Determinants of Translocation and Folding of TreF, a Trehalase of Escherichia coli *

One isoform of trehalase, TreF, is present in the cytoplasm and a second, TreA, in the periplasm. To study the questions of why one enzyme is exported efficiently and the other is not and whether these proteins can fold in their nonnative cellular compartment, we fused the signal sequence of periplasmic TreA to cytoplasmic TreF. Even though this TreF construct was exported efficiently to the periplasm, it was not active. It was insoluble and degraded by the periplasmic serine protease DegP. To determine why TreF was misfolded in the periplasm, we isolated and characterized Tre+ revertants of periplasmic TreF. To further characterize periplasmic TreF, we used a genetic selection to isolate functional TreA-TreF hybrids, which were analyzed with respect to solubility and function. These data suggested that a domain located between residues 255 and 350 of TreF is sufficient to cause folding problems in the periplasm. In contrast to TreF, periplasmic TreA could fold into the active conformation in its nonnative cellular compartment, the cytoplasm, after removal of its signal sequence.

Secretory and cytoplasmic proteins differ not only by the signal sequence but also in their folding properties. It is thought that the export competence of secretory proteins is the result of slow folding prior to export. Cytoplasmic proteins are believed to fold more rapidly and thus are not substrates of the cellular secretion apparatus. For a better understanding of the mechanism of translocation, the folding of cytoplasmic and secretory proteins needs to be characterized. Folding of polypeptides is determined by the primary amino acid sequence but also could be influenced by the particular properties of a cellular compartment. However, little is known about whether and how the cytoplasm and the periplasm specifically influence protein folding, i.e. whether there are other elements besides redox state and the presence/absence of ATP. To study these aspects, we used TreA and TreF, the two trehalases of Escherichia coli, as model proteins.
Periplasmic TreA is synthesized as a precursor of 565 amino acids. The signal sequence of 30 amino acids is rather long.
Mature TreA has a molecular mass of 58 kDa (1,2). The K m of the purified enzyme is 0.8 mM, the V max is 66 mol of trehalose hydrolyzed/min/mg of protein, and the pH optimum is 5.5 (3). The expression of treA is independent of the presence of trehalose in the growth medium but is stimulated 10-fold at high osmolarity (1,2). Also, treA is regulated by RpoS, the stationary phase sigma factor (4).
Cytoplasmic TreF has 549 amino acids and a molecular mass of 64 kDa. The K m of the purified enzyme is 1.9 mM, the V max is 54 mol of trehalose hydrolyzed/min/mg of protein, and the pH optimum is 6.0. Like treA, the expression of treF is independent of trehalose, is dependent on RpoS, and is induced at high osmolarity. Both TreA and TreF are monomeric enzymes, and they share an amino acid sequence identity of 47%. TreA has an extended C terminus of about 30 residues, which is not present in TreF, and TreF has a 61-residue extension at its N terminus, which is not present in TreA (5).
Trehalose metabolism in E. coli can occur either in the periplasm via TreA or after transport into the cytoplasm via a trehalose-specific enzyme II encoded by treB (6). Transport results in the formation of trehalose 6-phosphate. Trehalose 6-phosphate is cleaved by TreC into glucose 6-phosphate and glucose (7). When trehalose is synthesized in the cytoplasm in response to high osmolarity of the growth medium, it can be degraded by cytoplasmic TreF. Cells expressing treF from the chromosome exhibit very low trehalase activity (5).
We are interested in the phenomenon that, despite their high similarity, TreA and TreF are localized in different cellular compartments, and we wished to determine whether these compartments have an effect on folding and specific activity. To study these questions we expressed both enzymes in their nonnative cellular compartment, i.e. treA in the cytoplasm and treF in the periplasm, and investigated whether they would fold into the active conformation. Whereas treA was actively expressed in the cytoplasm, periplasmic TreF was inactive. To determine why TreF is misfolded in the periplasm, we isolated and characterized Tre ϩ revertants of periplasmic TreF. To identify regions in TreF responsible for misfolding in the periplasm, we selected and characterized TreA-TreF hybrid proteins.
Construction of Plasmids-p⌬sstreA was generated by cleaving ptre11 with BsgI followed by T4 DNA polymerase treatment and PstI digestion. The 2.7-kilobase DNA fragment containing ⌬sstreA was ligated into pBAD22, which was cleaved with Asp178I, followed by treatment with Klenow enzyme and PstI digestion. Because of these manipulations, treA was expressed without its signal sequence and the first five amino acids of mature TreA (Glu, Glu, Thr, Pro, and Val) were replaced by Met, Val, and Leu.
ptreAЈ-ЈtreF was constructed by subcloning a 2.27-kilobase EagI fragment from ptre11 into pBADtreF which was cleaved with EagI. This plasmid contains the wild-type treA promoter and the 404 codons of the 5Ј-end of wild-type treA followed by treF containing a deletion of the first 165 codons. p⌬sstreAЈ-ЈtreF was constructed by subcloning the 3.05-kilobase EagI-ScaI fragment from pBADtreF into p⌬sstreA which was cleaved with EagI and ScaI.
ptreAЈ-treF expresses a gene fusion of the first 265 codons of treA and all of treF. It was constructed by cloning a polymerase chain reaction product of treF flanked by a KpnI restriction site at the 5Ј-end and an EcoRI restriction site at the 3Ј-end into ptre11 cleaved with KpnI and EcoRI. Primers used were 5Ј-GGGGTACCatgctcaatcagaaaattcaaaac-3Ј and 5Ј-CGGAATTCttatggttcgccgtaca aacc-3Ј. p⌬sstreAЈ-treF was constructed by subcloning of a 2.41-kilobase KpnI-PstI fragment into p⌬sstreA cleaved with KpnI and PstI.
psstreF was constructed by the subcloning of a 789-base pair polymerase chain reaction fragment containing the treA promoter up to the treA signal sequence flanked by a KpnI restriction site at the 3Ј-end as a BamHI-KpnI fragment into ptreAЈ-treF cleaved with BamHI and KpnI. Primers used were 5Ј-GGGGTACCttcttctgcctgcaccgat-3Ј and 5Јaatgggcatgcaaggagatgg-3Ј. This construct contained a treA sequence including the first two codons of mature treA, followed by codons for Gly and Thr and the entire treF gene. The additional Gly and Thr residues were a consequence of the introduced KpnI site immediately upstream of the translational initiation codon of treF.
Trehalase Assay-To determine the K m for trehalose of the various trehalase constructs, treA treC mutant strains KU92 or KU95 expressing the trehalase constructs from plasmids were grown overnight in rich medium. Cultures were washed twice in minimal medium and were broken in a french pressure cell at 9000 p.s.i. The remaining intact cells were removed by centrifugation (15,000 rpm, 30 min, 4°C, SS34 rotor). The DNA of the cell extract was precipitated with 2% streptomycin sulfate and pelleted by centrifugation (15,000 rpm, 30 min, 4°C, SS34 rotor). Subsequently, the protein concentration of the cell extract was adjusted to 0.2-0.4 mg/ml. Alternatively, to assay periplasmic trehalase activity of whole cells, cells were grown overnight in rich medium. Cultures were washed twice in minimal medium and were resuspended in trehalase assay buffer.
Trehalase assays were performed in 100 l of 20 mM potassium phosphate buffer, pH 6.0. The reaction was started by the addition of 0.25 to 10 mM trehalose and incubated for 5-15 min at room temperature. The reaction was stopped by boiling the samples for 5 min. Cleavage of trehalose was assayed by determining glucose concentration by the glucose test kit from Merck.
Labeling of Cells, Cell Fractionation, and Antibodies against TreF-Protein was labeled in cultures of strain KU101 expressing either treA or sstreF from plasmids ptre11 and psstreF, respectively, growing exponentially (A 600 0.4) in minimal medium 9 (11), 0.2% glucose, 1 g/ml thiamine, 50 g/ml ampicillin, supplemented with each common amino acid except Cys and Met. 1 ml of cells was exposed to [ 35 S]methionine (Ͼ1000 Ci/mmol) at 50 Ci/ml for 1 min and subsequently cooled on ice for immunoprecipitation or the labeling period was followed by a chase with excess cold methionine (50 mM final concentration) for the time indicated. To assay the effect of the temperature-sensitive lepB9 mutation on secretion of ssTreF, strain IT41 (12) was grown at 28°C and shifted to 42°C for 30 min before exposure to [ 35 S]methionine. Immunoprecipitation and gel electrophoresis were done as described by Ito et al. (13) and Laemmli (14). To precipitate TreA, TreF, and DegP, polyclonal antibodies against TreA, TreF, and DegP were used, respectively.
Rabbit polyclonal antiserum against TreF was obtained by immunization with purified TreF. The antiserum against TreF cross-reacts only weakly with TreA. Rabbit polyclonal antisera against TreA, MBP, DegP, or SecA were kind gifts of W. Boos, J. Beckwith, and P. C. Tai. The cold osmotic shock procedure was carried out according to Neu and Heppel (15).
Selection of Tre ϩ Revertants of sstreF-Strain KU101 containing psstreF was grown in LB in the presence of diaminopurine (600 g/ml). psstreF was isolated and retransformed into KU101. Tre ϩ revertants were selected on minimal medium A containing 0.2% trehalose (11). After purification on the selection medium, candidates were tested for elevated trehalase activity on MacConkey agar plates containing 1% trehalose. Candidates exhibiting red color on MacConkey plates were further tested by trehalase assays. The mutations of the Tre ϩ revertants were determined by nucleotide sequencing of the entire treF gene.
Selection of TreA/TreF and of ⌬ssTreA/TreF Hybrids-Strain KU92 containing ptreAЈ-ЈtreF was grown overnight in rich medium. A 25-ml culture was washed twice in minimal medium, and aliquots corresponding to 2.5 ml of cell culture were plated onto one minimal medium agar plate containing 0.2% trehalose. After incubation overnight, Tre ϩ candidates were purified twice on the selection medium. To show that complementation of the tre defect was linked to ptreAЈ-ЈtreF, plasmid DNA of such candidates was retransformed into strain KU92. In all cases growth on minimal trehalose media was detected. This indicated that plasmid-derived mutations were sufficient to generate growth on selection plates. These cells were used for further characterization. The fusion joints were determined by nucleotide sequencing. Selection of ⌬ssTreA/TreF hybrids was carried out as described above for selection of TreA/TreF hybrids except that strains KU93 and KU95 containing p⌬sstreAЈ-ЈtreF were used.
Electron Microscopy-For electron microscopy, strains were grown overnight at 28°C in LB medium. Cells were harvested by centrifugation (4000 ϫ g, 10 min, 4°C) and fixed for 1.5 h at room temperature with 2.5% glutaraldehyde in buffer A (0.1 M sodium phosphate buffer, pH 7.1). The pellet was washed three times with buffer A and embedded in 2% low melting agarose. These samples were fixed in buffer A containing 2% OsO 4 (1 h, room temperature), washed five times with buffer A, dehydrated in 50 and 70% acetone (10 min, room temperature), contrasted with 2% uranylacetate in 70% acetone (1 h, room temperature), and subsequently further dehydrated with 90 and 100% acetone. Embedding was done according to Spurr (16). After polymerization ultrathin sections were prepared and contrasted with 2% aqueous uranylacetate and lead citrate (17). The specimen were examined in a EM10 C2 (Zeiss, Oberkochen, Germany) under 80 kV at a primary magnification of ϫ17,600.

RESULTS
The two trehalases of E. coli are highly homologous and have similar enzymatic properties. The main apparent difference is their cellular localization. Therefore, these enzymes must contain signals allowing either efficient translocation and folding in the periplasm or exclusion from export and proper folding in the cytoplasm. To obtain information on how these isoenzymes have adapted to their environment we expressed both in their nonnative cellular compartment, periplasmic treA in the cytoplasm, and cytoplasmic treF in the periplasm.
Expression of treA and treF in Their Nonnative Cellular Compartments-To test whether periplasmic TreA can fold in the cytoplasm, we constructed a signal sequenceless TreA derivative. This plasmid was termed p⌬sstreA. ⌬sstreA was expressed in ⌬treA strain KU101. Signal sequenceless treA was actively expressed in the cytoplasm (Table I). The kinetic parameters of cytoplasmic TreA were determined in whole cell a Trehalase activity was determined by assaying the release of free glucose from trehalose. Assays were done in 20 mM potassium phosphate buffer (pH 6.0) in the presence of 100 mM trehalose. Trehalase activity determined in whole cells corresponds to periplasmic trehalase activity whereas both cytoplasmic and periplasmic trehalase activities contribute to the activity determined in whole cell extracts.
b -, no plasmid. c nd, not determined.
extracts of strain KU101. Signal sequenceless TreA had a K m of 0.16 mM, whereas periplasmic TreA had a K m of 0.31 mM trehalose. Cytoplasmic wild-type TreF had a K m of 1.5 mM. In addition, signal sequenceless TreA remained export competent, because it could be exported in prlA4 background (data not shown). prlA mutants are derivatives of secY allowing the export of signal sequenceless secretory proteins (18).
To test whether treF could be actively expressed in the periplasm, we fused the signal sequence of TreA to the N terminus of TreF. This plasmid was termed psstreF. TreF was efficiently exported to the periplasm. Pulse-chase experiments indicated that more than 95% of the signal sequences of the TreF precursor population were processed within 1 min (Fig.  1A). Only under nonpermissive conditions in the temperaturesensitive lepB9 strain IT41 (12) could the ssTreF precursor be detected (Fig. 1B). We concluded that processing of the signal sequence occurred as a consequence of translocation. Also, if a cytoplasmic protease would be responsible for generating the mature form of ssTreF, TreF would fold into the active conformation in the cytoplasm. Because we did not detect trehalase activity in whole cell extracts (see below), this explanation for efficient processing of ssTreF could be excluded.
sstreF was expressed at lower levels compared with treA, i.e. 37 Ϯ 6% of wild-type treA (Fig. 1). Because treF was expressed under treA promoter control and was fused to the TreA signal sequence, we expected comparable levels of expression. Be-cause the periplasmic DegP protease is known to degrade a large variety of misfolded proteins (19), we tested whether DegP is involved in degradation of periplasmic TreF. We detected an increase in treF expression in the degP null mutant strain KU104. This effect of the degP mutation was stronger in overnight than in log phase cultures (Fig. 1C). It should be noted that 60 times more cells were used for Western blotting compared with pulse-chase experiments. Other mutants of cell envelope proteases such as ompT, ptr, hhoA, and hhoB had no pronounced effects on TreF expression.
Enzymatic Activity of Periplasmic TreF-Periplasmic TreF had only background enzymatic activity regardless of whether TreF was expressed in degP ϩ or degP mutant strains (Table I). This finding corresponded to the weak growth of strain KU101 expressing sstreF on minimal trehalose agar plates. When growth was dependent on the presence of periplasmic TreA, single colonies were detected after overnight incubation at 28°C. When growth was dependent on periplasmic TreF, strains needed 3 days to form single colonies.
Trehalase activity was determined in either whole cells or whole cell extracts of treA::spec ⌬treBC strains, lacking periplasmic trehalase TreA and the trehalose transporter TreB. Trehalase activity of whole cells represented only periplasmic trehalase activity because in ⌬treB mutants trehalose is not transported across the cytoplasmic membrane. When trehalase activity was determined in whole cell extracts, periplasmic and cytoplasmic trehalase contributed to enzymatic activity. There was no difference when comparing trehalase activity of whole cells and cell extracts of cells expressing sstreF (Table I). This indicated that no residual TreF activity was present in the cytoplasm, which could be the result of inefficient export.
In addition, periplasmic TreF was present in an insoluble form. We were unable to extract TreF from cold osmotic shock fluids ( Fig. 2A). Electron microscopy suggested the presence of inclusion bodies in the periplasm (Fig. 2B). Similar results were obtained with MBP mutants (20). From these data we conclude that TreF can be exported to the periplasm where it is present in a misfolded and inactive form and is a substrate for DegP protease.
Genetic Selection of Periplasmic TreF with Increased Trehalase Activity-To determine why TreF is misfolded in the periplasm, we isolated and characterized Tre ϩ revertants of periplasmic TreF. To obtain mutants of periplasmic TreF with increased enzymatic activity, psstreF was mutagenized using 2-aminopurine. Mutagenized plasmids were transformed into treA::spec ⌬treBC strain KU101 and plated on minimal trehalose agar plates (Fig. 3). Colonies showing improved growth were screened for increased trehalase activity on MacConkey trehalose agar plates. Subsequently, trehalase activity of 25 candidates was assayed in whole cells. Two candidates showed more than a 5-fold increase in trehalase activity. Retransformation verified that the isolated mutations were linked to psstreF. Both mutants, termed treF82 and treF172, exhibited about a 6-fold increase in trehalase activity (Table I). Nucleotide sequencing indicated that they were independent mutants, because treF82 had a A82T and treF172 had a T172I exchange (Fig. 4). Interestingly, the T172I exchange detected in treF172 is located in the N-terminal trehalase signature. An amino acid sequence alignment shows that most trehalases have hydrophobic residues at position 172, whereas wild-type TreF has a Tyr residue (Fig. 5). This may explain why the T172I exchange leads to elevated trehalase activity.
The expression of treF82 and treF172 was found to be identical to periplasmic TreF. It is therefore likely that the mutations did not change the overall structure of periplasmic TreF but might have an effect only on folding of the catalytic site.
Characterization of a TreAЈ-TreF Hybrid-We reasoned that a hybrid protein composed of a significant fragment of TreA and TreF might mediate contact to potential periplasmic chaperones or folding catalysts and may thus influence folding of the TreF moiety. Fusing the N-terminal 263 amino acids of the TreA precursor to TreF (Fig. 6) led to translocation of the hybrid protein but did not increase trehalase activity (Table II) or solubility of the protein compared with periplasmic TreF (Fig. 2).
To test whether the presence of the 233 amino acids of mature TreA would interfere with folding of TreF we constructed a signal sequenceless derivative. It was expressed in the cytoplasm and exhibited high trehalase activity (Table II). Thus, the N-terminal 233 amino acids of TreA did not interfere with the folding of TreF in the cytoplasm and did not stimulate folding of periplasmic TreF.
Genetic Selection for Functional TreAЈ-ЈTreF Hybrids-To identify regions responsible for misfolding of periplasmic TreF, we wanted to test whether functional hybrids can be constructed composed of N-terminal fragments of TreA and Cterminal fragments of TreF. The genetic selection for active trehalase hybrids was identical to that used for selection of Tre ϩ revertants of sstreF. It was carried out in strain KU92 (treA::spec ⌬treBC), which requires periplasmic trehalase activity for growth on minimal trehalose medium (Fig. 3). To select for functional TreA-TreF hybrids, plasmids were constructed containing the 5Ј-end of treA until codon 373 fused to the 3Ј-end of treF starting at codon 166. The region of overlapping homology was about 750 base pairs. In this construct, termed treAЈ-ЈtreF, the reading frames of treA and treF were nonidentical. This construct did not confer trehalase activity, and no growth on minimal trehalose medium was observed after expression in strain KU92.
When plating cultures of KU92 containing ptreAЈ-ЈtreF spontaneous Tre ϩ revertants were isolated at high frequency. Plasmids of 12 candidates were retransformed into strain KU92 and checked for trehalase activity. Three candidates were chosen for further analysis. Nucleotide sequencing indicated that treA-F1 encodes a hybrid protein precursor of 535 amino acids composed of the N-terminal 335 residues of TreA precursor and the C-terminal 200 residues of TreF. treA-F8 encodes a hybrid protein precursor of 534 amino acids composed of the N-terminal 239 residues of TreA precursor and the C-terminal 295 residues of TreF. treA-F10 encodes a hybrid protein precursor of 535 amino acids composed of the N-terminal 342 residues of TreA precursor and the C-terminal 193 residues of TreF (Figs. 4 and 6). Processing of each precursor protein removes the N-terminal signal peptide of 30 residues. Because the fusion joints of treA-F1 and treA-F10 were nearly identical, we did not further investigate treA-F10.
Enzymatic Activity of Periplasmic TreAЈ-ЈTreF Hybrids-The hybrid proteins differed in their enzymatic activity. Whereas TreA-F1 exhibited high trehalase activity, TreA-F8 had low trehalase activity. Like all other constructs described above, the hybrids were substrates of DegP protease. In degP null mutant derivatives of strain KU92, protein expression (data not shown) and enzymatic activity increased (Table II). Also, we detected a difference in solubility. Whereas about 50% of TreA-F1 could be released by cold osmotic shock, only about 10% of TreA-F8 was released, indicating that a larger population of TreA-F8 was misfolded (Fig. 7). Because TreA-F8 had similar properties to periplasmic TreF but had only about 50% of TreF sequence, we suspect that the C-terminal half of TreF contains a major determinant for misfolding of periplasmic TreF. Because TreA-F1 exhibited enzymatic properties similar to TreA (see below) even though it contains the 200 C-terminal residues of TreF, the region causing folding problems could be further restricted to residues 255-350 of TreF.
Because wild-type TreA and TreF differ in their K m values for trehalose, i.e. 0.3 mM for TreA and 1.5 mM for TreF, we asked whether the trehalase hybrid TreA-F1 would resemble more closely TreA or TreF. The K m value for trehalose was determined as 0.3 mM, which was identical to that of periplasmic TreA. Thus, the high affinity binding site should be located in the TreA part of TreA-F1. A possible candidate for the high affinity binding site is the N-terminal trehalase signature (Figs. 4 and 5).
Genetic Selection and Characterization of Functional ⌬ssTreAЈ-ЈTreF Hybrids-We repeated the selection described above using signal sequenceless versions of ptreAЈ-ЈtreF. These were expressed in prlA4 strain KU93 allowing export of signal sequenceless secretory proteins. As above, we selected for periplasmic trehalase activity. Tre ϩ revertants were obtained at a 1000-fold lower frequency compared with the original selection using ptreAЈ-ЈtreF encoding the TreA signal sequence. Fortuitously, one isolate had exactly the same fusion joint as TreA-F1 and was thus termed ⌬ssTreA-F1. The fusion joint was such that a new trehalase gene was generated, lacking the extended N terminus of TreF and the extended C terminus of TreA. This trehalase was composed of 503 residues, whereas wild-type TreA has 535 residues and wild-type TreF has 549 residues. ⌬ssTreA-F1 contained three additional residues (Met, Val, and Leu) at its N terminus, which are a consequence of subcloning, 300 residues of TreA (37-334), and the 200 Cterminal residues of TreF (350 -549). The enzymatic activity of ⌬ssTreA-F1 was about 12-fold higher as detected for the signal sequence-containing construct TreA-F1 and was mostly localized in the cytoplasm (Table II). After expression in prlA degP mutant strain KU105, the activity of ⌬ssTreA-F1 was 0.89 Ϯ 0.02 units, which was only about 40% higher than detected in prlA ϩ degP ϩ strain KU101 (Table II) indicating that ⌬ssTreA-F1 was only a poor substrate of the prlA secretion apparatus. Therefore, we conclude that the C-terminal 200 residues of TreF were sufficient to block export by the PrlA secretion apparatus. The K m of ⌬ssTreA-F1 was determined in whole cell extracts as 0.16 mM for trehalose, which was identical to that of ⌬ssTreA. This result and the finding that the activity of ⌬ssTreA-F1 was the same as that of ⌬ssTreA (Tables I and II) indicated that the exclusion from export was not the consequence of improper folding of ⌬ssTreA-F1. DISCUSSION We used genetic approaches to study the intramolecular signals and the effects of the cellular compartments on translocation and folding of the two trehalases of E. coli. Periplasmic trehalase TreA folded properly in the periplasm and in the cytoplasm. Also, TreA-TreF hybrid proteins containing 334 amino acids of TreA and 200 residues of TreF had enzymatic properties comparable to those of wild-type TreA, no matter whether these constructs were expressed in the cytoplasm or in the periplasm. The experimental data obtained from these hybrid proteins indicated that the main determinant of substrate affinity must be localized in the N-terminal 334 residues of TreA. Trehalases have two conserved signatures typical for glycosyl hydrolases. One is localized at the N terminus and the other at the C terminus (Fig. 4). Thus, we speculate that the N-terminal trehalase signature is responsible for the lower K m values of TreA. In contrast to TreA, cytoplasmic TreF could not fold in its nonnative cellular compartment. Periplasmic TreF was misfolded and enzymatically inactive. One TreF mutation, T172I, located in the N-terminal trehalase signature, led to 6-fold higher periplasmic trehalase activity. This finding supported the model that the N-terminal Tre box is important for enzymatic activity. However, this mutation did not abolish the problems of solubility and protease sensitivity of periplasmic TreF.
It is difficult to secrete native cytoplasmic proteins of E. coli, which is consistent with the idea that cytoplasmic proteins fold too rapidly to be substrates of the secretion apparatus. Therefore, successful translocation of native cytoplasmic E. coli proteins has been reported only for a few cases. Thioredoxin 1 can be translocated to the periplasm when fused to the signal sequence of alkaline phosphatase or of DsbA (21,22), and ␤-galactosidase was exported as tripartite protein, LamB-LacZ-PhoA (23), or when fused to the signal sequence of OmpA (24). Compared with thioredoxin and ␤-galactosidase, secretion of TreF was more efficient, which may be because of the exceptionally long and hydrophobic signal sequence of TreA. We are currently investigating whether a more hydrophobic signal sequence is leading to a more efficient export of native cytoplasmic proteins.
A model predicting that the cytoplasm and the periplasm have different properties influencing the folding of polypeptides may explain why TreF did not fold properly in the periplasm. Periplasmic E. coli amylase MalS is another example because signal sequenceless MalS cannot fold into the active conformation in the cytoplasm (25). The cellular factors responsible for these effects are unknown. Possible candidates are different sets of molecular chaperones, for example ATPdependent chaperones are present in the cytoplasm but not in the periplasm. Also, because TreF contains three Cys residues, which tend to form intermolecular disulfide bonds in purified cytoplasmic TreF, these Cys residues could be responsible for misfolding of periplasmic TreF. When testing whether the inability of periplasmic TreF to fold was dependent on DsbA, a catalyst for disulfide bond formation, no increase in TreF activity or solubility was observed in dsbA knock out strains. 1 Also, there are no Cys residues in the TreF part of the hybrid TreA-F8, which was as insoluble as full-length ssTreF. However, one region comprising residues 255-350 of TreF was 1 K. Uhland, unpublished results.  a Trehalase activity was determined by assaying the release of free glucose from trehalose. Assays were done in 20 mM potassium phosphate buffer (pH 6.0) in the presence of 100 mM trehalose. Trehalase activity determined in whole cells corresponds to periplasmic trehalase activity, whereas both cytoplasmic and periplasmic trehalase activities contribute to the activity determined in whole cell extracts.

FIG. 7.
TreA-F1 is released by cold osmotic shock, whereas TreA-F8 is not. Western blot of periplasmic and pellet fractions of strain KU104 (degP::Tn10) expressing treAЈ-ЈtreF1 and treAЈ-ЈtreF8 after overnight growth in LB. To detect TreAЈ-ЈTreF, SecA, or MBP, polyclonal antibodies against TreA, SecA, or MBP were used. SecA and MBP were used as controls for cytoplasmic and periplasmic proteins, respectively. The presence of MBP in the pellet fraction results from intact cells. The position of TreAЈ-ЈTreF and its derivatives, SecA and MBP, are indicated. S is cold osmotic shock fluid, and P is the pellet of osmotically shocked cells.
identified as sufficient to cause folding problems in the periplasm. These 95 residues, of which only 24 are nonconserved with respect to TreA, are located between the fusion joints of treA-F1 and treA-F8. We are expecting that further work on TreF will allow us to study the determinants of solubility, folding of the active site, and protease sensitivity by using genetic methods.