxalatewere added to the reaction mixture as described in “Materials and Meth-ods.”

bInitial concentrations of amino acids were 2 mM.

KIM AND SWARTZ: REGENERATION OF ATP FROM GLYCOLYTIC INTERMEDIATES 315

ing and selecting functional proteins by using cell-free translation.Curr Opin Biotechnol 9:534–548.

Kigawa T, Muto Y, Yokoyama S. 1995. Cell-free synthesis and acid-selective stable isotope labeling of proteins for NMR analysis. J Bio-mol Nucl Magn Reson 6:129–134.

Kim D-M, Kigawa T, Choi C-Y, Yokoyama S. 1996. A highly efficientcell-free protein synthesis system fromE. coli Eur J Biochem 239:881–886.

Kim D-M, Swartz JR. 1999. Prolonging cell-free protein synthesis with anovel ATP regeneration system. Biotechnol Bioeng 66:180–188.

Kim D-M and Swartz JR. 2000a. Prolonging cell-free protein synthesis byselective reagent additions. Biotechnol Progr 16:385–390.

Kim D-M, Swartz JR. 2000b. Oxalate improves protein synthesis by en-hancing ATP supply in a cell-free system derived fromEscherichiacoli. Biotechnol Lett 22:1537–1542.

Laemmli UK. 1970. Cleavage of structural proteins during the assembly ofthe bacteriophage T4. Nature 227:680–685.

Pratt JM. 1984. Coupled transcription–translation in prokaryotic cell-freesystems. In: Hames BD, Higgins SJ, editors. Transcription and trans-lation: A practical approach. New York: IRL Press. p 179–209.

Rattan SI, Kristensen O. 1990. Continuous gene expressionin vitro: TheSpirin system. Trends Biotechnol 8:275–276.

Ryabova LA, Vinokurov LM, Shekhovtsova EA, Alakhov YB, Spirin AS.1995. Acetyl phosphate as an energy source for bacterial cell-freetranslation systems. Anal Biochem 226:184–186.

Shaw WV. 1975. Chloramphenicol acetyltransferase from chlorampheni-col-resistant bacteria. Meth Enzymol 43:737–755.

Shaw-Goldstein LA, Gennis RB, Walsh C. 1978. Identification, localiza-tion, and function of the thiamine pyrophosphate and flavin adeninedinucleotide dependent pyruvate oxidase in isolated membranevesicles ofEscherichia coliB. Biochemistry 17:5605–5613.

Stiege W, Erdmann VA. 1995. The potentials of thein vitro protein bio-synthesis system. J Biotechnol 41:81–90.

316 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 74, NO. 4, AUGUST 20, 2001