Protein Engineering vol.9 no.12 pp.1197-1202, 1996

Relationship between thermal stability, degradation rate andexpression yield of barnase variants in the periplasm ofEscherichia coli

Won Shik Kwon, Nancy A.Da Silva2 andJames T.KeUis Jr1*2

Department of Chemical and Biochemical Engineering, University ofCalifornia, Irvine, CA 92717 and 'Genencor International Inc.,925 Page Mill Road, Palo Alto, CA 94J5O4, USA

2To whom correspondence should be addressed

An advantage of exporting a recombinant protein to theperiplasm of Escherichia coli is decreased proteolysis in theperiplasm compared with that in the cytoplasm. However,protein degradation in the periplasm also occurs. It hasbeen widely accepted that the thermodynamic stability ofa protein is an important factor for protein degradation inthe cytoplasm of E.coli. To investigate the effect of thethermodynamic stability of an exported protein on theextent of proteolysis in the periplasm, barnase (an extracel-lular ribonuclease from Bacillus amyloliquefaciens) fusedto alkaline phosphatase leader peptide was used as a modelprotein. A set of singly or doubly mutated barnase variantswere constructed for export to the E.coli periplasm. It wasfound that the half-life of the barnase variants in vivoincreased with their thermodynamic stability in vitro. Adominant factor for the final yield of exported barnase wasnot exportability but the turnover rate of the barnasevariant. The yield of a stabilized mutant was up to 50%higher than that of the wild type. This suggests thatexporting a protein to the periplasm and using proteinengineering to enhance the stability can be combined asa strategy to optimize the production of recombinantproteins.Keywords: barnase/periplasmic degradation/thermal stability

Introduction

To maximize the production of recombinant proteins in Escher-ichia coli cultures, various strategies have been attempted.High copy number plasmids or strong promoters are commonlyemployed. However, these efforts to enhance expression canbe significantly diminished by proteolysis. Proteolysis canbe so severe that the low yield of recombinant protein ismisinterpreted as a lack of expression. Proteolytic degradationof recombinant proteins has been observed in both rapidlygrowing and non-growing cultures (Pine, 1970, 1973). There-fore, avoiding undesirable proteolysis, and consequentlyincreasing the final yield of recombinant protein production,has long been a major goal of many researchers.

Because only 15-20% of known E.coli proteases reside inthe periplasm (Maurizi, 1992), artificially targeting a proteininto the periplasm has been effective at reducing the extent ofdegradation by proteases. Indeed, Talmadge and Gilbert (1982)showed that die half-life of insulin was increased 10-fold byexporting the protein into the periplasm of E.coli. In addition,export of a protein offers many advantages over cytoplasmicproduction (Baneyx et al., 1991). Ease of purification, disulfide

bond formation and the reduction of possible toxic effects ofrecombinant proteins are among these advantages.

However, some exported proteins have been reported to bedegraded rapidly even in the periplasm of E.coli (Strauchet al., 1989; Baneyx and Georgiou, 1990). In some cases, theexport process of foreign proteins was found to be coupledwith proteolysis by either membrane-associated or periplasmicproteases, resulting in a low yield of the exported foreignprotein (Gentz et al., 1989). Although the construction ofperiplasmic protease-deficient strains can substantially reducethe level of degradation (Meerman and Georgiou, 1994), thesestrains cannot be used in cases where the responsible protease(s)are unknown. Furthermore, the slower growth rate of suchstrains can pose a problem. The addition of protease-inhibitingmetal ions (e.g. zinc ion; Baneyx et al., 1991) or chemicals(e.g. phenylmethylsulfonyl fluoride; Anba et al., 1988) to theculture medium is an alternate way of reducing proteolysis inthe periplasm, because the outer membrane of E.coli, unlikethe inner membrane, is quite permeable to molecules withmolecular weights <600 Da (Decad and Nikaido, 1976).However, these additional components were found to affectthe cell's viability and the synthesis capacity of recombinantproteins. Moreover, these additions invariably create additionalexpense and concerns for downstream processing. Loweringthe cultivation temperature is another effective method todecrease the extent of proteolysis (Chesshyre and Hipkiss,1989; Battistoni et al, 1992). Again, the slower growth rateand suboptimum synthesis rate of recombinant protein can bea problem.

If the desired protein could be engineered with improvedintrinsic resistance against proteolysis, it might be possible toenhance the final yield of the foreign protein without modifyingthe host strain, culture medium or cultivation conditions. Anintriguing problem in proteolysis has been how cells distinguishproteins to be degraded (abnormal and non-essential proteins)from proteins to be protected (normal and essential proteins).Several properties of a protein have been shown to be respons-ible for its degradation rate. Among these, the thermodynamicstability of the folded structure was found to be a keydeterminant of the half-lives of proteins in the cytoplasmof E.coli (McLendon and Radany, 1978). Examples of thismechanism have been confirmed for the N-terminal domainof X-repressor protein (Parsell and Sauer, 1989) and T4lysozyme (Inoue and Rechsteiner, 1994); the half-lives oftemperature-sensitive variants of these proteins have beenfound to correlate with their thermal stabilities. This indicatesthat proteolysis occurs primarily with the unfolded form ofa protein and is controlled by the thermodynamic foldingequilibrium. If this mechanism also holds true in the periplasm,it should be possible to increase foreign protein productionby engineering the protein to have higher diermal stabilitycompared with wild-type protein, and exporting it to theperiplasm of the cell.

Here we have used barnase, an extracellular ribonuclease

© Oxford University Press 1197

W^.Kwon, N.A.Da SUva and J.T.KeUis Jr

EcoRl

So/11

Fig. 1. Schematic drawing of the plasmid pKK-EC2. Important uniquerestriction sites are underlined, and two basal mutations in the 'pseudo wild-type barnase' are in bold.

from Bacillus amyloliquefaciens, fused to the signal peptideof alkaline phosphatase as a model protein. To avoid thecomplication of co-expression of its intracellular inhibitorbarstar, barnase was inactivated by substituting a catalyticallyessential histidine at position 102 with alanine. Single ordouble mutations were introduced into the barnase gene toconstruct variants with decreased or increased thermodynamicstability. The relationship between thermal stability anddegradation of these proteins in the periplasmic space ofE.coli was studied. In addition, the potential of applying thisrelationship to increase the overall yield of recombinant proteinwas also considered.

Materials and methodsConstruction of the plasmid pKK-EC2Plasmid pTZ416-2 (kindly provided by Prof. Alan Fersht)contains the barnase gene fused to the PhoA (alkaline phosphat-ase) signal peptide, the barstar gene and a V36M mutationwithin the bamase gene (Sancho and Fersht, 1992). UsingPCR (Clackson et al., 1991), we amplified the PhoA-barnasefusion gene and introduced another basal mutation, HI02A(histidine at position 102 replaced by alanine), while creatingconvenient restriction sites (EcoRl and HindUl) that flank thefusion gene. The two oligonucleotides for PCR were as follows:N-term, 5'-gtacatggagagaattcaatgaaacaaagc-3', and C-term, 5'-ccgtttttaagcttatctgatttttgtaaaggtctgataagcgtccgt-3'. The ampli-fied fragment was inserted into the multiple cloning site of theexpression vector, pKK223-3 (Pharmacia), by digestion withEcoKI and HindUl, followed by ligation, resulting in pKK-EC1. An Afllll site in the vector DNA was mutated to a flg/IIsite; the remaining AflUl site at the N-terminus of the barnasegene becomes unique (the signal peptide is flanked by uniqueEcoRl and AflUl sites). The final plasmid was named pKK-EC2, and its structure is shown in Figure 1. The entire fusiongene was sequenced to verify the correct construction. AnE.coli strain JM105 [F' traD36 lad* A(lacZ) M15proA+B+/thirpsL (StrO endA sbcB15 sbcC? hsdR4(rK-mK

+)A(lac-proAB);Pharmacia] was transformed with pKK-EC2 and used forbarnase purification as well as the in vivo studies.Site-directed mutagenesis

We used Transformer™ (Clontech) for high-efficiency site-specific mutagenesis of double-stranded plasmid DNA. E.coli

1198

strain BMH 71-18 mutS {hi, supE, A(lac-proAB),[mutS::TnlO][F'proAB, lacY* ZAM15]} was used for mutagen-esis. For the selection oligonucleotide, either 5'-TAA-GCCCGGTACCTCTACGC-3' or 5'-TAAGCCCGGATCCT-CTACGC-3' was used to change a unique BamHl site to aKpnl site, or vice versa, respectively. Mutagenic oligonucleo-tides were as follows: N58A, 5'-GCCTTCCCTAGCTGAGAA-GATGT-3'; H18Q, 5'-CAGGTAGCTTCTGATATGTCTG-3';Q15I, 5'-CTTATGATATGTAATAAGATAATCCG-3'; andK108R, 5'-GCTTATCTGATACGTGTAAAGGTCT-3'. Eachmutation was verified by dideoxy sequencing (Sanger et al.,1977) using [a-35S]dATP (Amersham) and Sequenase™(USB).

Barnase purification

Cells were grown overnight in 10 ml LB with 100 (ig/mlampicillin at 37°C. 1 1 NZCYM medium (Sambrook et al.,1989) was inoculated with this overnight culture and cultivateduntil mid-log phase. 1 mM IPTG was added to induce theexpression of barnase, and the culture was harvested after 3 hof induction. The purification method was identical to thatdescribed in Kellis et al. (1989), except that Fast Flow S-Sepharose (Sigma) was used instead of SP-Trisacryl. The finalproduct purified by this method was homogeneous, as judgedby SDS-PAGE.

Flask culture and immunoblot analysis

Cells were grown in 50 ml LB medium with 100 (ig/mlampicillin at 37°C and 250 r.p.m. When the cell densityreached O D ^ = 0.8, barnase expression was induced by theaddition of IPTG to a final concentration of 1 mM. Throughoutthe culture and induction periods, specific growth rates werecarefully monitored to ensure that every culture had the sameexperimental conditions; the difference in specific growth ratesup to 4 h post-induction was <5%. After 2 h of induction,1 ml of cells was harvested by centrifugation and the pelletresuspended in 50 |il TE8 (Tris 10 mM, EDTA 1 mM, pH 8.0)and 50 uJ SDS-PAGE loading buffer. Each sample wasincubated in boiling water for 5 min and then fractionated by15% SDS-PAGE (Laemmli, 1970). A rabbit antiserum againstbarnase (prepared by Bethyl Laboratories) was used to detectbamase in the immunoblot analysis. The relative amounts ofprotein were estimated from densitometric scanning of the blot(Model GS-670 Imaging Densitometer, Bio-Rad).

Pulse—chase experiments

Cells were grown at 37°C in 10 ml M9 minimal medium(Sambrook et al., 1989) supplemented with 0.4% glucose,100 n̂ g/ml ampicillin, 0.05% thiamine and all amino acidsexcept cysteine and methionine. When the cell density reachedODeoo = 0.5, the entire culture volume was transferred toa prewarmed flask which contained IPTG to give a finalconcentration of 1 mM. Barnase production was induced for30 min. The culture was then transferred to a prewarmed flaskcontaining 10 \lC\lm\ [35S]methionine (DuPont NEN). After3 min of labeling, 5 ml prewarmed M9 minimal mediumcontaining an excess amount of unlabeled methionine (3% w/v) were added. At each time point indicated in Figure 3,150 \i\ of cells were drawn from the culture medium, mixedimmediately with the equivalent volume of SDS-PAGE loadingbuffer and boiled for 5 min. Samples were fractionated by15% SDS-PAGE. The dried gel was exposed to a phosphorim-ager high-sensitivity screen for 5-7 h, and band densities were

Protein stability and periplasmic expression

determined with a phosphorimage analyzer (Model GS-250Molecular Imager, Bio-Rad).

Thermal unfolding of barnase variantsThermally induced unfolding, monitored by fluorescence spec-troscopy, was carried out with 10 \iM of each barnase variant in50 mM (Af-morpholino)ethanesulfonic acid (pH 6.3). Jacketedcuvettes were heated by a thermostated circulating water bath,and the actual temperature of the solution was monitoredcarefully by a thermocouple immersed in the cuvette underobservation. The excitation wavelength was 290 nm and theemission wavelength was 315 nm (Aminco-Bowman series 2luminescence spectrometer, SLM Instruments).

Thermal unfolding was analyzed by non-linear regressionusing the data-fitting program GraFit (Erithacus Software).The fluorescence data were fitted to a transformation of theGibbs-Helmholtz equation, where:

Fraction Native = 1/[1 + exp({ACp[(rm -T) + T\n(T/TJ] +A#m(l - T/Tm)}/RT)].

Fraction Native is equal to (F - Fu)/(Fn - FJ , where F is theobserved fluorescence at a given temperature and Fu and Fn

are the fluorescence values of the unfolded and native proteinsat that temperature, obtained by linear extrapolation of thetemperature dependence of the fluorescence outside the regionof the unfolding transition (Pace, 1986). AHm is the enthalpychange for unfolding at the midpoint of the thermal unfoldingtransition, Tm. ASm = AHJTm, because AG is zero at Tm. Avalue of 1.4 kcal/mol.K for the heat capacity change ofunfolding (ACp) was obtained from calorimetric studies ofbarnase unfolding (Griko et al., 1994).

ResultsThe experimental systemPlasmid pKK-EC2 constructed for this study has a strong tacpromoter that can be regulated by the addition of IPTG, andthe signal peptide can be readily replaced because it is flankedby two unique restriction sites (FcoRI and AflUT); signalpeptide manipulations were not carried out in our study (seeMaterials and methods). The base protein engineered for thisstudy contains two mutations. Histidine 102 was mutated toalanine to eliminate the RNase activity of the enzyme toobviate co-expression with barstar (Paddon and Hartley, 1987).Valine 36 was mutated to methionine to provide a site forthe incorporation of [35S]methionine for in vivo pulse-chaseexperiments (wild-type bamase lacks the sulfur-containingamino acids methionine and cysteine). Both the His 102 —»Ala and Val36 —> Met mutations have been characterizedpreviously and have little effect on the stability and level ofexpression of bamase (Serrano and Fersht, 1989; Sancho andFersht, 1992). With this engineered bamase (referred to as'pseudo wild type') as a template, several bamase variantswere constructed by single or double mutations. Therefore,the variants actually have triple or quadruple mutations com-pared with the real wild-type bamase. Specific mutants werechosen to represent a wide range of thermal stabilities, includ-ing destabilized and stabilized mutants, compared with thewild-type protein (Kellis et al., 1989; Serrano and Fersht,1989; Serrano et al., 1993). However, as mentioned above,bamase variants constructed in our laboratory have two basalmutations that differ from the bamase variants studied byother researchers. Thus, thermal stability measurements of thebamase variant were an essential step for this study.

§

1

1.0

0.8

0.6

0.4

0.2

0.0

1 1 1 1

I

-

-_ o'- •7 •

'_ au

-

i i • • i i i • i • i i • • • i i • • • i • ' ' • _

\ \ A °l \N58A V V \ \ T J

H18Q \ \ A fa \ •Pseudo WT V \ \ ^ \ :Q151 \ \ \ N l "Qi5i/Kio8R \ J O J V :

30 35 40 45 50

Temperature (°C)

55 60

Fig. 2. Thermal unfolding of bamase variants. One-letter codes for theamino acids were used. Each mutation is denoted as the original aminoacid, followed by its position and then the substituted amino acid. Observedmelting temperatures (at pH 6.3) are given in Table I.

Table I. Thermodynamic parameters obtained by a Gibbs-Helmholtzanalysis of thermal unfolding of the barnase variants

Variant(kcal/mol) (kcal/mol.K)

AGU (37°C)(kcal/mol)

N58AH18QPseudo wild typeQ15IQ15I/K108R

114.8146.8144.7126.7176.8

0.3620.4580.4480.3900.541

43.847.549.652.053.9

2.44.65.35.48.5

The analysis used the calorimetrically determined value of ACp for bamase(1.4 kcal/mol.K) from the literature (Griko el al., 1994). AHm and A5m arevalues at the midpoint temperature for unfolding of the respective variants(Ta). ACU was extrapolated to 37°C using these parameters. The standarderrors of all values are <69b.

Thermal stability of barnase variantsTo measure their thermodynamic stability, all of the bamasevariants constructed for this study were subjected to thermalunfolding experiments. The results are shown in Figure 2. Thedata were fitted to the Gibbs-Helmholtz equation (see Materialsand methods), and the resulting thermodynamic parametersare given in Table I. Also shown in Table I is the free energyof unfolding of the variants calculated at 37°C, the temperatureof the in vivo experiments of this study. The stabilities andmelting temperatures of the variants are in excellent agreement,as shown previously for other site-specific variants of bamase(Kellis et al., 1989). Thus we have created an assortment ofbamase variants with melting temperatures (defined as fractionnative = 0.5) spanning 10°C and AGU values covering a rangeof 6 kcal/mol, with the pseudo wild-type enzyme in the middle.

In vivo half-lives of barnase variantsAn E.coli strain (JM105) was transformed with plasmidsencoding the five bamase variants. Standard pulse-chaseexperiments were performed to measure the rate of degradationof each variant in the periplasm during overproduction. Cellswere grown to mid-log phase, induced with IPTG, pulse-labeled with [35S]methionine and transferred to medium con-taining excess unlabeled methionine (for details see Materials

1199

W^.Kwon, N.A.Da Silva and J.T.Kellis Jr

(a) Chasing Time (hrs)0 0.25 1 2 3 4 6

p •»-

p •

N58A

H18Q

pWT

Q15I

Q15I/K108R

2 3 4Chasing Time (hrs)

Fig. 3. (a) Pulse-chase of barnase variants. P, pre-bamase; M, maturebamase. (b) Normalized amount of mature bamase. Normalization wasperformed by dividing the amount of mature bamase at each chasing timeby the amount of mature barnase at the beginning of the chase. Specificgrowth rates were monitored throughout the culture and induction periods,and the differences were <5% up to 4 h post-induction.

and methods). At various times indicated in Figure 3, cellswere solubilized in SDS-PAGE buffer and the proteins fraction-ated on 15% polyacrylamide gels. The rate of periplasmicproteolysis was measured by determining the disappearanceof labeled barnase by phosphorimage analysis. As seen inFigure 3, while 70% of the pseudo wild-type barnase remainedat the end of the pulse-chase (6 h), <30% of the thermo-dynamically destabilized variants (H18Q and N58A) remained.The half-lives of these variants in the periplasm of E.coli weremuch shorter than that of the pseudo wild-type enzyme. Incontrast, no significant decrease was observed for the moststabilized barnase mutant (Q15I/K108R). For a stabilizedvariant (Q15I), 80% remained after 6 h. A comparison of thein vivo half-lives and in vitro stability measurements for allbarnase variants clearly demonstrates a correlation between

1200

Table II. Comparison of the change in melting temperature, free energy ofunfolding relative to pseudo wild-type bamase, periplasmic degradation, andyield of barnase variants

Variant AACU (37°C) Amount remaining Final yield(kcal/mol) after 3 h chase (%) (mg/1)

N58AH18QPseudo wild typeQ15IQ15I/K108R

-5.8-2.1

02.44.3

-2.9-0.7

00.13.2

20377286

100

2-414-2025-3035-4040-45

Note the concordance between all of these parameters.

(a)

- • P- * M

1II8Q pWT QI5I QI5I K108K NS8A

(b)

1.6

1.4 -

1.2 -

1.0-

0.8-

"5 0.6 H

0.4 -

S

2 0.0

PrebamascMature barnase

N58A H18Q pWT Q151 Q15I/K108R

Fig. 4. (a) Western blot analysis of barnase expression. P, pre-barnase;M, mature bamase. The same cell concentrations were loaded in each lane,(b) The final yield of bamase. The graph shown is the result of threeseparate experiments. The amounts of pre- and mature bamase arenormalized by taking the amount of mature bamase for the pseudo wildtype in the periplasm as 1.0.

the two values (Table II). These results verify that thermalstability and proteolytic susceptibility of a protein are inverselyrelated in the periplasm, as in the cytoplasm of E.coli.

Increase in the yield of exported barnaseThe yield of pre- and mature bamase was determined byimmunoblotting. As shown in Figure 4, the amount of maturebarnase variants in the periplasm at the end of 2 h post-induction varied from 10 to 150% of the pseudo wild type;the amount of pre-bamase in the cytoplasm varied from 5 to500% of the pseudo wild-type. As with the degradation rates,the amounts of protein correlated with the thermal stabilitiesof the variants. Because the thermodynamic stability of pre-bamase was found to be the same as mature bamase (unpub-lished data), it is possible that the effect of thermal stabilityon proteolysis is more extreme in the cytoplasm than in theperiplasm, or faster folding variants get stuck inside the cell

Protein stability and periplasmic expression

(Hardy and Randall, 1991). However, this conclusion awaitsfurther studies.

By exporting barnase to the periplasm of E.coli, we routinelyproduced 25-30 mg/1 of the pseudo wild-type barnase. Whenmutations were introduced into the barnase gene to improvethe thermal stability of the protein, the final yield was increasedup to 40-̂ 45 mg/1. On the other hand, for a less stable variant(N58A), the final yield was decreased to 2—4 mg/1 (Table II).Therefore, engineering the protein for higher thermal stabilitywas successful in increasing the yield of the overexpressedrecombinant barnase. In control experiments, we checked forthe presence of inclusion bodies and for leakage of bamasefrom the periplasm of E.coli to the culture medium. None ofthe variants tested in this study exhibited the formation ofinclusion bodies or any significant leakage of periplasmicproteins (results not shown).

DiscussionThis study with five barnase variants of differing thermalstabilities clearly demonstrates the relationship between theconformational stability determined in vitro and the half-lifeof the protein in the periplasm of E.coli determined in vivo.The barnase mutations were designed to be non-disruptive,and no known proteolytic cleavage sites were introduced ordeleted by the mutations. We therefore conclude that thedifferent degradation rates were a result of the altered structuralstabilities of the variants. This is the first report to show sucha relationship in the periplasm of E.coli. In addition, we haveverified that the final yield of the exported foreign protein inthe periplasm of E.coli can be increased by improving thethermodynamic stability of the protein (Figure 4). It shouldbe noted that the mutations which increased the stability ofbarnase were outside the active site; it has been shown thatthe thermodynamic stability of barnase can be increasedwithout altering its catalytic activity (Serrano et al., 1993).

There are numerous ways of making proteins more thermo-dynamically stable. Creating disulfide bonds (Perry and Wetzel,1984; Gokhale et al., 1994), engineering metal chelation sitesinto a protein (Kellis et al., 1991) or even substituting one ora few amino acids (Eijsink et al., 1992; Miyazaki et al., 1993;Serrano et al., 1993; Anderson et al., 1994) are just a fewexamples of effective strategies. As demonstrated in this study,regardless of the stabilization method used, efforts to increasestability should be viewed as important not only for in vitropurposes [such as prolonged shelf-life (Brems et al., 1992)and preparing enzymes for harsh reaction conditions], but alsofor the in vivo purpose of increasing overall production.

During the export process of a protein, a pre-protein isthought to remain in the so-called 'secretion-competent state',i.e. the pre-protein must be in a non-folded conformation topass through the membrane (Park et al., 1988; Hardy andRandall, 1991). Once the pre-protein attains its folded state, itbecomes 'secretion incompetent' and cannot pass through themembrane. Any change in the amount of unfolded pre-proteincan be attributed to the kinetics and/or thermodynamics ofprotein folding. While slower folding would increase the poolsize of pre-protein to interact with the secretion machinery, adecrease in thermodynamic stability would also have the sameeffect of increasing the fraction of unfolded pre-protein. Thepositive effect of increasing the unfolded pre-protein fractionhas been observed for the export of a slower folding variantof bovine pancreatic trypsin inhibitor (Nilsson et al., 1991)and of ribose-binding protein (Teschke et al., 1991) in E.coli

and the import of a destabilized mouse dihydrofolate reductasevariant into yeast mitochondria (Vestweber and Schatz, 1988).However, the lack of a tightly folded structure is known toaccelerate protein degradation (Parsell and Sauer, 1989; Inoueand Rechsteiner, 1994). Therefore, a competition may existbetween the export and the degradation of pre-protein in thecytoplasm. Furthermore, as we have shown here, a destabilizedmature protein can also be degraded faster in the periplasm.Thus it is imperative to determine which is the more dominantfactor affecting the final yield of exported protein.

In our studies with bamase we have observed that morebarnase is exported in a shorter time when the mutant isthermodynamically destabilized (W.S.Kwon et al., manuscriptin preparation; this finding also ruled out the possibility ofsecretion machinery saturation as a limiting factor for theexpression of exported barnase). However, as seen in Figure3, a major determinant for final protein yield was not theexportability of the pre-protein but the turnover rate of themature protein in the periplasm. Therefore, when barnase wasstabilized through mutation, the metabolic stability of theprotein in both the cytoplasm and the periplasm was improved,resulting in an increased final yield.

In the case where the exportability of pre-protein dominatesthe yield, engineered stabilization may actually decrease theyield of exported protein. The two opposing effects (exportabil-ity and degradation) of increasing the unfolded fraction of pre-protein by protein engineering can pose a true optimizationproblem. An optimum thermodynamic stability of a proteinmay exist which promotes translocation of the protein across themembrane but thwarts proteolysis. This remains an interestingchallenge in the use of protein engineering for the purpose ofincreasing the final yield of exported recombinant proteins.

AcknowledgementsWe are indebted to Prof. Alan Fersht for plasmid pTZ416-2. This work waspartially supported by grant BCS-911021 (to J.T.K. and N.A.D.) from theNational Science Foundation W.S.K. acknowledges a predoctoral traininggrant fellowship from the National Institutes of Health (THS-NIH grant no.5T32 GM07311-20).

ReferencesAnbaJ., Bemadac.A., Lazdunsld.C. and PagesJ.M. (1988) Biochimie, 70,

727-733.Anderson,D.E., HurleyJ.H., Nicholson,H., Baase,W.A. and Matthews,B.W.

(1994) Protein Sci., 2, 1285-1290.Baneyx,F. and Georgiou,G. (1990) J. Bacteriol., 172, 491^*94.Baneyx.F., Ayling.A., Palumbo.T., Thomas.D. and Georgiou.G. (1991) Appl.

Micwbiol. Biotechnol., 33, 14-20.Battistoni.A., Carri.M.T., Mazzetti,A.P. and Rotilio.G. (1992) Biochem.

Biophys. Res Commun., 186, 1339-1344.Brems.b.N. el al. (1992) Protein Engng, 5, 519-525.ChesshyreJ.A. and Hipkiss,A R. (1989) Appl. Micwbiol. Biotechnol., 31,

158-162.Clackson.T., GUssow.D. and Jones.P.T. (1991) In McPherson.MJ., Quirke.P.

and Taylor.G.R. (eds), PCR: A Practical Approach. IRL Press, Oxford, UK.Decad,G.M. and Nikaido.H. (1976) J. Bacteriol, 128, 325-336.Eijsink,V.G., Vriend.G., van den Burg.B., van der ZeeJ.R. and Venema,G.

(1992) Protein Engng, 5, 165-170.Gentz,R., Kuys.Y., Zwieb.C, Taatjes.D., Taatjes.H., Bannwarth.W., Stueber.D.

and IbrahimU. (1989) /. Bacteriol, 170, 2212-2220.Gokhale,R.S., Agarwalla.S., Francis.V.S., Santi.D.V. and BalaramJ5. (1994) J.

Mol. Biol, 235, 89-94.Griko.Y.V., Makhatadze.G.I., Privalov.P.I. and Hartley.R.W. (1994) Protein

Sci, 3, 669-676.Hardy.SJ.S. and Randall.L.L. (1991) Science, 251, 439-443.Inouej. and Rechsteiner,M. (1994) /. Biol. Chem., 269, 29241-29246.KellisJ.T.Jr, Nyberg.K. and Fersht,A.R. (1989) Biochemistry, 2S, 4914-4922.KellisJ.T.Jr, Todd.RJ. and Arnold,F.H. (1991) Bio/Technology, 9, 994-995.

1201

W^.Kwon, N.A.Da SUva and J.T.Kellis Jr

Laemmli.U.K. (1970) Nature, 227, 680-685.Maunzi,M.R. (1992) Expcrientia, 48, 178-201.McLendon.G. and Radany,E. (1978) J. Biol. Chem., 253, 6335-6337.Meerman.HJ. and Georgiou.G. (1994) Bio/Technology, 12, 1107-1110.Meiering.E.M., Serrano.L. and FershtAR. (1992) J. Mol. Biol., 225, 585-589.Miyazaki.Y., Matsufuji,S., Murakami.Y. and Hayashi.S. (1993) Eur. J.

Biochem., 214, 837-844.Nilsson.B., Berman-Marks.C, KuntzJ.D. and Anderson.S. (1991) J. Biol.

Chem., 266, 2970-2977.Pace.C.N. (1986) Methods Enzymol., 131, 266-279.Paddon.C. and Hartley.R. (1987) Gene, 53, 11-19.Partc.S., Liu.G., Topping.T., Cover.W. and Randall.L.L. (1988) Science, 239,

1033-1035.Parsell,D.A. and Sauer.R.T. (1989) / Biol. Chem., 264, 7590-7595.Perry.LJ. and Wetzel.R. (1984) Science, 226, 555-557.Pine.M.J. (1970) J. Bacteriol, 103, 207-215.Pine,MJ. (1973) J. Bacteriol, 115, 1097-1116.SambrookJ., Fritsch.E.F. and Maniatis.T. (1989) Molecular Cloning: A

Laboratory Manual. 2nd edition, Cold Spring Harbor Laboratory Press,Cold Spring Harbor, NY.

SanchoJ. and Fersht,A.R. (1992) J. Mol. Biol., 1U, 14\-141.Sanger.F., Nicklen.S. and Coulson.A.R. (1977) Proc. Natl Acad. Sci. USA,

74, 5463-5467.Serrano.L. and Fersht,A.R. (1989) Nature, 342, 296-299.Serrano.L., Day.A.G. and Fersht.A.R. (1993) J. Mol. BioL, 233, 305-312.Strauch.KX., Johnson.K. and BeckwithJ. (1989) /. Bacteriol., 171, 2689-

2696.Talmadge.K. and Gilbert,W. (1982) Proc. Natl Acad. Sci. USA, 79, 1830-1833.Teschke.C.M., KimJ., Song.T., Park.S., Park.C. and Randall.L.L. (1991) J.

Biol. Chem., 266, 11789-11796.Vestweber.D. and Schatz,G. (1988) EMBO J., 7, 1147-1151.

Received March 27, 1996; accepted August 13, 1996

1202