Folding of a mutant maltose-binding protein of Escherichia coli which forms inclusion bodies.

The maltose-binding protein (MalE) of Escherichia coli is the periplasmic component of the transport system for malto-oligosaccharides. We have examined the characteristics of a Mal- mutant of malE corresponding to the double substitution Gly32 --> Asp/Ile33 --> Pro, MalE31, previously obtained by random mutagenesis. In vivo, the MalE31 precursor is efficiently processed, but the mature protein forms inclusion bodies in the periplasm. Furthermore, the accumulation of insoluble MalE31 is independent of its cellular localization; MalE31 lacking its signal sequence forms inclusion bodies in the cytoplasm. The native MalE31 protein can be purified by affinity chromatography from inclusion bodies after denaturation by 8 M urea. The renatured protein exhibits full maltose binding affinity (Kd= 9 x 10(-7) M), suggesting that its folded structure is similar to that of the wild-type protein. Unfolding/refolding experiments show that MalE31 is less stable (-5. 5 kcal/mol) than the wild-type protein (-9.5 kcal/mol) and that folding intermediates have a high tendency to form aggregates. In conclusion, the observed phenotype of cells expressing malE31 can be explained by a defective folding pathway of the protein.

The maltose-binding protein (MalE) of Escherichia coli is the periplasmic component of the transport system for malto-oligosaccharides. We have examined the characteristics of a Mal ؊ mutant of malE corresponding to the double substitution Gly 32 3 Asp/Ile 33 3 Pro, MalE31, previously obtained by random mutagenesis. In vivo, the MalE31 precursor is efficiently processed, but the mature protein forms inclusion bodies in the periplasm. Furthermore, the accumulation of insoluble MalE31 is independent of its cellular localization; MalE31 lacking its signal sequence forms inclusion bodies in the cytoplasm. The native MalE31 protein can be purified by affinity chromatography from inclusion bodies after denaturation by 8 M urea. The renatured protein exhibits full maltose binding affinity (K d ‫؍‬ 9 ؋ 10 ؊7 M), suggesting that its folded structure is similar to that of the wild-type protein. Unfolding/refolding experiments show that MalE31 is less stable (؊5.5 kcal/mol) than the wild-type protein (؊9.5 kcal/mol) and that folding intermediates have a high tendency to form aggregates. In conclusion, the observed phenotype of cells expressing malE31 can be explained by a defective folding pathway of the protein.
In Escherichia coli, the export of proteins to the periplasm follows the general secretion pathway (Ref. 1). Considerable genetic and biochemical information about the export machinery has accumulated (2), but little is known about the conformational state of the polypeptide chain, either during its translocation through the membrane or upon its release into the periplasm.
Maltose-binding protein (MalE or MBP), the malE gene product, serves as the periplasmic receptor for the high affinity transport of maltose and maltodextrins (reviewed in Ref. 3). Because of its key role in maltose transport, correct export of MalE into the periplasm is essential for cells to utilize maltose as a carbon source. This feature facilitated the use of genetic selections for analyzing MalE export. Furthermore, the isolation of strains synthesizing MalE-LacZ hybrid proteins led to a novel genetic approach that identified several genes encoding the secretory machinery, the sec genes (4). As a protein model of translocation across the cytoplasmic membrane, MalE has also been extensively studied in the laboratories of the late P. Bassford and of L. Randall (5,6). The MalE protein is synthesized in the cytoplasm as a precursor protein, pre-MalE, with an amino-terminal extension, the signal sequence. As the precursor enters the periplasm, the signal sequence is cleaved by signal peptidase. Translocation requires that the precursor exist in an export-competent conformation representing a partially unfolded state (7). Both the signal peptide (8) and the binding of SecB, the molecular chaperone involved in protein export (1), participate in the maintenance of this initial conformation. Although Randall and Hardy (7) showed that MalE folding and export are kinetically competing processes, we do not know the extent of secondary and tertiary structures of precursor proteins. Another important unsolved problem concerns the releasing and folding steps of the polypeptide chain in the periplasm. Do specific proteins or molecular chaperones catalyze or facilitate this step in the periplasm? Two integral membrane proteins, SecD and SecF, are required for efficient protein translocation (9). Interestingly, treatment of spheroplasts with anti-SecD antibodies decreased the processing of pre-MalE and increased the protease susceptibility of the mature species (10). These results would be consistent with a molecular chaperone function of SecD on the periplasmic face of the inner membrane (11).
Protein aggregates are formed in vivo as a result of: (i) mutations that affect the folding pathway; (ii) expression of heterologous proteins; (iii) exposure of the cell to certain environmental stresses. It is generally accepted that aggregation and folding are competing processes and that aggregation is favored at elevated protein concentrations. A model suggesting that a folding intermediate is responsible for inclusion body formation has been proposed by Mitraki and King (12). Such a kinetic competition between the folding pathway and an aggregation reaction has previously been recognized as a major determinant for apparent irreversibility in refolding studies in vitro (13). Since aggregation is a second (or higher) order process, it can be much faster than first-order folding and therefore, at high protein concentrations, aggregation dominates over folding and leads to the formation of insoluble protein (14). The observation that temperature-sensitive folding (tsf) mutations in the tailspike protein of phage P22 increased the fraction of newly synthesized tailspike chains forming aggregates demonstrated the importance of folding intermediates in the mechanism of inclusion body formation (15). The tsf mutations prevent formation of the native protein at the restrictive temperature, but once folded at a lower permissive temperature, the mutant tailspike proteins have stabilities similar to that of the native protein. Intragenic suppressors of these mutations inhibit aggregation of folding intermediates and increase the folding efficiency of tsf mutants (16). Mutations influencing inclusion body formation were also identified for human interleukin-1␤, expressed in the cytoplasm of E. coli (17). The mutated residues were mainly located either in a ␤ strand or in a flexible loop. However, there was no strong correlation between the thermodynamic stability of the mutant interleukin-1␤ proteins and their tendency to form inclusion bodies. Another example of studies of inclusion body formation in E. coli is provided by the enzyme ␤-lactamase (18). In the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. case of such an exported protein, the formation of inclusion bodies depends on the protein synthesis rate and the amino acid sequence of the signal peptide (19). High level of expression of ␤-lactamase resulted in the accumulation of precursor chains: aggregation occurred both in the cytoplasm from the precursor and in the periplasm from the mature ␤-lactamase. Despite the widespread reports of inclusion body formation resulting from very high level production of recombinant proteins in E. coli, there is little information on the molecular basis governing this process.
In this study, we characterized a Mal Ϫ mutant of the maltose binding protein (MalE31), previously isolated by random linker insertion into the malE gene (20). Our in vivo analysis revealed that the mutant MalE31 polypeptide chain forms inclusion bodies after translocation and signal peptide processing. We further report the in vitro analysis of the stability and aggregation of this mutant protein compared with the wild-type protein. These data confirmed that the observed phenotype of cells expressing malE31 can be explained by a defective folding pathway of the protein.

MATERIALS AND METHODS
Bacterial Strains and Plasmids-E. coli K12 strain pop6499, which is a derivative of MC4100 (F Ϫ araD139 thiA relA ⌬lacU169 ⌬(srlR-recA)306::Tn10 rpsL malT c 1 ⌬malE444) was used as the host strain for plasmids encoding MalE proteins. This strain carries a nonpolar deletion of the chromosomal malE gene (21) and harbors the malT c 1 allele which confers constitutive expression on the maltose operons (20). Therefore maltose was not added to the growth medium. The expression of a plon::lacZ fusion was monitored in strain SR1364, which is a derivative of MC4100. 1 Plasmids pPD1 and pPD364 carrying the wild-type malE gene (21) and the mutant malE31 gene (20), both under the control of the mal promoter, were designated here as pHCME and pHCME31, respectively. Plasmid p⌬709, which carries a deletion of the ribosome binding site and the first 256 codons of the malE gene (22), was used as a negative control.
Cell Fractionation-Cultures were grown in LB medium supplemented with 0.1 mg/ml ampicillin at 30°C. After 3 h, cells were harvested by centrifugation and fractionated by spheroplast preparation. The cell pellets, normalized to the same absorbance reading at 600 nm (5 ϫ 10 8 cells/ml), were suspended in 10 mM Tris-HCl (pH 7.5) containing 0.7 M sucrose and 1 mM phenylmethylsulfonyl fluoride. Lysozyme (0.2 mg/ml) and EDTA (10 mM) were added, and the suspensions were incubated for 20 min at 4°C. Then, the samples were centrifuged for 5 min at 10,000 rpm in an Eppendorf microcentrifuge, and the supernatants (periplasmic fraction) were removed. The pellets were resuspended, washed, freeze-thawed twice, then centrifuged as described above. Supernatants were kept (cytoplasmic fraction) and the pellets resuspended in Tris-HCl buffer and washed (membrane fraction). The three fractions were mixed with an equal volume of 2% SDS sample buffer (0.1 ml) and heated to 100°C for 3 min before being analyzed by SDS-polyacrylamide gel electrophoresis (SDS-PAGE). 2 The gels were stained with Coomassie Blue and scanned with a MasterScan I interpretive densitometer (Scanalytics).
Radiolabeling and Immunoprecipitation-Cells were grown at 30°C in M63B1 medium supplemented with 40 mg/ml amino acids minus methionine, containing 0.1 mg/ml ampicillin and 0.4% glycerol. During the mid-exponential phase of growth, cells were labeled with 60 mCi of [ 35 S]methionine/ml for 15 s, then chased with 4 mg/ml nonradioactive methionine for 10, 30, and 60 s. In experiments in which export was blocked, sodium azide was added to a final concentration of 5 mM, 5 min before labeling. All the samples were precipitated with 10% cold trichloroacetic acid and solubilized in 50 mM Tris-HCl buffer (pH 8) containing 1% SDS, 1 mM EDTA, and 1 mM phenylmethylsulfonyl fluoride. MalE-wt and MalE31 were immunoprecipitated using rabbit anti-MalE serum as described previously by Ito et al. (24). Immune precipitates were analyzed by SDS-PAGE and autoradiography. Transmission Electron Microscopy-Cells were grown in LB at 30°C as described above, harvested, fixed with 2.5% glutaraldehyde in 0.1 M cacodylate (pH 7.4) for 1 h at room temperature, and post-fixed in 1% osmium tetroxide, stained with 0.5% uranyl acetate, then dehydrated through an ascending series of ethanol and embedded in Spurr. Thin sections were examined using a Philipps CM12 transmission electron microscope.
Protein Purification-The wild-type MalE protein was purified by osmotic shock procedure as described previously (20). The MalE31 protein was purified from inclusion bodies. Cultures were grown to late exponential phase at 30°C in LB broth (2 liters) with 0.1 mg/ml ampicillin. Cells were harvested and suspended in 50 ml of 25 mM Tris-HCl buffer (pH 7.5) (buffer A) containing 0.1 mg/ml DNase I and lysed in a French press at 12,000 p.s.i. The lysate was centrifuged at 14,000 ϫ g for 15 min, and the pellet (inclusion bodies and membranes) was suspended in 20 ml of buffer A containing 2% Triton X-100. After centrifugation at 14,000 ϫ g for 15 min, the pellet was solubilized by 8 M urea in buffer A and incubated 30 min at room temperature. Renaturation was achieved by dilution (20-fold) into buffer A at 4°C. The cloudy renatured mixture was centrifuged at 14,000 ϫ g for 15 min, and the supernatant was loaded onto a cross-linked amylose column at 4°C (25).
Preceding all biochemical characterizations, both purified proteins were extensively dialyzed three times against 3 liters of buffer A at 4°C. The first five amino acid residues of purified MalE31 protein were identified by using an Applied Biosystems 473 protein sequencer. The protein concentration of MalE-wt and MalE31 solutions was determined from the absorbance at 280 nm with an extinction coefficient of ⑀ ϭ 68,750 M Ϫ1 ⅐cm Ϫ1 . 3 Maltose Transport and Binding Assay-The initial rate of maltose transport was estimated by measuring uptake of [ 14 C]maltose as described by Szmelcman and Hofnung (26) with 1.2 M maltose (final concentration). Equilibrium dissociation constants (K d ) were assessed by fluorescence titration using both purified proteins essentially as described by Miller et al. (27). The tryptophanyl fluorescence quenching was recorded on a Fluoromax spectrofluorometer (Spex) with an excitation wavelength of 290 nm. 2-l aliquots of successive 10-fold dilutions of a 1 M maltose solution were added to protein samples (2 ml at 0.5 M), and the emission spectra were recorded between 300 and 400 nm.
In Vitro Unfolding-Refolding Studies-Unfolding and refolding experiments were carried out in buffer A at 25°C. The unfolding transitions of MalE-wt and MalE31 by guanidine hydrochloride (GdnHCl) at equilibrium were measured by fluorescence emission at 345 nm (excitation was 290 nm). The final concentration was 0.5 M for both proteins. Concentrations of GdnHCl solutions were determined by refractive index measurements (28). The experimental data were analyzed by using the model of linear dependence of ⌬G upon denaturant concentration x (29).
In order to take into account the effect of the solvent on the fluorescence signal, the following equation was used, where Y x is the fluorescence signal in the presence of ϫ molar GdnHCl, Y n and Y d are the signal of the native and denatured forms, respectively, at the same denaturant concentration; M n and M d are the solvent effects on the native and denatured protein signal, respectively. Experimental data were fitted by using a simplex procedure based on the 1 S. Raina, unpublished results. 2 The abbreviations used are: PAGE, polyacrylamide gel electrophoresis; GdnHCl, guanidine hydrochloride. 3 J.-M. Betton, unpublished result.
Nelder and Mead algorithm (30). To study the reversibility of the unfolding process, MalE-wt and MalE31 proteins at two concentrations (1 and 10 M) were incubated with varying concentrations of GdnHCl for 6 h. They were then diluted in the Tris-HCl buffer, incubated for 15 h at 25°C, and centrifuged at 10,000 rpm for 10 min in an Eppendorf microcentrifuge. Finally, the amount of native protein in the supernatant was determined by using the two-antibody sandwich assay described by Martineau et al. (31). The monoclonal antibody 56.5, which was used in this assay, bound to a conformational epitope of MalE or MalE31 protein with the same affinity (K d ϭ 5 ϫ 10 Ϫ10 M). 4

RESULTS
We previously described a set of MalE mutant proteins generated by random insertion of an oligonucleotide linker into the malE gene (20). Among this collection, six mutations prevented both growth on maltose as a carbon source and the periplasmic release of the corresponding proteins by cold osmotic shock (32). In all cases but one, a deletion of several amino acid residues from the mature region of the protein could explain their defective phenotypes. However, in one case, malE31, the linker insertion did not modify the length of the protein, but changed six nucleotides, resulting in Gly 3 Asp and Ile 3 Pro substitutions at position 32 and 33 of the mature sequence, respectively (20).
Properties of MalE31 and Its Derivatives-To identify the particular codon in malE31 responsible for the Mal Ϫ phenotype, derivatives carrying either the single mutation at position 32 (MalE-G32D) or at position 33 (MalE-I33P) were constructed. The biological properties of the corresponding mutants are indicated in Table I. Although bacteria carrying malE31 had altered growth properties, as indicated by their doubling time in maltose minimal medium or their phenotype on MacConkey dextrin plates (Table I), production of MalE31 did not affect growth in rich medium (data not shown). The single substitution G32D led to a mutant with biological properties similar to those of cell expressing wild-type malE. In contrast, the Ile 33 3 Pro substitution affected both growth and maltose transport, but the defects were less severe than in malE31. While the level of total expression in cells producing these various proteins was similar, as assessed on SDS-PAGE, the amount of MalE31 and its derivatives in the periplasmic fraction correlated with their ability to grow on maltose minimal medium (Table I). The cellular localization of the MalE31 protein and its derivatives was determined after spheroplast preparations (Fig. 1). The periplasmic and cytoplasmic localization of wild-type MalE expressed from pHCME or from pHCME⌬2-26, coding for signal sequence deleted MalE pro-tein, confirmed that the cell fractionation procedure was correct. However, the pHCME31-encoded MalE31 protein was entirely recovered in the subcellular membrane fraction. This result suggested that export of MalE31 protein was blocked somewhere in the membrane or, alternatively, that MalE31 formed inclusion bodies either in the cytoplasm from the precursor, before translocation, or in the periplasm from the mature species, after translocation. Interestingly, a large fraction of pHCME31⌬2-26-encoded MalE31 (73%) was also found in the membrane fraction. Close inspection of the gel revealed the presence of increased amounts of two proteins of about 60 and 70 kDa, in the cytoplasmic fraction of cells carrying pHCME31⌬2-26. These proteins correspond to the major heatshock proteins GroEL and DnaK (see below). While the MalE-G32D protein was localized entirely in the periplasmic fraction (98%), most of MalE-I33P (63%) was in the membrane fraction. Although, some synergistic effect was observed with the double substitution, these data confirmed that the Ile 33 3 Pro substitution is mainly responsible for the defective phenotype of cells expressing malE31.
MalE31 Forms Inclusion Bodies-In the experiment described above, MalE31 sedimented with the membranes, irrespective of whether or not it was made as a signal peptidebearing precursor. To determine whether these proteins were in fact periplasmic or cytoplasmic, respectively, or were indeed membrane-associated, we performed two kinds of experiments.
First, the kinetics of signal sequence processing was determined using a pulse-chase analysis. Processing of the precursor, a late step in export, is a reliable indicator that the protein has, at least partially, crossed the inner membrane (1). Cells were pulse-labeled with [ 35 S]methionine and then chased with nonradioactive methionine for different times. The labeled MalE proteins were analyzed by immunoprecipitation. Fig. 2 shows that the MalE31 precursor is processed very rapidly and that an export defect cannot be observed at the earliest time point (10 s). In conclusion, there were no differences in the processing kinetics observed for pre-MalE31 species compared with those of wild-type pre-MalE. Furthermore, amino-terminal microsequence analysis of purified MalE31 protein (see below) confirmed that pre-MalE31 was processed at the normal cleavage site.
Second, transmission electron microscopic studies of cells harboring pHCME31 revealed the presence of electron dense material between the inner and outer membranes of the bacteria (Fig. 3A). Periplasmic inclusion bodies of MalE31 were small and frequently included more than one per cell. In contrast, cells producing the signal sequence deletion MalE31 protein (pHCME31⌬2-26) had rare inclusion bodies that were clearly localized within the cytoplasm and that had blurred boundaries (Fig. 3B). By analogy with other proteins which form inclusion bodies in E. coli (33), the amount of insoluble MalE31 was temperature-dependent, with higher growth temperatures promoting inclusion body formation (data not shown). These results indicated: (i) that the MalE31 precursor was correctly processed and thus that the mature protein is localized in the periplasm, (ii) that the malE31 mutation causes MalE aggregation whatever its cellular localization, and (iii) that inclusion body formation in the periplasm arose from the association of processed MalE31 protein Heat-shock Response Induced by MalE31 Production-From Fig. 1 it is evident that cytoplasmic accumulation of MalE31 (pHCME31⌬2-26) increased the synthesis of two proteins which correspond to the GroEL (60 kDa) and DnaK (70 kDa) proteins. In contrast, no such elevated level of heat-shock proteins was visible with periplasmic MalE31 aggregates (pH-CME31). To determine quantitatively the extent of heat-shock 4   gene induction caused by MalE31, we used a plon-lacZ promoter fusion construct transcribed by the sigma factor 32 , the rpoH gene product (strain SR1364). 1 This strain was transformed with plasmids pHCME31 and its derivatives, grown at 30°C, and ␤-galactosidase activities were determined according to Miller (34). The same level of ␤-galactosidase induction was observed in all cases, except for pHCME31⌬2-26 (Fig. 4). In this case, the synthesis of ␤-galactosidase directed from the lon promoter was 2-fold higher than for strains containing the other plasmids. This result confirmed that periplasmic inclusion bodies of MalE31 did not induce the 32 -dependent heatshock response.
In Vitro Folding and Aggregation of MalE31-The production of misfolded proteins that could form inclusion bodies is one of the main factors responsible for induction of heat-shock response in E. coli (35,36). To explore this possibility, we studied the in vitro folding characteristics of purified MalE31. We purified soluble MalE31 protein to homogeneity from periplasmic inclusion bodies by a simple denaturation-renaturation step and affinity chromatography (Fig. 5). The maltose binding affinity for purified wild-type and MalE31 proteins were assayed by fluorescence quenching experiments. As indicated in Table II, the two proteins display the same maltose binding affinity (K d ϭ 9 ϫ 10 Ϫ7 M).
To investigate the folding properties of MalE31, the GdnHClinduced unfolding transitions of the wild-type and MalE31 proteins were determined using tryptophanyl fluorescence as a probe of the tertiary structure (Fig. 6). At the final protein concentration of 0.2 mM, both transitions were found to be reversible (see below) and symmetrical. The concentration of GdnHCl at the midpoint of the transition (C m ) for MalE-wt was 0.92 Ϯ 0.02 M, whereas the midpoint occurred at 0.62 Ϯ 0.02 M GdnHCl for MalE31 (Table II). This reflects the lower stability of the mutant MalE31; the ⌬G H 2 O value obtained by curve fitting using Equation 2 is only Ϫ5.5 kcal/mol, compared with Ϫ9.5 kcal/mol for MalE-wt (Table II).
Light scattering and apparent irreversibility were detected when the MalE31 concentration was increased in the unfolding experiments described above. To test the effects of protein concentration on renaturation, we performed a two-step dilution experiment (37). Starting from the native state, MalE31  was incubated in various GdnHCl concentrations for 6 h at 25°C and then diluted into the renaturation buffer (Fig. 7). Two different protein concentrations in the first step were explored, and we used a quantitative immunoassay to monitor the recovery of native MalE31 because of its high sensitivity (31). The renaturation became partially irreversible when the protein had been exposed to concentrations of GdnHCl corresponding to the end of the unfolding transition zone. However, when refolding was initiated from 3 M GdnHCl, this phenomenon was less pronounced. The extent of irreversible denaturation of MalE31 was strongly dependent on the protein concentration, while no such effect of the MalE-wt protein concentration could be detected up to 10 mM. This phenomenon was detected only upon monitoring the native protein with the immunoassay. Indeed, using the same procedure as described for Fig. 7, but with a MalE31 protein concentration of 1 M, complete reversibility was found by fluorescence measurements (data not shown). These results revealed the occurrence of an aggregation reaction arising mainly from MalE31 equi-librium folding intermediates. It is tempting to speculate that the aggregation of MalE31 folding intermediates is correlated with the formation of inclusion bodies.

DISCUSSION
In vivo and in vitro analyses of MalE31 folding revealed that the variant polypeptide chain had a high tendency to aggregate. The three-dimensional structure of MalE consists of two globular domains separated by a cleft in which maltose and maltodextrins bind (38). Each of the two domains is constructed from secondary structural elements belonging to both the amino-and carboxyl-terminal halves of the protein, thereby forming a central ␤-sheet framed by ␣-helices. The structural location of amino acid substitution corresponding to the malE31 mutation is shown in Fig. 8. Positions 32 and 33 are located in a turn connecting helix I to strand B forming the first ␤␣␤ supersecondary structure of the amino-terminal domain. In other proteins, similar structural locations were also found for amino acid substitutions which promote aggregation (17,   6. Equilibrium denaturation curves. The GdnHCl-induced transition was assessed by the fluorescence emission at 345 nm (excitation wavelength: 290 nm). q, MalE-wt; f, MalE31. The normalized data were obtained by using the relation where Y x refers to the observed fluorescence intensity, Y n and Y d are the fluorescence intensity of the native and unfolded protein, respectively, and f N is the fraction of folded protein at ϫ molar GdnHCl concentration.

40)
. This observation is in agreement with the significant role of turns and loops in the folding pathway of ␣/␤ proteins (41). The double substitution, Gly 32 3 Asp/Ile 33 3 Pro, destabilizes the protein by about 4 kcal/mol, probably as a result of changes to geometrical constraints at the turn region. Furthermore, despite a synergistic effect, the single Ile 33 3 Pro substitution was the major factor responsible for the aggregation. It is possible that a turn which includes Pro 33 could introduce a conformational strain in the folded state because of the limited value of the angle (42), even though the turn is in a solventaccessible region. Moreover, if the introduction of a proline residue in this turn destabilizes the native structure, it does not affect the affinity of maltose binding of MalE (Table II).
By studying the unfolding-refolding transition of MalE variants corresponding to intragenic suppressors of export-defective signal peptides, Randall and co-workers (43) have suggested that the rate-limiting step in the MalE folding pathway could be the formation of a supersecondary structure localized in the amino-terminal domain. Strikingly, the altered turn which leads to in vivo and in vitro aggregation of MalE31 connects an ␣-helix and a ␤-strand belonging to the same substructure of the amino-terminal domain. Intragenic sup-pression of the export-defective signal peptide is accomplished by slowing the folding of the precursor, thereby increasing the time during which the protein is competent to enter the export pathway (44). cis-trans isomerization of prolyl imide bonds might be expected to introduce slow steps in protein folding, thus increasing the lifetime of intermediates and promoting inclusion body formation. Preliminary results on the kinetics of refolding have shown that, like intragenic suppressors of defective peptide signal, MalE31 refolds more slowly than MalE-wt in vitro. However, this slowest refolding step could not be explained by a slow isomerization step in the unfolded state of MalE31. 5 Inclusion body formation and in vitro aggregation are similar phenomena, originating from the failure of folding intermediates to achieve final folding. The putative cellular pathway for export and folding of precursors is shown in Fig. 9. As the precursor protein elongates, it begins to acquire secondary structure. This newly synthesized precursor may face many competing pathways, including binding molecular chaperones, folding, degradation, aggregation, and export. All of these possible outcomes depend on the intrinsic folding properties of the precursor being synthesized (or its folding rate relative to the rate of other processes). In this schematic drawing, molecular chaperones play an essential role in determining the choice between folding and export, but they may also participate in proteolysis by maintaining proteins in an unfolded state accessible to proteases. MalE belongs to a subset of precursors that interact preferentially with SecB, which seems to be entirely dedicated to export (45). However, heat-shock proteins can substitute for SecB during protein export. Indeed, overproduction of DnaK and DnaJ improved the export of wild-type pre-MalE in a secB::Tn5 mutant (46). Although periplasmic molecular chaperones involved in general protein folding have not yet been identified, we assume that such molecules could exist in this compartment. Another class of periplasmic proteins is enzymes that can either directly catalyze disulfide bond formation or maintain an appropriate oxidized environment in the periplasm. The lack of Cys residues in MalE eliminates the possible complication of intra-and/or extramolecular disulfide bond formation.
Our results suggest that the step at which productive folding of MalE31 is blocked is likely to be a folding intermediate localized in the periplasm. These data do not fit the model of alternative pathways open to protein precursors within the cell, as defined by Randall and Hardy (48). Indeed, they proposed that the kinetic partitioning between folding and aggregation depends strongly on binding to SecB in the cytoplasm. Although direct experiments on the interactions between SecB and MalE31 have not been performed, we have independently confirmed by several approaches that the MalE31 precursor did not accumulate in the cytoplasm. This behavior would indicate that the altered turn of the MalE31 precursor affects neither its export-competent conformation nor its ability to bind SecB. Structurally, a nascent precursor in an exportcompetent conformation may be similar to misfolded proteins. Indeed, Wild et al. (49) showed that accumulation of secretory protein precursors in strains lacking the SecB chaperone generates a signal for induction of heat-shock proteins. Obviously, the absence of increased 32 activity in bacteria producing pre-MalE31 eliminates this possibility. However, the formation of periplasmic inclusion bodies of MalE31 specifically triggers the 24 -dependent stress. 6 This alternative heat-shock sigma factor is involved in response to stress occurring in extracyto- 5  of the x-ray structure of MalE was generated from coordinates kindly provided by F. Quiocho using the program Molscript (39). The location of the altered ␣␤ turn is arrayed. B, the two substituted residues, Gly 32 and Ile 33 , connecting helix 1 and strand B are shown by thick lines.
plasmic compartments (50). It is worth noting that the soluble fraction of MalE31 that is periplasmic (2%) is less than the soluble fraction of MalE31 that is cytoplasmic (27%). This observation suggests that in the latter case, the increased level of GroEL and DnaK, that results from induction of the heatshock response by MalE31, can partially suppress the cytoplasmic aggregation of MalE31. It appears that under these experimental conditions, there are insufficient molecular chaperones in the periplasm to prevent or reduce aggregation of MalE31 Finally, this work on the aggregation of MalE31 opens the way to the selection of mutants affected in protein folding and/or aggregation in the E. coli periplasm. Indeed, malE31 carried by plasmid pHCME31 was not able to complement the deletion ⌬malE444 of strain pop6499. If general molecular chaperones exist in the periplasm of E. coli, they might modulate the aggregation reaction from the defective folding intermediate of MalE31 as they do in the cytoplasm. Thus, mutations that improve protein folding in the periplasm could be selected by their ability to promote growth on maltose.