Protein Engineering vol.6 no.7 pp.733-738, 1993

Breakdown in the relationship between thermal andthermodynamic stability in an interleukin-1/3 point mutant modifiedin a surface loop

Boris A.Chrunyk1 and Ronald Wetzel2

Macromolecular Sciences Department, SmithKline Beecham Pharmaceuticals,709 Swedeland Road, King of Prussia, PA 19406, USA

'Present address: Central Research Division, Pfizer Inc., Eastern Point Rd,Groton, CT 06340, USA2To whom correspondence should be addressed

Sequence variants of the /3-barrel protein interleukin-1/3 havebeen analyzed for their stabilities toward irreversible thermalinactivation by monitoring the generation of light scatteringaggregates on heating. Hie derived temperatures for the onsetof aggregation ( 7 ^ values) correlate well with the freeenergies of unfolding of these proteins with the exception ofone variant, Lys97—Val (K97V), which undergoes aggrega-tion at a temperature 7°C lower than expected based on itsthermodynamic stability. This lower than expected thermalstability may be due to generation of an aggregation-proneunfolding intermediate at a temperature lower than the 7m

of the global transition. This hypothesis is supported by thelocation of residue 97 in the long 86-99 loop which hasstructural features suggesting it may comprise a small,independent folding unit or microdomain. The excellentcorrelation of thermal and thermodynamic stabilities of sevenof the eight variants tested is consistent with accepted modelsfor thermal inactivation of proteins. At the same time the poorfit of the K97V variant underscores the risk in using thermalstability data in quantitative analysis of mutational studiesof the folding stability of proteins.Key words: aggregation/microdomain/unfolding intermediate

IntroductionThermodynamic and thermal stability are related but not identicalproperties of a protein (Wetzel ex al., 1990; Wetzel, 1992).Thermodynamic stability is a measure of the energy required toinduce cooperative unfolding of the protein under reversibleconditions, while thermal stability is a measure of the energyrequired to induce irreversible loss of protein structure and/orfunction. The two processes are usually related because theaggregation event which is often the source of irreversibilityinvolves an unfolded state of the protein generated in thecooperative unfolding transition (Wetzel et al., 1990). Althoughthere is a substantial data base supporting this generalization [forreviews, see Mitraki and King (1989) and Wetzel et al. (1990)]no formal studies have been reported examining this relationshipin a series of point mutants. Despite this, thermal stabilities areoccasionally used as substitutes for thermodynamic stabilities inmutational analyses of the molecular basis of protein stability.

As part of an ongoing study of the molecular basis of inclusionbody formation in the expression of interleukin-1/3 in Escherichiacoli (Wetzel and Chrunyk, 1992; Chrunyk et al., 1993a,b) wehave purified and analyzed the thermodynamic stabilities of aseries of point mutants of this protein (Chrunyk et al., 1993a).Here we describe a method for assessing and ranking the

stabilities of proteins towards thermally induced irreversibleaggregation and present a direct comparison of these thermalstability values with thermodynamic stabilities of the samemutants. The data for most sequence variants examined isconsistent with the simple model that thermally induced aggrega-tion of IL-1/3 is mediated by cooperative global unfolding. Inaddition however, one sequence variant, Lys97—Val, undergoesthermal aggregation at a temperature substantially lower thanexpected based on its thermodynamic stability. The location ofresidue 97 in an organized surface loop suggests a mechanismfor this effect.

Materials and methodsProduction of IL-lfi sequence variantsThe construction of IL-1/3 variant-encoding cDNA and theproduction and purification of proteins is described elsewhere(Chrunyk et al., 1993a). Briefly, IL-1/3 was produced fromplasmid-encoded cDNA by thermal or nalidixic acid (Mott et al.,1985) induction of the X PL promoter in E.coli (Myers et al.,1987). All proteins were purified from the soluble fraction ofnative lysates using either ion exchange (Myers et al., 1987)or hydrophobic interaction (Roe, 1989) chromatography.All proteins were purified to homogeneity as assessed bySDS-PAGE.

Temperature for the onset of aggregationThe buffer used in aggregation experiments was 10 mM MES,2 mM EDTA, 90 mM NaCl, pH 6.5, 10 mM DTT. Temperatureof aggregation was determined by fluorescence light scatteringexperiments. Samples of 100 /xl containing 1 mg/ml purifiedprotein were heated in a microcuvette (Hellma Cells) from 25to 60°C at 0.5°C/min. Temperature was controlled by program-ming a circulating water bath and confirmed during a blank runby monitoring temperature directly in the cell using a thermo-couple. Aggregation was monitored by light scattering in theSLM-8000 spectrofluorimeter with excitation and emissionwavelengths set at 500 nm. Data was collected every 30 s. Ineach case a sharp increase in signal indicated the onset ofaggregation with generation of light scattering particles. Thetemperature of aggregation was obtained by linear extrapolationof the heating curves to the temperature axis. Since the slopeof these curves changes slightly as light scattering develops, wedevised an objective method for the graphical analysis. In thismethod the point on the curve representing 40% of uie finalamplitude is identified, then the linear portion of the curve atamplitudes immediately higher man the 40% mark is used forthe extrapolation. Other graphical methods produce slightlydifferent numbers, but choice of graphical analysis method doesnot change the basic trends described in diis paper. Studies wereconducted in reducing agent to suppress aggregation via disulfidebond formation by the unpaired Cys residues of the protein. Theeffectiveness of this treatment was confirmed by the absence ofmultimeric forms in non-reducing SDS-PAGE ofiodoacetamide-capped samples of thermal aggregates (data notshown).

733

at Old D

ominion U

niversity on July 4, 2014http://peds.oxfordjournals.org/

Dow

nloaded from

B.A.Chrunyk and R.Wetzd

FTg. 1. Ribbon diagram of wild type main chain of interieulrin-1/S generated from the X-ray crystallographic coordinates of Verrapandian a aL [BrookhavenData Base (Bernstein a aL, 1977; Abola et al., 1987)] using the graphics package Insight n on a Silicon Graphics IRIS. The side chains of the two residuesmutated, positions 9 and 97, as well as the side chain of the fluorescence probe Trpl20, are shown in CPK representation.

Fig. 2. Ribbon diagram (Insight) showing the residues in the 86—99 loop of interleukiD-1/3. The loop is in the same orientation as it is in Figure 1 except his tilted toward the reader to reveal the intra-loop pairing as viewed from the center of the protein. The clustering of hydrophobic side chains and the cispeptidyl—prolyl bond at Pro91 are evident in the CPK representations of these atoms shown in Figure 2a (left)- Figure 2b (right) shows the same residues andorientation plus the side chain of Tyr68 and the ribbon representation of the 67-71 backbone.

734

at Old D

ominion U

niversity on July 4, 2014http://peds.oxfordjournals.org/

Dow

nloaded from

Thermal and tbennodynamic stability relation

ResultsThe structural locations of the mutations explored in this studyare shown in Figure 1. Position 9, Thr in the wild type, is locatedin an extended chain which forms a strand in the j3-barrel coreof this all-/3 protein, with a side chain pointing toward the solventbut still mostly buried. Position 97, Lys in the wild type, is locatedin a long loop which connects two /3-strands, with a side chainthat packs on the outer surface of the molecule and is thereforeexposed. This loop is shown in more detail in Figure 2. Thestructure suggests that the stability and structure of the loop isderived to a large extent from the interaction of a number of sidechains from the loop which interact with each other (Figure 2a)and with Tyr 68 (Figure 2b) in a hydrophobic cluster. In additionto these hydrophobic contacts the internal structure of the loopis supported by hydrogen bonds and side-chain electrostaticinteractions in the three turns within this loop (Clore et al., 1991).Position 97 is not involved in the hydrophobic cluster but ratheris turned outward. The loop is also distinguished by the presenceof a cis-peptidyl—prolyl bond at Pro91. Mean isotropictemperature factors from X-ray crystallography (Finzel et al.,1989) suggest that the 86-99 loop is of only modest flexibility.Detailed NMR studies have uncovered two conformational formsof native IL-1/3 which differ primarily in the positions of residuesof the 86-99 loop (Clore et al., 1990; DriscoU et al., 1990).

60000

50000

•j 40000 •

- 30000

20000

20 25 30 35 40 45 50 / 55 ' 60 65 70

5O0OO

- 40000

20000

CD

O

25 35 40 45 50

Temperature 'C

70

Fig. 3. 7"™ determinations of IL-1/3 derivatives, (a) Wild type IL-1/3 andIL-10 (K97V). Extrapolation of the scattering curve back to baseline yieldsa value for wild type of 58.4°C and for the K97V mutant of 53.0°C. Thehorizontal line shows the apparent fluorescence value at 40% of themaximum signal, the value used to define the point of extrapolation(Materials and methods), (b) T9A run on two successive days illustratingthe reproducibility of the determination and of the aggregation process.

Figure 3 shows examples of the estimation of thermal stabilitiesof IL-1/3 mutants. A 1 mg/ml solution of a purified IL-1/3 variantin native buffer was heated in a cuvette so that the temperatureincreased by 0.5°C/min. Figure 3a shows that each sampleundergoes a dramatic increase in apparent fluorescence at acharacteristic temperature, coincident with the generation of lightscattering precipitate. Extrapolation of the scattering curve to thebaseline as described in Materials and methods generates atemperature value for the onset of aggregation, 7 ^ . The 7 ^value for the wild type IL-1/3 is 58.6°C and for the K97V mutantis 53.4CC. Figure 3b shows the 7 ^ curves for two samples ofthe T9A mutant, prepared from the same protein stock solution,run on two successive days.

Table I lists the 7 ^ values for the sequence variantsexamined in this study. It also lists the AAG°H2O values (thedifference between the free energies of unfolding for a mutantand for the wild type) for each sequence variant, as determinedin reversible unfolding experiments in guanidine hydrochloridemonitoring the fluorescence of the single Trp at position 120[described elsewhere (Chrunyk et al., 1993a)].

Figure 4 shows the correlation between the 7 ^ values forirreversible thermal aggregation and the AAG°HjO values forreversible unfolding in Gdn—HC1. The figure shows the least-squares fit of all of the data points except that for the K97Vmutant, which fits the data with a correlation coefficient of 0.901.The K97V variant is well off the line. Although K97V is more

Table I. Thermal and thermodynamic stabilities of sequence variants ofinterleukin-lj3

Variant(kcal/mol)1

Wild typeT9AT9LT9QT9GK97RK97GK97V

-0 .8-0 .7-1 .9-2 .6-0 .5-1 .2

0.8

±±±±±±±

0.20.10.40.50.10.20.2

58.658.657.453.052.656.854.453.4

"Data from Chrunyk ex al. (1993a).

Fig. 4. Correlation of thermodynamic ( A A C ^ Q ) and thermal ( 7 ^ )stabilities of sequence variants of IL-10. Errors shown are: ;y-axis, from theF^p plot fits for the A A C J ^ Q determination of each variant (Chrunyk etal., 1993a); x-axis, ±0.5°C, estimated maiimnm error in determination ofinternal cell temperature derived from direct temperature probing of thecuvette in a blank run.

735

at Old D

ominion U

niversity on July 4, 2014http://peds.oxfordjournals.org/

Dow

nloaded from

B.A.Chrunyk and R.Wetzel

stable than the wild type to reversible unfolding it is less stablethan wild type to irreversible thermal aggregation.

Although, as described here, 7"agg values appear closely tiedto a thermodynamic Tm and/or AAG, these values may beinfluenced to some extent by solvent conditions and heating rate.Not surprisingly Tm values are dependent on protein concen-tration. The T^g ofK97V decays from 53.4°C to 49.2°C whenthe concentration of K97V is increased from 1 to 4 mg/ml; thevariation of 7 ^ with K97V concentration suggests a limitingvalue of ~48°C for this variant (Chrunyk et al., 1993a). TheJagg values shown in Table I were all generated from studies at1 mg/ml. It would seem legitimate to compare Tagg values ifthey are all determined at the same protein concentration andthe quality of the correlation shown in Figure 4 supports this.As shown in Figure 3b, when experimental parameters are fixedthe reproducibility of the measurement is surprisingly good.

Discussion

The observation of the steep temperature dependence of proteinaggregation near the beginning of this century was the firstindication of the cooperative nature of protein stability (Chickand Martin, 1912). Since then the development of solutionbiophysical methods has allowed the characterization of thecooperative unfolding of proteins under reversible conditions,making the process amenable to thermodynamic analysis (Schulzand Schirmer, 1979; Creighton, 1984). Nonetheless, kineticstability studies in which loss of protein activity or solubility ismonitored under irreversible conditions continue to be useful incharacterizing protein stability, especially where protein samplesare only available in small amounts or where reversible unfoldingconditions have not been identified.

The cooperative nature of protein folding and stability wasapparent in the thermal aggregation results of Chick and Martin(1912) because in many proteins thermal aggregation is dependentupon prior thermal unfolding (Wetzel et al., 1990). There istherefore in theory a tight link between reversible unfolding andirreversible aggregation/inactivation. Furthermore, a number ofcareful studies have implicated unfolding intermediates in thethermal aggregation process [for reviews, see Mitraki and King(1989) and Wetzel et al. (1990)]. For these reasons stability dataacquired from studies under irreversible conditions hasoccasionally been used in lieu of true thermodynamic stabilitydata to analyze the structural basis of protein stability.

There are significant potential problems with this practicehowever. Proteins can undergo inactivation by many processesbesides aggregation, which may not always be unfoldingdependent (Wetzel et al., 1990). Furthermore, it is known thatin some cases protein variants can be significantly more stabletoward irreversible thermal aggregation than expected based ontheir stability toward reversible unfolding (Wetzel et al., 1990;Wetzel, 1992). For example, introduction of a disulfide bondinto T4 lysozyme increased its Tm in reversible unfolding byonly 3°C (Wetzel et al., 1988) but increased its stability toirreversible inactivation by 40°C or more (Wetzel et al., 1990).The simplest explanation for this additional stabilization is thatthe cross-link decreases the rate by which a thermally generatedunfolded state undergoes aggregation; in essence, if the thermallyunfolded state is not susceptible to aggregation the protein willremain in solution at the elevated temperature and refold/reactivate on cooling (Wetzel et al., 1990; Wetzel, 1992).

In this paper we hope to communicate two major points. First,that linkage between thermodynamic and thermal stability can

be surprisingly tight, so that a parameter which measures stabilitytoward irreversible thermal aggregation can be a good indicatorof the relative thermodynamic stabilities of a series of mutants.The second point, however, is that this generally tight relation-ship between thermodynamic and thermal stabilities can breakdown in ways at present unpredictable for particular mutants.Specifically we report here a mutation which produces a proteinwhich is considerably less stable to irreversible thermal aggrega-tion than expected based on its thermodynamic stability.

In a recent series of experiments we surveyed a series of 67point mutants in IL-1/3 for their tendencies, relative to the wildtype protein, to form inclusion bodies when produced in thecytoplasm of E.coli (Wetzel and Chrunyk, 1992; Chrunyk et al.,1993a,b). The mutants surveyed were originally constructed forthe study of structure—function relationships in receptor andantibody binding (Simon et al., 1993) and thus are not biasedby design criteria based on protein stability considerations.Sequence variants were purified from the soluble fraction ofE.coli lysates and evaluated for their diermodynamic stabilitiesagainst reversible unfolding induced by guanidine hydrochloride(Chrunyk et al., 1993a). Figure 1 shows the structural locationsof the two sites concentrated on, positions 9 Qocated in /3-sheet)and 97 (located in a surface loop). Of all the variants tested onlyone, resulting from the replacement of Lys97 with Val (K97V),was more thermodynamically stable than the wild type (Table I).

Previously, thermal stabilities (i.e. to irreversible inactivation)have been assessed in a number of ways. Typically samples areincubated at the test conditions then restored to mild conditionsand assessed for their enzymatic, immunochemical or otherbiological activity. Alternatively, in cases where diermalaggregation is the predominant mechanism of inactivation it issufficient to monitor the generation of light scattering particlesby the development of turbidity in a sample heated in a cuvette(Mulkerrin and Wetzel, 1989; Wetzel, 1992). We describe herea modification of the light scattering method which leads to aparameter, the temperature for the onset of aggregation or 7"^,which can be easily defined and reproducibly measured and whichcorrelates well with the thermodynamic stabilities of a series ofIL-1/3 mutants.

Figure 3a shows how these values are obtained. A 1 mg/mlsample of protein is heated in a jacketed fluorescence cell by acirculating water bath to generate a ramp rate of 0.5°C/min.When the protein begins to aggregate it generates precipitatewhich scatters the input excitatory radiation generating anapparent emission signal which is read as fluorescence. This rapidsignal increase is extrapolated back to baseline to generate the7 ^ value, the temperature at which thermally induced aggrega-tion is initiated.

In attempting to relate thermal stability in vitro to anothertemperature-dependent property such as inclusion body formationwe expect the Tm in some cases to be more relevant than thecalorimetrically determined Tm. This is because the 7 ^ valueshould be predictive of the lowest temperature at whichaggregation is observable, while die calorimetric Tm, even if itaccurately measures an unfolding event which is tightly linkedto the aggregation, will fall several degrees higher than theminimum effective temperature, at the midpoint of the unfoldingtransition. This is in fact seen in the comparison of thecalorimetric Tm and the 7 ^ values of several interleukin-1/3variants. For example, me calorimetric I'm for the wild typeprotein (obtained under conditions which generate appreciablesample precipitation) is 62°C (C.Brouilette, personal communica-tion) while its Tm is 58.6°C (Table I).

736

at Old D

ominion U

niversity on July 4, 2014http://peds.oxfordjournals.org/

Dow

nloaded from

Thermal and thermodynamlc stability relation

The Tagg values determined for the set of IL-1/3 variants arelisted in Table I. None of the proteins is more stable towardirreversible thermal aggregation than the wild type. The leaststable mutant T9G is 6°C less stable than the wild type proteinby this measure of stability. Figure 4 shows the relationshipbetween the 7]™ values described here and the correspondingfree energies ofunfolding. Seven of the eight proteins are well-correlated, exhibiting a correlation coefficient of 0.901 in a leastsquares fit, as shown in Figure 4. The eighth data point, asignificant outlier, is the variant K97V. When this point isincluded in the linear regression analysis the correlation co-efficient drops to 0.233 (line not shown). By extrapolation ofthe fit for the other seven variants K97V would be expected toexhibit a 7 ^ of ~60.5°C based on its thermodynamic stability.In fact the experimental 7 ^ for this protein is 53.4°C, 7°Clower than predicted by the correlation. Perhaps more signific-antly, Figure 4 shows that an experimenter taking the 7 ^ valueof K97V as an indication of its thermodynamic stability wouldunderestimate that stability by over 2.5 kcal/mol.

What is the molecular basis for this dramatic effect? Anattractive hypothesis is suggested by the protein structure in theneighborhood of position 97. This residue is on the C-terminusof a long loop which bridges two /3-strands of the structure. Thisloop exhibits an unusually high degree of internal structure. Thereare three turns in the loop, at residues 86-89, 89-92 and97-100 (Clore et al., 1991). In addition, inspection of thegraphics representation shows a cluster of hydrophobic side chainscontributed by residues Val85, Tyr90, Met95 and Phe99, inaddition to the core residue Tyr68 (Figure 2). We hypothesizethat this loop forms a microdomain which undergoes independentthermal unfolding/misfolding prior to the global unfolding of the/3-barrel and in doing so forms an unfolding intermediateconsisting of an intact /3-barrel and an unfolded or misfolded86—99 loop. This hypothesis is especially attractive given theobservation in the NMR structure of two interchanging groundstate conformations involving changes in environments of 19residues, including residues at positions 68, 81, 82, 85, 86, 88,89, 90, 92, 93 and 96 (Clore et al., 1990; Driscoll et al., 1990).

It seems reasonable that this hypothetical partial unfoldingmight go undetected in the Gdn-HCl unfolding experiments,which rely on the fluorescence of the single Trp at position 120(Figure 1). At the same time a fluorescence change does occurat low Gdn-HCl concentrations in all IL-1/3 molecules examined(Chrunyk et al., 1993a,b) and it remains possible that this changeis linked to the proposed unfolding of the 86—99 loop.

To explain the data in Figure 4 one of two mechanisms ispossible. Either (i) the 86-99 loop must unfold independentlyonly in the case of the K97V mutant or (ii) an independentunfolding in this region exhibited by all the IL-1/3 molecules mustgenerate an aggregation-prone intermediate only for this mutant.The latter possibility is consistent with the increased potentialfor hydrophobic interaction anticipated for a Lys—Val replace-ment. If further experiment validates this explanation, the roleof the 86—99 loop in protein aggregation would thus be somewhatakin to that of the third helix of mammalian growth hormones(DeFelippis et al., 1993).

Further mutagenesis and biophysical analysis will be requiredto test the above hypothesis. However, the experiments describedhere show that useful insights into protein stability can be obtainedby the comparative analysis of thermodynamic and thermalstabilities. This data may help explain the high tendency for theK97V mutant to form inclusion bodies in vivo (Chrunyk et al.,1993), as well as the unique ability of this mutant of all IL-1/3

mutants so far examined to generate light scattering aggregatesduring folding in vitro (Chrunyk and Wetzel, 1993). Theinvolvement of folding intermediates in the formation of threedifferent IL-1/3 aggregates (the folding-dependent aggregate citedabove, the thermally induced aggregate described in this paperand inclusion bodies formed in vivo) is supported by recent FTIRstudies on the aggregates (Oberg, Chrunyk, Wetzel and Fink,manuscript in preparation). Structural and folding implicationsof the in vitro aggregation studies reported here, if verified byfurther experiment, would suggest that protein aggregationphenomena in general may be guided by and reflect importantaspects of the protein folding process. This concept is alsosupported by recent work on al-antitrypsin (Lomas et al., 1992)as well as the observation of dramatic mutational effects on levelsof inclusion body formation in a number of systems, includingIL-1/3 (Wetzel, 1992a,b). The importance of loop stability toin vivo aggregation is also illustrated by the recent identificationof an Asp -~ His mutation in the 64 - 76 loop of human lysozymeas a lesion leading to amyloid fibril formation in a rare humangenetic disorder (Pepys et al., 1993).

Since most of the sequence variants tested provide a strongcorrelation between thermal and thermodynamic stability thiswork might be viewed as establishing an experimental basis forthe often tacitly made assumption of interchangability of thesetwo measures of protein stability. The dramatic example of theK97V mutant however should provide a cautionary tale todiscourage the use of such thermal stabilities in the quantitativeanalysis of the molecular basis for folding stability of proteins.The real value of measures of kinetic stability to irreversibleinactivation, besides their contribution to practical evaluation ofprotein stability, may prove to be the additional insights they allowwhen viewed in parallel with thermodynamic data.

AcknowledgementWe gratefully acknowledge Dr Jim Nicholson for aid with the computer graphics.

ReferencesAbola.E., Bernstein.F.C, Bryant.S.H., Koetzle.T.F. and WengJ. (1987) In

Allen.F.H., Bergerhoff,G. and Sievers.R. (eds), Cryaallographic Databases—Information Content, Software Systems, Scientific Applications. DataCommission of the International Union of Crystallography, Bonn, Germany,pp. 107-132.

Bernstein,F.C, Koetzle.T.F., Williams.G.J.B., Meyer.E.F.Jr, Brice.M.D.,RodgersJ.R., Kennard.O., Shimanouchi.T. and Tasumi,M. (1977) J. Mol.Biol., 112, 535.

Chick.H. and Martin.C.J. (1912) J. PhysioL, 45, 261-295.Chrunyk.B. and Wetzel,R. (1993) Manuscript in preparation.Chrunyk.B.A., Evans,J., LillquistJ., Young.P. and Wetzel.R. (1993a) J. Biol.

Oiem., 268, in press.Chrunyk.B.A., Evans J. and Wetzel.R. (1993b) In Cleland.J.L. (ed.), Protein

Folding In Vivo and In Vitro. American Chemkal Society Books, Washington,DC, pp. 46-58.

Clore.G.M., Driscoll,P.C, Wingfield.P.T. and Gronenborn.A.M. (1990)Biochemistry, 29, 7387-7401.

Clore.G.M., Wingfield.P.T. and Gronenborn.A.M. (1991) Biochemistry, 30,2315-2323.

Creighton.T.E. (1984) Proteins: Structures and Molecular Properties.W.H.Freeman, New York, NY.

DeFelippis.M.R., AlteT.L.A., Pekar.A.H., Havel.H.A. and Brcms.D.N. (1993)Biochemistry, 32, 1555-1562.

DriscoU.P.C, Gronenborn.A.M., Wingfield.P.T. and Clore.G.M. (1990)Biochemistry, 29, 4668-4682.

Finzel.B.C, Clancy.L.L., Holknd.D.R., Muchmore.S.W., Watenpaugh.K.D.and Einspahr.H.M. (1989) / Mol. Biol, 209, 779-791.

Lomas.D.A., Evans.D.L., FinchJ.T. and CarreU.R.W. (1992) Nature, 357,605-607.

Mhraki.A. and KingJ. (1989) Biotechnology, 7, 690-697.

737

at Old D

ominion U

niversity on July 4, 2014http://peds.oxfordjournals.org/

Dow

nloaded from

B.A.Chnmyk and R.Wetzd

MotU.E., Grant,R.A., Ho,Y.-S. and Platt,T. (1985) Proc. NatlAcad. Sd. USA,82, 82-92.

Mulkerrin.M.G. and Wetzel.R. (1989) Biochemistry, 28, 6556-6561.Myers.C.A., Johanson.K.O., Miles.L.M., McDevitt.P.J., Simon.P.L.,

Webb.R.L., Chen,M.J., Holskin.B.P., LillquistJ.S. and Young.P.R. (1987)J. Biol. Chem., 262, 11176-11181.

Pepys.M.B., Hawkins.P.N., Booth.D.R., Vigushin.D.M., Tennent.G.A.,Soutar.A.K., Totty.N., Nguyen.O., Blake.C.C.F., Terry.C.J., Feest.T.G.,Zalin.A.M. and HsuanJ.J. (1993) Nature, 362, 553-557.

Roe.S. (1989) In Harris^E.L. V. and Angal.S. (eds), Protein Purification Methods.IRL Press, Oxford, pp. 175-244.

Schulz.G.E. and Schirmer.R.H. (1979) Principles of Protein Structure. Springer-Verlag, Berlin, Germany.

Simon.P.L., Kuma,V., LfllquistJ.S., Bhatnagar.P., Lee^.C, Porter.T.,Greea.D., Satne.G. and Young.P.R. (1993)/ Biol Chem., 268, 9771-9779.

Wetzel.R. (1992a) In Rees.A.R., Stemberg.M.J.E. and Wetzel.R. (eds), ProteinEngineering—A Practical Approach. IRL Press, Oxford, pp. 191-219.

Wetzel.R. (1992b) In Anem.TJ. and Manning.M.C. (eds), Stability of ProteinPharmaceuticals: In Vivo Pathways of Degradation and Strategies for ProteinStabilization. Plenum Press, New York, pp. 43-88 .

Wetzel.R. and Chrunyk.B.A. (1992) In Himmel.M.E. and Georgiou.G. (eds),Biocatatyst Design for Stability and Specificity. American Chemical Society,Washington, DC, pp. 116-125.

Wetzel.R., Perry.L.J., Baase.W.A. and Becktel.W.J. (1988) Proc. NatlAcad.Sd. USA, 85, 401-405.

Wetzel.R., Perry.L.J., Mulkerrin.M.G. and Randall.M. (1990) In Poste.G. andHookJ.B. (eds), Protein Design and the Development of New Therapeuticsand Vaccines; Proceedings of the Sixth Annual Smith, Kline and French ResearchSymposium. Plenum, New York, NY, pp. 79-115.

Received on February 1 1993

738

at Old D

ominion U

niversity on July 4, 2014http://peds.oxfordjournals.org/

Dow

nloaded from