Molecular Microbiology (1995) 16(5), 835-845

MicroReviewRepair, refold, recycle: how bacteria can deal withspontaneous and environmental damage to proteins

Jonathan E. Visick and Steven Clarke*Department of Chemistry and Biochemistry, and theMolecular Biology Institute. University of California.Los Angeles, California 90024-1569, USA.

Summary

Proteins, like DNA, are subject to various forms ofdamage that can render them non-functional. Confor-mational changes and covalent chemical alterationsoccur spontaneously, and the rates of these reactionscan be increased by environmental stresses such asheat, oxidative agents, or changes in pH or osmoticconditions. Although affected proteins can tie replacedby de novo biosynthesis, cells — especially thosesubjected to stress or nutrient limitation — havedeveloped mechanisms which can either restoredamaged polypeptides to an active state or removethem. Such mechanisms can spare the biosyntheticcapacity of the cell and ensure that the presence ofnon-functional molecules does not disrupt cell phys-iology. Three major mechanisms, which operate inbacteria as well as eukaryotic organisms, have beendescritied. First, chaperones not only assist in properde novo folding of proteins but also provide an Impor-tant means of restoring activity to conformationallydamaged proteins. Second, enzymatic 'repair' sys-tems exist to directly reverse certain forms of proteindamage, including proline isomerization, methionineoxidation and the formation of isoaspartyl residues.Finally, proteotysis provides a 'last-resort' means ofdealing with abnormal proteins which cannot berepaired. Protein maintenance and repair may be ofspecial importance for bacteria preparing to surviveextended periods In stationary phase: both constitu-tive and induced mechanisms are utilized to permitsurvival despite greatly reduced protein synthesis.

Received 20 December, 1994; revised 9 February, 1995: accepted13 Fabruafy, 1995. 'For correspondence. E-mail, CLARKE©EWALD.MBI.UCLA.EDU; Tel. (310) 825 8754: Fax (310) 206 7286.

tO 1995 BlacKw^l Science Ltd

Introduction

Escherichia coli and other bacteria can survive — andeven grow — under an impressive range of environmentalconditions, including temperature and pH extremes, nutri-ent limitation, osmotic variation or the presence of noxiouschemicals (for a review see Watson. 1990), This ability isdependent upon the maintenance of cellular proteins inactive forms. However, enzymes and other proteins arenever completely stable even under optimai conditions,and they can become much less so when subjected toenvironmental stress. Therefore, prevention of damageto proteins and the repair or replacement of those whichdo become altered may be key features of both the bac-terium's response lo stressful conditions and its main-tenance programme under 'normal' conditions.

Protein damage can be of two general types: (i) confor-mationai damage, or alteration of the three-dimensionalstnjcture of the protein without changing its chemicalmakeup; and (ii) covalent damage, in which the primarystructure of the protein is modified by covaient bond break-age and formation (summarized in Fig. 1). E. coli has anumber of mechanisms for dealing with such changes,including systems which are induced or enhanced inresponse to environmental stimuli, as well as thosewhich are expressed constituttvely (or at a significantbasal level) during normal growth. The short generationtime and correspondingly rapid rate of protein synthesisin bacteria might suggest that damaged proteins —unlike DNA — could simply be degraded and replaced,and proteolysis is indeed one possible response to proteindamage. However, protein synthesis is severely limited instarved bacteria (Reeve etai, 1984), and is at least tran-siently depressed in response to other stresses (Neid-hardt and VanBogelen, 1987): under these conditions,much of the cells' synthetic capacity may be turnedtoward the expression of stress-specific proteins. Thus,protein repair systems may permit bacteria to surviveunder sub-optimal conditions (the rule rather than theexception in the natural environment) by making the bestpossible use of their existing proteins. Many of thesemechanisms are also present in eukaryotic cells, whichare similarly afflicted by spontaneous protein damage,especially in the ageing process.

836 J. E. Visick and S. Clarke

edt. pH changes, ge^ianol. salts,

coid, etc.CroEL DnaK. etc

O3

Functional Native Protein

IsomerizedProlines

OtherAmi no-acid

Modifications. Methionine Isoaspartyl

sulfoxtdes Formation

Covalent Damage

Fig. 1. Summary of mechanisms of proteindamage and repair. Proteins are subject toconformational damage (top), which can bealleviated by chaperones. as well as tocovalent damage (bottom). Prolineisomerizalion, methionine oxidation andisoaspartyl lormation are common lorms ofcovalent damage, repaired by peptidyl-prolylds-lrans isomerases (PPIases), methioninesulphoxide (Met[O]) reduclase andL-isoaspartyl protein carboxyl methyi-transferase (PCfVl), respectively; additionalrepair mechanisms may also exist. Degra-dation of damaged proteins occurs whenrepair or refolding are unsuccessful (right).

Chaperones can reverse conformational damage afterinitial foiding

The conspicuous effect of heat on protein conformation isthe basis for the hypothesis that the universally conservedheat-shock response serves primarily to prevent proteindenaturation and/or to allow renaturation (Pelham, 1986;1988). All the information required for correct folding of aprotein is believed to reside within its amino acid sequence(Anfinsen, 1973). For many proteins, however, folding isnot necessarily rapid, and partially folded or misfoldedpolypeptides can aggregate instead of proceeding to thenative state (Nilsson and Anderson, 1991; Seckler andJaenicke, 1992). The folding process can be facilitatedby 'molecular chaperones' (Ellis, 1987), many of whichare heat-shock proteins. Chaperones. typified by the bac-terial proteins DnaK and GroEL, can interact with a varietyof protein substrates (Bochkareva etai. 1988; Van Dyk etBi, 1989) and prevent aggregation or unfold misfoldedstnjctures, increasing the likelihood that the protein will

ultimately fold properly. The role of chaperones in thefolding of nascent and newly synthesized proteins is nowwell characterized and has been reviewed recently (seeBecker and Craig, 1994: Craig et ai, 1993; HartI et ai,1994).

Despite con'ect initial folding, there is always some ten-dency for proteins to subsequently unfold or aggregate. Itis now clear that GroEL and DnaK also help maintain thefunctional conformations of existing proteins. Their abilityto bind unfolded polypeptides and facilitate renaturationhas been demonstrated both in vitro and in vivo (Kawataet ai, 1994; Martin et ai. 1992; Skowyra et ai., 1990).Both of these chaperones are required for ttie properfunctioning of many cellular processes even at normalgrowth temperatures (Fayet et ai, 1989; Paek andWalker, 1987). Heat exacerbates the problems of unfold-ing and aggregation (Beckmann etai. 1992; Nguyen etai.. 1989), as do many other environmental conditions.Induction of GroEL and DnaK synthesis under thesecircumstances provides additional chaperone molecules

1995 Biackwell Science Ltd, Molecular Microbiology. 16. 835-845

Protein repair in Escherichia coli 837

to bind denatured proteins, prevent aggregation (whichmay be a major cause of death in heat-stressed cells)and allow refolding {Parsell etai, 1993).

Although E. coli synthesizes distinct sets of proteins tocounter specific environmental changes, increased pro-duction of chaperones is common to nearly all knownstress responses. Heat, ethanot. nutrient limitation, oxidiz-ing agents, high or low pH, nalidixic acid, u.v. irradiation,puromycin. phage infection and alkylating agents are allfactors that induce GroEL and DnaK synthesis (Neid-hardt and VanBogeien, 1987; Nystrom, 1993; Watson.1990). Similarly. \he quaternary, tertiary or secondarystructure of proteins can be perturbed by many factors,including heat, cold, salts, acid or alkaline pH. highosmotic pressure, alcohols and oxidizing agents, both invitro (Davies, 1987; Privalov. 1990; Tanford. 1968) andin vivo {Nguyen et ai, 1989). This striking parallel sug-gests that conformational damage to proteins may beone important consequence of many environmentalstresses, and the pre-eminence of chaperones in the Ecoli stress responses is consistent with a critical functionin the cell's reaction to such damage. Significantly,denatured proteins themselves stimulate a heat-shockresponse: GroEL and DnaK are induced in E. coli in pro-portion to the expression of a mutant \ repressor whichis unable to fold (Parseil and Sauer. 1989).

Conformationally damaged proteins may thus triggerthe synthesis of chaperones and other heat-shock pro-teins, which may then be able to alleviate the damage. Ifits primary structure is unaffected, a protein which becomesdenatured should be able to regain activity, provided itcan refold fo a native state before aggregating or beingdegraded. By extending the protein's opportunity torefold, GroEL and DnaK can, in effect, reverse conforma-tional damage. Chaperones can. therefore, be thought ofas 'repair' mechanisms for conformationally damagedproteins.

Covalent damage and mechanisms of repair

Just as DNA is subject to covalenf alteration (such asdepurination or deamination) through spontaneous chemi-cal reactions (Kushner. 1987), the amino acids which makeup proteins are also susceptible to chemical changes.Environmental factors can cause specific alterations inthe amino acid sequence of a protein or may acceleratespontaneous changes. This presents two serious prob-lems for the cell: how a damaged protein may be recog-nized — particularly difficult since a protein could beinactivated by a very subtle change in its active-site chem-istry — and how the damage might be reversed. Severalforms of such damage, as well as possibilities for repair.are considered here (see Fig. 1).

Repair of covalent damage affecting protein folding

Disulphide bond formation

The formation of disulphide bonds is an example of acovalent change which greatly affects three-dimensionalstructure: protein disulphide isomerases, which catalysethis process, are important members of the protein-foldingpathway (Freedman, 1989). In E. ooli. the peripiasmicproteins DsbA (Bardwell etal., 1991) and DsbC (Wulfingand Pluckthun, 1994) are active in the formation ofdisulphide bonds needed for activity of peripiasmic andouter-membrane proteins. An additional component ofthis pathway. DsbB, serves to reoxidize DsbA (Bardwellet ai, 1993). In addition to catalysing initial disulphidebond formation, the Dsb proteins can potentially reoxidizeexisting peripiasmic proteins in which disulphide bondshave become reduced, thus serving as a repair system(Bardwell and Beckwith. 1993; Wulfing and Pluckthun,1994).

In contrast to the extracellular environment, the E. colicytoplasm is reducing, and disulphide-containing proteinsare rare (Derman etai. 1993). Since exposure of cytoplas-mic proteins to oxidizing agents may result in the inap-propriate formation of disulphide linkages, maintenanceof functional cytoptasmic proteins would require reductionof disulphides. The cytoptasmic enzyme thioredoxin reduc-tase has been implicated in this form of 'repair' sincemutations in its gene permit the formation of disulphidebonds which othenwise would not occur in the cytoplasm(Derman etai, 1993).

Proline isomerization

Proline peptide bonds can exist in either the 'normal' trans-configuration or the c/s-form (Fischer et ai., 1983; seeFig. 2A). The presence of the inappropriate c/s-isomerscan greatly retard protein folding or refolding (Lang etai.1987) and can also inhibit proteolysis (Yaron and Naider.1993). Whether such rotation around the peptide bondreflects a covalent change or a conformational one canbe debated (see Stein. 1993); it will be treated here as acx)va!ent alteration to emphasize the partial double-bondcharacter of the peptide bond and the necessary disrup-tion of this bond in the isomerization event. This positionis reinforced by the suggestion that at least the PpiC-typeisomerases (see below) use a covalent intermediate incatalysis (Rudd etai, 1995).

Peptidyl-prolyl cis-trans isomerases. enzymes capableof catalysing proline isomerization. have been identifiedin many organisms and shown to increase the rate of pro-tein folding (Schmid et ai, 1993). E. coii has genes for atleast three such isomerases: ppiA and ppiB. whichencode peripiasmic and cytoplasmic enzymes, respec-tively (Hayano et ai, 1991), and ppiC, encoding a newly

I.C. 1995 BiackwBll Science Ltd. Molecular Microbiology. 16. 835-845

838 J. E. Visick and S. Ciarke

trans

^ L-aspartylpresidue

L-succinimiJe 0intermediate

BL-a^araginyl

residue

L-succinimideintermediate ,.

L-aspartyl residue

L-isoaspartyl _residue V

L-aspartyl nresidue T

A ''SAM

PCMMethyKransferase ,^ .

L-isoaspartylmethyl ester

.N^ ..-

L'Succinimide o

Fig. 2. Mechanisms of covalent damage and repair.A. Isomerization of prolyl residues trom Ihe norma! /rans-configuration to the c/s-configuration (adapted from Fischer etal., 1983).B. Deamidation of an asparaginyl residue by intramolecular attack and formation of a succinimide intermediaie. followed by hydrolysis to yieldetther L-aspartate or i-isoaspartateC. Isomerization of aspartyl residues via a succinimide intermediate generated by intramolecular attack paralleling that in (B); hydrolysis of theintermediate leads either to regeneration of the L-aspartate or to an L-rsoaspartyl residue.D. Methylation of L-isoaspartyl residues by PCM. facilitating reformation of the succinimide intermediate; hydrolysis produces either L-aspartateor L-isoaspartate. The merhyl donor is S-adenosylmethionine (SAM), converted to S-adenosylhomocysteine (SAH) in calatysis. Throughout thefigure, the peptide 'backtwne' is shown as a heavy line.

discovered lOkDa isomerase of undetermined cellularlocalization (Rahfeid et ai, 1994a: 1994b: Rudd ef ai,1995). These enzymes probably expedite the initiai fold-ing of nascent proteins but can also function in protein'repair" by the reconversion of c/s-linkages which occurspontaneously after synthesis in either folded or unfoldedproteins. The ability to accelerate refolding of heat-denatured proteins in vitro has been direcfly demonstratedfor both PpiA and PpiB (Compton etai. 1992; Knappik etai. 1993). Uniike at least one eukaryotic isomerase. theseE. coii enzymes are not heat induced; however, they dointeract with the heat-shock protein HtpG. implicatingthem in the cellular response to stress (Nadeau et ai,1993).

The close relationship of peptidyl-prolyl isomerase activ-ity to protein folding places this enzyme near the borderbetween conformationai and covalent repair mechan-isms. It has been suggested that these enzymes maythemselves be chaperones (Freskgard et ai, 1992). butrecent results have shown that their role in acceleratingrepair of conformational damage is solely through theirability to reverse isomerization of prolines (Kern et ai,1994). However, other proteins involved in this processmay have novel functions. The SurA protein, required forlong-term stationary-phase survival (Tormo etai, 1990).shows significant amino acid sequence similarity to PpiC(Rahfeid ef ai, 1994a). SurA contains two copies of thepresumed isomerase domain which makes up nearly all

!£: 1995 Biackwell Science Ltd. Moiecuiar Microbiology, 16, 835-845

Protein repair in Escherichia coli 839

of the small PpiC protein (Rudd et ai, 1995). Similarsequences are also found within two putative chape-rones: PrsA, involved in secretion and maturation ofextracytoplasmic proteins in Bacillus subtilis (Jacobs etai, 1993). and NifM. a protein required for assembly ofnitrogenase in Klebsieila pneumoniae (Harris et ai.1990). Based on these similarities. SurA is likely to be apeptidyl-prolyl cis~trans isomerase and may be specific-ally required for the maintenance and repair of proteinsduring nutrient limitation; it could also act as a chape-rone. Recent work indicates that SurA is localized in theperiplasm and that it may affect the level of outer-membrane proteins (Rudd etai, 1995).

Covalent damage by molecular oxygen: many formsof damage, one known repair pathway

Oxidative stress, in addition to damaging DNA (Farr andKogoma, 1991). can alter the primary structure of pro-teins. Oxidizing agents such as H2O2 may be presentin the environment, or hydroxyl (OH*), hydroperoxyl(HOO*), superoxide (O2 ) or lipid peroxyl radicals may begenerated by ionizing radiation or as side products ofmetabolism (Farr and Kogoma. 1991). These highly reac-tive oxygen species can attack virtually any amino acid,particularly methionine. cysteine. histidine. tyrosine. fryp-topfian and phenylalanine; metal-binding proteins andthose with iron-sulphur clusters are particularly suscep-tible (Stadtman. 1992). Commonly observed modifica-tions include oxidation of methionine to form methioninesulphoxide (Farr and Kogoma. 1991). formation of bityro-sine (Davies et ai, 1987). conversion of histidine toasparagine or aspartate and of proline to glutamylderivatives and even fragmentation of the protein (Stadt-man, 1992). Such alterations often eliminate or reducethe activity of the affected enzymes (Farr and Kogoma,1991).

E. coli has only one known repair mechanism specificfor oxidative damage; the enzyme methionine sulphoxidereductase recognizes methionine sulphoxide as an abnor-mal residue in proteins and can reduce it back to methio-nine. using reduced thioredoxin as a substrate (Lunn andPigiet. 1987; Rahman et ai. 1992). Oxidative damage afother sites, however, is likely to be difficult to recognizeand repair, because one amino acid may be convertedinto another ordinary residue, and, unlike DNA repair,there is no nearby 'template' to preserve the correctsequence. Furthermore, covalent damage is often theunderlying cause of protein denaturation observed afteroxidative attack (Davies and Delsignore. 1987), so evenchaperone-assisted refolding may fail to restore activity.These difficulties may explain why preventive mechan-isms, such as the production of catalase and superoxidedismutase. dominate the response of E. coli to oxidative

IQ 1995 Btackwell Science Ltd. Molecular MtnkMlogy. 16. 835-845

stress (Farr and Kogoma. 1991). Possibly, however, addi-tional repair mechanisms remain to be discovered.

Covalent modification of asparaginyl and aspartylresidues and repair pathways

Asparaginyl and glutaminyi residues deamidate spon-taneously to form aspartate and glutamate. respectively,a change in both chemical structure and charge that canaffect protein stability and activity (Stadtman, 1988). Thereis no known means of distinguishing a glutamate or aspar-tate residue which arose by deamidation from one which isencoded in the genome, or of reversing the change. In atleast the case of asparagine. however, deamidation isnot direct, and the intramolecular reaction by which itoccurs also results in isomerization and racemization.the products of which serve as a telltale trace and makeenzymatic recognition and restoration of normal L-aspartylresidues possible.

Direct hydrolysis of the amide linkage is too slow toaccount for the observed rate of asparagine deamida-tion. Instead, evidence favours the formation of a succin-imide intermediate by an intramolecular attack of thepeptide-bond nitrogen on the side-chain carbonyl carbon,releasing the amine as ammonia in the process (Fig. 2B;reviewed by Clarke et ai. 1992). This unstable intermedi-ate can then be hydrolysed on either side of the nitrogen.In one case, an L-aspartyl residue will result, but in theother, the residue will be isomerized to an L-isoaspartylresidue; the peptide bond is shifted to what was previouslythe side-chain carbonyl carbon, placing a 'kink' in the pep-tide backbone (Geiger and Clarke. 1987; see Fig. 2B).In addition, the succinimide ring is prone to racemiz-ation between the L- and D-forms; if formation of the D-succinimide precedes hydrolysis, then a D-asparfyi orD-isoaspartyl residue will be generated, To complicatematters further, a succinimide can also be formed by aparallel attack upon fhe side-chain carboxyl group of anL-aspartyl residue (Fig. 2C). and hydrolysis again pro-duces either aspartyl or isoaspartyl forms (Clarke et ai,1992). in summary, both asparaginyl and aspartyl resi-dues in proteins are susceptible to covalent alteration bythe spontaneous formation of a succinimide derivative,the ultimate products of either reaction being a D- or L-aspartyl or D- or L-isoaspartyl residue.

Both the change in charge (from neutral asparagine toacidic aspartate or isoaspartate) and the change in chemi-cal structure (from asparagine or aspartate to an iso-merized or racemized form) produced by these reactionscan be detrimental to the function of a protein. Clearly,alteration of a single amino acid at the active site candisrupt enzymatic activity; the damage might also occurat a residue critical for maintenance of protein structure.The formation of isoaspartate, by creating a 'kink' in the

840 J. E. Visick and S. Ciarke

poiypeptide. could affect protein conformation even if it isnot in a region ordinarily crucial for structure. Lysozyme.calmodulin. calbindin, human epidermal growth factorand growth hormone releasing factor all show reducedactivity correlating with the presence of isomerized aspar-tate residues (Clarke. 1992; Clarke etai, 1992), and theisoaspartyl peptide linkage also inhibits proteolysis ofaffected peptides (fvlurray and Clarke, 1984).

The presence of isoaspartate in a protein signals thatisomerization has occurred and allows recognition by anenzyme capable cf reversing — or at least mitigating —the damage. The E. coli L-isoaspartyl protein carboxylmethyItransferase (PCM) specifically binds t-isoaspartylresidues in proteins (Fu et ai, 1991) and methylates theside-chain carboxy! group using S-adenosylmethionineas a methyl donor (Fig. 2D). The reactive methyl esterthus formed stimulates spontaneous reformation of thesuccinimide intermediate. A significant fraction of themolecules resulting from hydrolysis of the succinimidewill be L-aspartyl residues, and fhe continued action offhe methyItransferase, therefore, ultimately reconverts L-isoaspartyl residues to normal L-aspartyl residues (Bren-nan etai. 1994; Johnson etai, 1987a,b; McFadden andClarke. 1987).

In addition to E. coli, L-isoaspartyl methyltransferaseactivity has been found in several other bacterial groups(Li and Clarke. 1992a). plants (Mudgett and Clarke,1993) and a variety of invertebrates and vertebrates(Barten and O'Dea, 1990; Clarke, 1985; Haklai andKloog. 1987; O'Connor. 1987). Although the genes arewell conserved from bacteria to humans (Fu et ai, 1991;Mudgett and Clarke. 1993). the mammalian methyltrans-ferases have the additional ability to recognize o-aspartylresidues, allowing them to repair some racemized resi-dues (Lowenson and Clarke, 1992). Possibly, the racemi-zation reaction, which is much slower than isomerization(Geiger and Clarke, 1987). may be less problematic forbacteria, where the lifespan of an individual protein isshorter than in a mammalian cell. In any case, no knownmethyltransferase can recognize D-isoaspartyl residues(McFadden and Clarke, 1987), and L-aspartyl residuesresulting from asparagine deamidation are similarly unde-tectable. The methyltransferase may repair those forms ofdamage which are least well tolerated by the cell, cr it mayeven function mainly to return proteins to forms which canbe degraded (Clarke. 1992).

The pcm gene, encoding the £. coli methyltransferase,has been cloned (Fu ef a/.. 1991). and pcm deletionmutants have been characterized (Li and Clarke. 1992b).The mutation has no apparent effect during exponentialgrowth, but survival is significantly impaired when mutantsare exposed to various stresses. After 10 days in station-ary phase, at least 10% of wild-type cells remain viable,but only 0.1 to 1% of pcm mutants are still capable of

forming colonies (Li and Clarke, 1992b). The percentageof cells surviving exposure to lethal (55 C) temperatures(Li and Clarke, 1992b). high salt concentrations. (Li etai,1994). H2O2. rapid chilling and ethanol (J. Visick and S.Clarke, unpublished results) is also much lower than forthe wild type. The effect of L-isoaspartyl residues, there-fore, seems most significant when E. coli is forced torespond to stress; in fact, conditions which can result inprotein denaturation would appear to correspond withreduced viability of pcm mutants, it is, therefore, temptingto speculate that the immediate effect of isomerization onprotein function may often be minor, but that once analtered protein becomes unfolded, the isoasparty! 'kink'may produce significant consequences by delaying or inhi-biting refolding (see Fig. 3). Isoaspartyl residues mightalso make the protein particularly vulnerable to denatura-tion. Accumulation of damaged proteins could compro-mise the cells' ability to recover from stress by reducingthe number of functional proteins present at this criticaltime. Iscaspartyl-containing peptides which cannot berefolded could also 'overload' the chaperone system,since folding-deficient peptides are thought to formunusually stable complexes with chaperones (Welch,1993).

isoAsp

Fig. 3. Possible explanation lor reduced viability of pcm mutantsafter exposure to stress. In the presence of PCM, aspartyt andasparaginyl residues are maintained, and proteins can be refoldedin case of conformational damage caused by heat or other stresses(left). When the methyltransferase is not present, L-isoaspartylresidues may Increase unfolding of proteins upon exposure todenaturing stresses or may interfere with refolding (righf),

1995 Blackwell Science Ltd. Molecu/ar Microbiology. 16. 83S-B45

Protein repair in Escherichia coli 841

Degradation of damaged proteins: the cell's 'lastresort'

Despite these repair systems, not every damaged proteinwill be recognized or properly restored to activity. Indeed,abnormal proteins, such as those produced in the pre-sence of amino acid analogues, are subject to rapiddegradation in E co//(Parsell and Lindquist, 1993). Con-sidering the rapid rate of growth and protein synthesis,could degradation, and not repair, be the primaryresponse of E. coli to damaged proteins? The existenceof many complex, specific mechanisms for the reactiva-tion of conformationally or covalently damaged proteinsargues against this idea, as does the requirement thatstarved or stressed cells be able to survive unfavourablecircumstances without extensive de novo protein syn-thesis. It seems likely that returning damaged proteins toa functional condition would be preferable to degradingthem. Proteolysis. however, remains an important alter-native for dealing with damage that cannot be resolvedby the repair machinery.

At least three proteolyfic systems are induced inresponse to heat and other stresses in E. coli. implicatingthe degradatlve machinery in the cellular response toadverse conditions, (i) The Lon protease is a member ofthe heat-shock regulon and is transcribed by RNA poly-merase containing the sigma factor a^^ (Neidhardt andVanBogelen. 1987). The involvement of Lon in degrada-tion cf damaged proteins is well known: its substratesinclude a variety of mutant peptides. puromycyl fragmentsand proteins made in the presence of amino acid ana-logues (Gottesman. 1989). (ii) DegP is presumed to fill asimilar role in fhe periplasm (Lipinska et ai, 1990). whereit degrades truncated proteins as well as slow-foldingmutant proteins (Wulfing and Pluckthun. 1994). Althoughit is heat inducible and essential at high temperatures.DegP is controlled not by a^ but by o^ (Erickson andGross. 1989; Lipinska etai, 1990; Strauch etai, 1989),which in turn may respond to stress-induced changes inthe outer membrane (Mescas etai. 1993). (iii) Three Clpproteins involved in proteolysis, CipB. CIpP and CIpX.have been identified as members of the n^^ regulon (Got-tesman etai, 1993; Kroh and Simon. 1990; Squires etai,1991).

The Clp proteins demonstrate a close relationshipbetween chaperone activity and proteolysis. CIpP. theproteolytic subunii, degrades short peptides but mustform a multimeric complex with an ATPase subunif inorder to break down larger proteins (Thompson et ai.1994). Such a complex can be formed with either ClpAor CtpX as the ATPase, and the two resulting enzymeshave different specificities (Gottesman ef ai, 1993);the existence of CIpB-CIpP complexes is uncertain buthas been suggested as a possible means of degrading

heat-denatured proteins (Kitagawa et ai, 1991). In theabsence of CIpP, it is thought that the ATPase subunitsmay act as chaperones (Squires and Squires. 1992), anda DnaK-like activity has now been direcfly demonstratedfor ClpA (Wickner etai. 1994). Binding of a damaged pro-tein to ClpA may afford it the opportunity to refold; if itcannot rapidly return to fhe native state, it may remainbound to ClpA and thus be targeted for degradationwhen CIpP joins the complex. Such a mechanism wouldallow for repair of existing proteins while preventing thechaperone system from becoming congested with irrepar-ably damaged polypeptides. Interestingly, the CIpA-CIpPprotease has recently been connected with the specificdegradation of proteins induced at the onset of stationaryphase (Damerau and St John. 1993).

Other chaperones seem to function in proteolysis aswell. Mutations in groEL. dnaK. grpE. or dnaJ impair pro-tein degradation (Keller and Simon, 1988; Straus ef ai.1988), even though these proteins have no known pro-teolytic activity. These chaperones may help to decidebetween the alternative fates of refolding and degradationfor a damaged protein. Two hypotheses have beenadvanced (Parsell and Lindquist, 1993) to explain thisrole: the chaperones may be members of a proteasecomplex, as suggested by the co-immunoprecipitation ofDnaK and GrpE with Lon and a mutant alkaline phos-phatase protein (Sherman and Goldberg. 1992); or theymay maintain abnormal proteins in a disaggregated,protease-sensitive conformation, supported by the abilityof GroEL to stabilize proteins in conformations which aresusceptible to cleavage (Kandror et ai, 1994; Martin efai. 1991).

Some E. coii proteotytic systems seem adapted tothe remediation of particular forms of covalent proteindamage. Dipeptides of the form L-isoaspartate-X arespecifically recognized by the E co//L-isoaspartyl dipepti-dase, the product of the /ad gene (Gary and Clarke. 1995;Haley, 1968). This appears to be a second mechanism fordealing with the isoaspartyl residues discussed above.Since the altered peptide bond at an isoaspartyl residueblocks proteolysis (Murray and Clarke, 1984). the iso-aspartyl dipeptidase may be needed to permit the com-plete breakdown of isoaspartyl-containing peptides whichevade repair by the isoaspartyl methyltransferase. AnATP-independent protease which preferentially attacksoxidatively damaged proteins may account for increasedproteolytic activity in cells exposed to oxidative stress(Davies and Lin. 1988; Roseman and Levine, 1987) andmay be triggered by denaturation resulting from covalentprotein damage (Davies and Delsignore. 1987). Increasedproteolysis has also been observed upon osmotic upshift(Meury, 1994); a protease which responds specifically todamage resulting from osmotic shock may, therefore,remain to be discovered.

1995 Blackwell Science Ltd, Midacular MIcrotjiology. 16. 835-845

842 J. E. Visick and S. Ciarke

Stationary phase as a stress response: the need forrepair systems in ageing bacteria

Upon exhaustion of a nutrient. E coli and other non-sporulating bacteria enter stationary phase, a non-dividingstate in which they can remain viable for extended periods(Siegele and Kotter, 1992). Defences against adverseenvironmental conditions appear to be important tc thesenutrient-limited cells, since they gain added resistance toheat. salt, peroxide and other stresses as stationaryphase begins (Jenkins etai. 1988; 1990). Characteriza-tion of the genes required for short- and long-term survivalin stationary phase is ongoing, but results to date clearlysuggest that the protection, maintenance and repair ofproteins, as well as other cellular components, are criticalelements of the response of E. coliXo starvation.

More than 20 genes are induced as cells make thetransition from exponential growth to stationary phase(Nystrom. 1993). Many of these genes are transcribed byRNA polymerase containing the alternative sigma factora^. the product of the rpoS or katF gene (Hengge-Aronis. 1993a.b; Nguyen et ai, 1993). Among the CT®-dependent genes whose functions are known, severalhave products which seem to be involved in protectionagainst stress: otsA. otsB and treA are involved in thesynthesis and transport of trehalose, a thermoprotectantwhich stabilizes both membranes and proteins (Hengge-Aronisefa/., 1991; Strom and Kaasen, 1993); katEencodesHPII catalase. providing protection against peroxide (Saket ai.. 1989); Dps is a histone-like protein involved in pro-tecting DNA against oxidative stress (Hengge-Aronis.1993b); XthA is a DNA-repair exonuclease (Sak ef ai,1989); and CbpA has sequence similarity to DnaJ andmay be a stationary-phase-specific chaperone (Yama-shino ef ai. 1994). Genes which are not under the controlof n® are also turned on at this time: the heat-shock pro-teins GroEL, DnaK and HtpG are prominently expressedin stationary phase (Nystrom, 1993). and DnaK is requiredfor starvation survival (Spence et ai, 1990). The heat-shock Sigma factor n^ appears to be responsible for thisinduction (Jenkins et ai. 1991). Finally, starvation survivalalso requires repair enzymes which are constitutivelysynthesized. As discussed above, deletion of pcm (encod-ing the methyltransferase which allows reconversion ofisoaspartyl to aspartyl residues) significantly reduceslong-term viability (Li and Clarke. 1992b). The surA gene,whose product is probably a peptidyl-prolyl cis-transisomerase. is also required for stationary-phase survival(Tormo etai. 1990).

Heat. salt, peroxide and other environmental stressescan have a devastating effect on proteins; repair mechan-isms are understandably important in remedying theresulting damage. But why are protein repair and stressprotection also crucial to £. coli's response to nutrient

limitation? Starving cells are unable to rapidly synthesizeproteins in response to whatever changes might ariseprior to the return of available nutrients; hence, it hasbeen suggested that the ceils must be prepared' forfuture stress (Siegele and Kolter. 1992). Another con-sideration is the reduction in synthesis of alt proteins —not merely stress proteins — during starvation: station-ary-phase ceils must rely heavily upon existing proteinsto carry out cellular functions. Bacteria in fhe naturalenvironment seldom enjoy the luxury of exponentialgrowth and are generally in stationary phase, or at bestgrowing slowly (Roszak and Colwell, 1987). Under theseconditions, protein turnover is slow, so even if the celtsencounter no specific form of stress, proteins will haveample opportunity to become denatured or otherwisedamaged.

Stationary phase in bacteria is the equivalent of ageingin eukaryotic systems, where Ihe occunence of both con-formational and covalent damage in ageing proteins iswell documented (Rothstein. 1984; Stadtman. 1988; Stadt-man ef ai. 1992). It is likely that such damage is alsoimportant in ageing' prokatyotes and that protein mainte-nance and repair are. therefore, of particular importanceduring stationary phase, tn support of this idea, suscep-tibility to oxidative damage has been correlated with lossof long-term viability in stan/ed £. coli (Eisenslark ef ai.1992). Interestingly, stationary-phase cells do not increaseproteolysis in response to heat shock as growing cells do(Soumillion and Fastrez. 1992), underscoring the impor-tance of preserving existing proteins, apparently even atthe risk of failing to degrade damaged ones.

The rpoS gene, encoding n^. is located at 59 min on theE. co//chromosome (Mulvey and Loewen. 1989). and thepcm gene is located just i300bp upstream and is tran-scribed in the same direction (Fu et ai, 1991; seeFig. 4). Two additional genes which affect sfationary-phase survival have now been found in this region: surE.the upstream gene of the surE-pcm operon (Li et at..1994), and nipD. located between pcm and katF (tchi-kawa ef ai. 1994). The function of the SurE protein isunknown, much like pcm cells; however, surE mutants

1000 bp

Fig. 4. Genes in the 59 min recpon of the E. co/t chromosome. Twooperons are present in this region, the PCM gane {pcm) and surE,encoding a protein required for stress and starvation survival, areco-transcribed; rpoS (kalF). encoding o^. and n/pD, encoding alipoprotein. are transcribed from a second promoter Majorpromoters are indicated by black arrows (Lange and Hengge-Aronis, 1994a,b; Takayanagi ef al.. 1994).

1995 Blackwell Sdence Ud. Moiecuiar Mcmblotogy. 16, B35-845

Protein repair in Escherichia coli 843

show diminished viability upon long-term starvation,lethal heat shock (55 C). osmotic shock (Li etai. 1994) oroxidative stress (J. Visick and S. Clarke, unpublisheddata). Interruption of the nIpD gene produces a weakerstationary-phase survival phenotype; this gene encodesa lipoprotein similar to ippB, an outer-membrane virulencefactor of Haemophiius somnus (Ichikawa et ai. 1994)which may affect cell-wall synthesis (Lange and Hengge-Aronis. 1994b). Transcripts from the nlpD promoterinclude rpoS. and additional promoters located withinntpD appear to be important for stationary-phase induc-tion of rpoS (Lange and Hengge-Aronis. 1994a; Takaya-nagi et ai, 1994). These four cistrons. ostensibly a genecluster required tor survival in starvation and otherstresses, may have descended from a single ancestraloperon which appears to be well conserved. Isoaspartylmethyltransferase activity has been found throughout thea- and y-purple groups of Gram-negative bacteria (Li andClarke. 1992a). and apparent homoiogues of two ormore of the four genes have been found together in H.somnus. Pseudomonas aeruginosa (Ichikawa et ai,1994) and Salmonella typhimurium (Prince et ai. 1994).It wiil. therefore, be interesting to determine whetherSurE and NlpD function in protein maintenance and repair.

Conclusion

It is tempting to conclude from the short generation time ofexponentially growing bacteria and their rapid rate of pro-tein synthesis that prokaryotic proteins are transitory andreadily replaced. However, the lifetime of most of theseproteins is much longer than the cells' optimal generationtime (Parsell and Lindquist. 1993). and environmentalstresses can result in greatly reduced synthesis of mostproteins (Watson. 1990). Furthermore, nutrients are veryoften limiting for bacteria in nature, so the rate of proteinsynthesis is not unlimited and the value of an individualprotein may be considerable. Protein damage — eitherconformational change or covalent alteration of the aminoacid sequence — can. therefore, be a significant problemfor the cell, whether it results from environmental condi-tions, such as heat or reactive oxygen species, or occursspontaneously as proteins in slow-growing cells 'age'.Although proteolytic systems can target altered polypep-tides for degradation, it may be preferable to regain theuse of existing proteins whenever possible. In fact, evi-dence suggests that E. coii and other bacteria havesophisticated repair systems allowing the reversal ofspecific forms of protein damage.

While DNA repair is a well-established phenomenon,protein repair is a relatively new idea. Chaperones havebeen shown to assist in the refolding of conformationallydamaged proteins as well as in the initial folding path-way. Enzymes which recognize particular types of

C' 1995 Blackwell Science Ltd. Molecular Microbiology. 16. 835-845

covalent damage have now been identified as well; theirprecise roles remain to be fully elucidated, but they sug-gest the tantalizing possibility of repair mechanisms forother problematic changes in the chemical makeup ofproteins.

One might ask. however, how prevalent covalent proteinrepair systems are likely to be. Considering the severalforms of repairable covalent alteration discussed above,it seems that their utility will probably not be unlimited. Allof these types of damage are common and occur inmany different proteins, and damaged proteins are ineach case relatively easily detected and can be repairedby a fairly straightforward mechanism. These characteris-tics are probably hallmarks of covalent protein repair: lack-ing any 'template* to show the correct primary structure ofproteins, the likelihood that random breakage of peptidebonds, interconversion of amino acids by oxidizing agentsor other variable or indistinguishable changes can berecognized and repaired is small. Nevertheless, additionaldiscoveries in this area can certainly be expected as ourcomprehension of both the structure of normal proteinsand the changes which can occur in that structureincreases.

!

References

Anfinsen. C.B. (1973) Scrence 181: 223-230.BarcJwell, J.C.A.. and Beckwith. J. (1993) Ce//74: 769-771.Bardwelt, J.C.A.. McGovern. K., and Beckwith. J. (1991) Cell

67: 581 -589.Bardwell, J.C.A.. Lee. J.-O.. Jander. G.. Martin, N., Beiln, D..

and Beckwith. J. (1993) Proc NatI Acad Sci USA 9Q: 1038-1042.

Barten, D.M.. and O'Dea. R.F. (1990) Life Sci AT. 181-194.Becker, J., and Craig. N. (1994) Eur J Biochem 2ABi 11-23,Beckmann. R.P., Lovett. M.. and Welch. W.J. (1992) J Cell

8(0/117: 1137-1150.Bochkareva. E.S.. Lissin. N.M.. and Girshovich, A.S. (1988)

Nature 336: 254-257.Brennan. T.V., Anderson, J.W.. Jia. Z.. Waygood. E.B.. and

Clarke. S. (1994) J Biol Chem 269: 24586-24595.Clarke, S. (1985) Annu Rev Biochem 54: 479-506.Clarke, S. (1992) Fund Med Cell Biol 3B:413-436.Clarke. S.. Stephenson, R.C., and Lowenson. J.D. (1992) In

Stabiiity of Protein Pharmaceuticals. Part A: Chemicai andPhysicai Pathways of Protein Degradation. Ahern, T.J., andManning, M.C. (eds). New York: Plenum Press, pp. 1-29.

Compton. L.A., Davis, J.M.. MacDonald. J.R.. and Bachinger,H.P. (1992) Eur J Biochem 206: 927-934.

Craig. E.A., Gambill, B.D.. and Nelson, R.J. (1993) Microbiolflei/57: 402-414.

Damerau, K.. and St. John. A.C. (1993) JBacfeno/175: 53-63.

Davtes. K.J.A. (1987) JS/o/C/iem 262: 9895-9901.Davies, K.J.A.. and Delsignore. M.E. (1987) J S/o/Chem 262:

9908-9913.Davies, K.J.A.. and Un. S.W. (1988) Free Radical Biol Med5:

225-236.

844 J. E. Visick and S. Clarke

Davies. K.J.A.. Delsignore. M.E.. and Un. S.W. (1987) J BioiChem 262: 9902-9907.

Derman. A.I., Prinz. W.A.. Beiin. D.. and Beckwith, J. (1993)Sc/ence 262: 1744-1747.

Eisenstark. A.. Miller, C, Jones. J.. and Leven, S. (1992)Biochem Biophys Res Commun 188: 1054-1059.

Ellis. J. (1987) Nature 32B: 378-379.Erickson. J.W.. and Gross, CA. (1989) Genes Dev3: 1462-

1471.Farr. S.B.. and Kogoma. T. (1991) Microbiol Rev 55: 561-

585.Fayet. O.. Ziegelhoffer. T.. and Georgopoulos. C. (1989) J

Bacteriol 171:1379-1385.Fischer, G.. Heins. J.. and Barth. A. (1983) Biochim Biophys

Acta 742: 452-462.Freedman. R.B. (1989) Ce//57: 1069-1072.Freskgard, P C , Bergenhem, N.. Jonsson. B.H., Svensson.

M., and Carlsson, U. (1992) Science 258: 466-468.Fu, J.C.. Ding. L., and Clarke. S. (1991) J Bioi Chem 266:

14562-14572.Gary. J.D.. and Clarke. S. (1995) J Bioi Chem 270: 4076-

4087,Geiger, T.. and Clarke. S. (1987) JB/o/C/iem 262: 785-794.Gottesman, S. (1989) Annu Rev Genel 23: 163-198.Gottesman. S.. Clark, W.P.. de Crecy-Lagard. V., and

Maurizi. M.R. (1993) J Biol Chem 268: 22618-22626.Haklai, R., and Kloog. Y. (1987) Biochemistry 26: 4200-

4206.Haley. E.E. (1968) J Biol Chem 243: 5748-5752.Harris, G.S., White. T.C., Flory, J.E.. and Orme-Johnson,

W.H. (1990) Jfi/o/C/iem 265: 15909-15919.HartI. F.-U.. Hlodan. R.. and Langer. T. (1994) Trends

Biochem Sci 19: 20-25.Hayano. T.. Takahashi, N., Kato, S., Maki. N.. and Suzuki, M.

(1991) Biochem30: 3041-3048.Hengge-Aronis, R. (1993a) In Starvation in Bacteria. Kjelle-

berg. S. (ed.). New York: Plenum Press, pp. 171-201.Hengge-Aronis. R. (1993b) Ce//72: 165-168.Hengge-Aronis. R.. Klein. W., Lange. R., Rimmele. M.. and

Boos. W. (1991) J Bacteriol 173: 7918-7924.Ichikawa. J.K.. Li, C, Fu, J.. and Clarke. S. {1994) J Bacteriol

176: 1630-1638.Jacobs. M.. Andersen. J.B., Kontinen. V., and Sarvas. M.

(1993) Mol MicrobiolS: 957-965.Jenkins. D.E., Schultz. J.E.. and Matin. A. (1988) J Bacteriol

170:3910-3914.Jenkins. D.E., Chaisson, S.A.. and Matin. A. (1990) J

Bacteriol 172: 2779-2781.Jenkins. D.E., Auger. E.A.. and Matin. A. (1991) J Bacteriol

173: 1992-1996.Johnson. B.A.. Langmack. E.L.. and Aswad. D.W. (1987a) J

BiolChem262: 12283-12287.Johnson, B.A.. Murray. E.D., Jr, Clarke. S.. Glass. D.B.. and

Aswad, D.W. (1987b) J Biol Chem 262: 5622-5629.Kandror, O., Busconi, L.. Sherman. M.. and Goldberg. A.L.

(1994) J Biol Chem 269: 23575-23582.Kawata. Y., Nosaka, K.. Hongo, K., Mizobata. T., and Nagai.

J. (1994) FEBS Leff 345: 229-232.Keller, J.A.. and Simon. L.D. (1988) Mol Microbioi2: 31-41.Kern. G.. Kern. D., Schmid. F.X,. and Fischer. G. (1994)

FEBS Lett 346: 145-148.

Kitagawa. M., Wada, C, Yoshioka. S.. and Yura. T. (1991) JBacterioi A73: 4247-4253.

Knappik. A.. Krebber, C, and Pluckthun. A. (1993) Bio/Technoiogy 11: 77-83.

Kroh. H.E.. and Simon. L.D. (1990) J Bacterioi M2: 6026-6034.

Kushner. S.R. (1987) In Escherichia coli and Salmonellatyphimurium Celluiar and Molecular Biology. Neidhardt,F.C. (ed.). Washington. D.C: American Society for Micro-biology, pp. 1044-1053.

Lang. K.. Schmid. F.X.. and Fischer. G. (1987) Nature 329:268-270.

Lange, R.. and Hengge-Aronis. R. (1994a) Genes Dev 8:1600-1612.

Lange. R., and Hengge-Aronis. R. (1994b) MolMicrobion3:733-743.

Li. C. and Clarke. S. (1992a) J Bacteriol 147: 355-361,Li. C and Clarke. S. (1992b) Proc NatI Acad Sci USA 89:

9885-9889.Li, C, tchikawa. J.K.. Ravetto. J.J., Kuo. H.-C, Fu. J.C. and

Clarke, S. (1994) J0acter/o/176: 6015-6022.Lipinska. B., Zylicz, M., and Georgopoulos, C. (1990) J

Bacterion72: 1791-1797.Lowenson, J.D., and ClarKe. S. (1992) J Biol Chem 267:

5985-5595.Lunn, CA.. and Pigiet. V.P. (1987) Int J Radiat Biol 5^: 29-

38.Martin, J.. Langer. T.. Boteva. R.. Schramel, A.. Honfl/ich,

A.L. and HartI. F.U. (1991) Nature3S2: 36-42.Martin. J.. Horwich, A.L.. and HartI, F.U. (1992) Sc/ence 258:

995-999.McFadden. P.N., and Clarke. S. (1987) Proc NatI Acad Sci

USA 84: 2595-2599.Mescas. J.. Rouviere, P.E., Erickson. J.W., Donohue, T.J..

and Gross. CA. (1993) Genes Dev 7: 2618-2628.Meury. J. (1994) FEMS Microbiol Lett 116: 287-292.Mudgett. M.B.. and Clarke. S. (1993) Biochem 32: 11100-

11111.Mulvey. M.R.. and Loewen. P.C. (1989) NucI Adds Res 17:

9979-9991.Murray. E.D.. Jr. and Clarite. S. (1984) J Biol Chem 259:

10722-10732.Nadeau, K.. Das. A., and Walsh. C.T. (1993) J Biol Chem

268: 1479-1487.Neidhardt. F.C, and VanBogelen, R.A. (1987) In Escherichia

coli and Salmonella typhimurium Cellular and MolecularBioiogy. Neidhardt. F.C (ed.). Washington. D.C: AmericanSociety for Microbiology, pp. 1334-1345.

Nguyen, L.H.. Jensen, D.B.. Thompson, N.E.. Gentry, D.R..and Burgess. R.R. (1993) Biochemistry 32: 11112-11117,

Nguyen, VT.. Morange. M.. and Bensaude. O. (1989) J BiolChem2BA: 10487-10492.

Nitsson, B., and Anderson. S. (1991) Annu Rev Microbioi A5:607-635.

Nystrom, T. (1993) In Starvation in Bacteria. Kjelleberg. S.(ed.). New York: Plenum Press, pp. 129-150.

O Connor. CM. (1987) J S(0/Crtem 262: 10938-10403.Paek, K.-H.. and Walker. G.C (1987) J Sacter/o/169: 283-

290.Parsell, D.A.. and Lindquist. S. (1993) Annu Rev Genet 27:

437-496.

1995 Bifickwell Science Ltd, Afotecu/ar Mlcrobloktgy. 16. S35-845

Protein repair in Escherichia coli 845

Parsell. D.A.. and Sauer. R.T. (1989) Genes Dev 3: 1226-1232.

Parsell. D.A.. Taulien. J., and Lindquist. S. (1993) PhilosTrans R Soc Lond 339: 279-286.

Pelham, H.R.B. (1986) Ce//46: 959-961.Pelham. H. (1988) Wafure 332: 776-777.Prince, R.W.. Xu, Y.. Libby, S,J.. and Fang. F.C (1994)

Biochim Biophys Acta 1219: 198-200.Privalov. P.L. (1990) Crit Rev Biochem 25: 281-305.Rahfeid. J.-U., Rucknagel, K.P.. Schelbert, B.. Ludwig. B.,

Hacker. J.. Mann. K., and Fischer. G. (1994a) FEBS Lett352: 180-184.

Rahfeid, J.-U.. Schierhorn. A.. Mann, K., and Fischer, G.(1994b) FEBS Lett3A3: 65-69.

Rahman. M.A.. Nelson, H., Weissbach. H.. and Brot. N.(1992) J S/o/CUem 267: 15549-15551.

Reeve. C.A.. Amy. P.S.. and Matin, A. (1984) jeac(eTO/160:1041-1046.

Roseman, J.E., and Levins. R.L. (1987) J Biol Chem 262:2101-2110.

Roszak. D.B.. and Colwell. R.R. (1987) Microbiol Rev 51:365-379.

Rothstein. M. (1984) In Moiecuiar Basis ot Aging. Roy. A.K.,and Chatterjee. B. (eds). Orlando: Academic Press, pp.209-233.

Rudd, K.E., Sofia. H.J.. Koonin, E.V.. Plunkett, G., Lazar, S.,and Rouviere, P.E. (1995) Trends Biochem Sci 20:12-24.

Sak, B.D.. Eisenstark, A., and TouatI, D. (1989) Proc NatIAcad Sci USA 86: 3271-3275.

Schmid. F.X., Mayr, L.M.. Mucke. M.. and Schonbrunner.E.R. (1993) Adv Protein Chem 44: 25-66,

Seckler. R.. and Jaenicke, R. (1992) FASEBJ6: 2545-2552.Sherman. M.Y.. and Goldberg. A.L (1992) E/WSO J11 : 7 1 -

77.Siegele. D.A., and Kolter. R. (1992) J Bacteriol 174: 345-348.Skowyra. D.. Georgopoulos, C. and Zylicz. M. (1990) Ce//62:

939-944.Soumillion. P., and Fastrez. J. (1992) Biochem J2B6: 187-

191.

Spence, J.. Cegielska. A., and Georgopoulos. C (1990) JBacterion72: 7157-7166.

Squires, C. and Squires. C.L. (1992) JSacfer/o/174: 1081-1085.

Squires, CL., Pedersen, S.. Ross, B.M., and Squires. C.(1991) J Bacteriol 173: 4254-4262,

Stadtman, E.R. (1992) Science 257: 1220-1224.Stadtman. E.R- 11988) JGeronfo/43: B112-120.Stadtman. E.R., Starke-Reed, P.E.. Oiiver, C.N., Carney.

J.M., and Floyd. R.A. (1992) In Free Radicais and Aging.Emerit, I., and Chance. B. (eds). Basel, Switzerland:Birkhauser Verlag. pp. 64-72.

Stein. R.L. (1993) Adv Protein Chem 44: 1-24.Strauch, K.L.. Johnson. K.. and Beckwith. J. (1989) J

Bacteriol 171: 2689-2696.Straus, D.B., Walter, W.A.. and Gross, CA. (1988) Genes

Dev 2: 1851-1858.Strom, A.R.. and Kaasen, I. (1993) Mol Microbiol S: 205-210.Takayanagi, Y., Tanaka, K., and Takahashi. H. (1994) Mol

Gen Genet 243: 525-531.Tanford. C (1968) Adv Protein Chem 23: 121-282.Thompson, M.W.. Singh. S.K., and Maurizi. M.R. (1994) J

S/o/C/Jem 269: 18209-18215.Tormo, A., Aimiron. M.. and Kolter. R, (1990) JSactenon72:

4339-4347.Van Dyk. T.K., Gatenby, A.A., and LaRossa. R.A. (1989)

Wafure 342: 451-453.Watson, K. (1990) Adv Microbiat Physiol 3^: 183-223.Welch. W.J. (1993) Philos Trans R Soc Lond 339: 327-

333.Wickner. S.. Gottesman. S.. Skowyra, D., Hoskins, J..

McKenney. K., and Maurizi, M.R. (1994) Proc NatI AcadSciUSA9V. 12218-12222.

Wulfing. C. and Pluckthun. A, (1994) Moi Microbiol 12: 685-692.

Yamashino. T., Kakeda. M., Ueguchi, C, and Mizuno. T.(1994) /Wo/M/crab/o/13: 475-483.

Yaron, A., and Naider. F. (1993) Crit Rev Biochem Mol Biol28:31-81.

(Q. 1995 BlackweH Science Ltd, Molecular Microbiology. 16, 835-845