INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Most therapeutically relevant proteins contain disulfide bridges and cannot be produced in their native conformation in the bacterial cytosol. In vitro refolding of inclusion body material is often laborious and costly. An alternative strategy to obtain these proteins in their native forms is to use their secretion into the periplasmic space. Targeting of proteins to the periplasm has both advantages and disadvantages. A major drawback of the periplasm is that space is limited. Thus, yields of recombinant proteins generally never match those obtained upon cytosolic expression. Also, translocation into the periplasmic space can limit the final yields of recombinant proteins. However, in the case of those proteins that bear multiple disulfide bonds of nonlinear connectivities in their native conformations and that are resilient to renaturation of inclusion body material, expression in the periplasmic space may offer the method of choice. The periplasm is a compartment where oxidation of thiols can occur due to the activity of the disulfide oxidoreductase (Dsb) system (for a review, see reference 28). The overall milieu of the periplasm is strongly oxidizing, with the DsbA protein being the major oxidant. However, Dsb components with disulfide isomerase functions, DsbC and DsbG, have also been described (5, 40). Still, presumably disulfide bond isomerization is insufficient in the periplasm, given that recombinant proteins that carry multiple disulfide bonds in their native conformations have a pronounced tendency to aggregate. Considering this major drawback of the expression of disulfide-containing proteins, the following strategies were devised to optimize folding in the periplasm: (i) modification of the medium composition by the addition of low-molecular-size compounds known to stimulate folding in vitro, (ii) addition of a redox component to allow reshuffling of wrongly formed disulfide bridges, and (iii) cosecretion of ATP-independent chaperones.

As model proteins, a truncated version of tissue-type plasminogen activator, consisting of the kringle-2 and the protease domain (BM 06.022, also known as rPA [23]), proinsulin, and a single-chain antibody fragment were chosen.

Our objective was to improve the yield of native rPA in the periplasm of Escherichia coli. A beneficial effect was observed upon the addition of low-molecular-size folding enhancers and reduced glutathione (GSH) and also upon cosecretion of either DnaJ or Hsp25. The general applicability of an optimized periplasmic expression compartment was confirmed with the two additional model proteins.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Genetic and protein analytic techniques. Cloning, transformation of E. coli cells, and DNA preparations were done by standard techniques (1). Oligonucleotides were purchased from Gibco BRL or MWG Biotech AG. Restriction enzymes were obtained from Roche Molecular Biochemicals GmbH, AGS GmbH, or New England Biolabs. Sequences of recloned DNA fragments were routinely confirmed by dideoxy sequencing (LiCor DNA-Sequencer 4000; LiCor, Lincoln, Nebr.). Sodium dodecyl sulfate-polyacrylamide gel electrophoresis and Western blotting were carried out as described in reference 8.

Strains, plasmids, proteins, and chemicals. E. coli strain BL21(DE3) was obtained from Novagen and used for gene expression; E. coli N4830/pPL-dnaJ-23 as a DnaJ overexpression strain was kindly provided by Thomas Langer (University of Munich, Munich, Germany).

Construction of expression plasmids. For cloning into pET20b(+) (Novagen), the coding sequence of rPA was PCR amplified from pA27fd7 (23) and inserted into pET20b(+). In this construct the second amino acid of rPA (Ser) is replaced by Ala. Proinsulin-encoding DNA was amplified by PCR from plasmid pRK-5-proinsulin (34) and ligated into pET20b(+). This vector mediates secretion via the pelB signal sequence. By QuikChange Mutagenesis (Stratagene), two surplus codons between signal sequence and proinsulin were removed. For coexpression of chaperones and model proteins, a two-plasmid expression system was chosen. After testing secretion of DnaJ and Hsp25, the genes were PCR cloned into pIN III ompA3 (16) and the coding sequences of DnaJ and Hsp25 with the regulatory sequences were recloned into plasmid pUBS520 (6), which bears the p15A replication origin and kanamycin resistance. This vector also carries the dnaY gene encoding the tRNA for the arginine codons AGA and AGG, which are rare in E. coli and thus often limit expression of genes with these codons (the gene for rPA contains seven of these rare arginine codons). The two-plasmid cosecretion system thus includes a vector for the secretion of the disulfide-bridged model protein on the ColE1-based pET vector and the chaperone on a p15A-based plasmid, which also carries the dnaY gene. Both the gene for the model protein and the chaperone were induced by the addition of isopropyl--D-thiogalactopyranoside (IPTG). Via PCR the coding and regulatory regions for scFv-ox were amplified from pHEN-scFv-ox (12) and cloned into pUBS520.

Cultivation of E. coli BL21(DE3) and rPA assay. In order to test rPA activity, cells were grown in Luria-Bertani medium at 24°C, induced with 1 mM IPTG at mid-log phase, and cultivated for a further 21 h. Medium additives (reduced GSH, 0 to 10 mM, and L-arginine, 0 to 0.4 M; formamide, 0 to 1 M; methylformamide, 0 to 1 M; acetamide, 0 to 1 M; methylurea, 0 to 1 M; or ethylurea, 0 to 1 M) were supplemented at the time of induction. After determination of the optical density at 600 nm (OD₆₀₀) (Pharmacia Ultrospec 3000; Pharmacia Biotech), 2-ml samples were collected and pelleted. For preparation of periplasmic extracts, the protocol described in reference 18 was downscaled to milliliter volumes. The soluble periplasmic fraction was assayed for rPA activity. For control purposes, cultures of E. coli BL21(DE3), transformed with pET20b(+) and pUBS520, were treated identically. Determination of functional rPA on microplates was performed according to a modified previously described protocol (38) with purified rPA as a standard. The concentration of rPA in the cellular extracts was determined by plotting the extinction against the square of the reaction time. The slope of a linear regression of this plot is directly proportional to the amount of rPA in the assay. The native state of rPA in extracts was tested in parallel assays after addition of 20 µl of 0.6-mg/ml fibrinogen fragments. The slope of the plot after addition of fibrinogen fragments divided by the slope in the absence of fibrinogen fragments defines the stimulation factor (23).

Expression studies and determination of scFv-TSH. E. coli BL21(DE3) transformed with pCANTAB5-TSH and pUBS520 was cultivated as described above in the presence of the indicated concentrations of reduced glutathione and L-arginine. Expression of scFv-TSH was determined via indirect enzyme-linked immunosorbent assay (ELISA) measurements (8) and detected using the ImmunoPure TMB substrate system (Pierce, Rockford, Ill.). The values of cell extracts without scFv-TSH were used for correction of background signal. scFv-TSH purified with the RPAS system (Amersham Pharmacia Biotech) was used as a standard.

Limited proteolysis of periplasmic DnaJ. E. coli XL1-blue cells, transformed with pIN III ompA3-dnaJ, secreting DnaJ, and N4830/pPL-dnaJ-23 cells, overexpressing DnaJ in the cytosol, were grown to mid-log phase and harvested 3 h after induction by centrifugation. The equivalent of 2 ml of bacteria of an OD₆₀₀ of 1 were converted to spheroplasts according to the method described in reference 37. The spheroplasts were resuspended in 30 µl of 50 mM Tris-HCl (pH 8.0)-100 mM NaCl with or without 0.1% Triton X-100. For limited proteolysis, aliquots of these fractions were incubated with 25 µg of trypsin per ml. Proteolysis was stopped by the addition of 20 M excess soybean trypsin inhibitor. In a control experiment, 0.1 mg of purified DnaJ per ml was incubated with 6 µg of trypsin per ml and treated in the same way as the spheroplast samples. The samples were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis and analyzed via Western blotting.

Cosecretion of proinsulin and DnaJ. E. coli BL21(DE3) cells harboring plasmids for cosecretion of proinsulin and chaperones were grown in Luria-Bertani medium at 25°C. One millimolar IPTG was added at an OD₅₀₀ of 1, and cells were harvested 6 h after induction. Soluble periplasmic protein was released by osmotic shock according to the method described in reference 22. For analysis and quantification of native proinsulin, an ELISA that specifically detects native (pro)insulin (Enzymun Test Insulin; Roche Diagnostics) was carried out.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Yields of secreted rPA and scFv-TSH in the presence of medium additives. Tissue-type plasminogen activator (tPA) converts the zymogen plasminogen to plasmin, a serine protease that degrades fibrin networks in thrombi (9). The tPA variant rPA contains nine disulfide bridges and aggregates upon cytosolic synthesis in inclusion bodies. In vitro refolding of rPA from inclusion body material is routinely performed (A. Stern, U. Kohnert, R. Rudolph, S. Fischer, and U. Martin, June 1993, U.S. patent application 5,223,256). As the native state of rPA can easily be assessed, it was chosen as a model protein for expression in the native conformation in the periplasm. For secretion of rPA, plasmid vector pET20b(+) (Novagen), containing the signal sequence of PelB (pectate lyase from Erwinia carotovora), was used. To determine the amount of functional rPA, protease activity was assayed according to the method described in reference 38 with minor modifications (see Materials and Methods).

View larger version (21K): [in this window] [in a new window]   FIG. 1.   Increases of the yields of secreted rPA and scFv-TSH upon addition of L-arginine to the cultivation medium. (A) Yields of native rPA in the periplasm of E. coli BL21(DE3)/pET20b(+)-rPA after cultivation (24°C) in the presence of the indicated concentrations of reduced GSH and L-arginine. Active rPA was determined as described previously (38). (B) Yields of native single-chain Fv (scFv-TSH) in the periplasm of E. coli BL21(DE3)/pCANTAB5-TSH after cultivation (24°C) in the presence of the indicated concentrations of L-arginine and GSH. Native scFv-TSH was determined using ELISA measurements. Mean values of at least three shake flask cultures and standard deviations are indicated.

View larger version (17K): [in this window] [in a new window]   FIG. 2.   Determination of the concentrations of L-arginine and GSH in the medium of E. coli BL21(DE3) after prolonged cultivation according to the methods described in references 13 and 17. (A) Determined L-arginine concentrations in the cultivation medium at the indicated time points after induction. At the time of induction L-arginine was added to the culture medium to final concentrations of either 0.2 M (filled circles) or 0.4 M (open circles). (B) Concentrations of GSH (closed triangles) and total GSH (GSH+GSSG; open triangles) in the cultivation medium after addition of 5 mM GSH.

Construction of a two-plasmid system for cosecretion of DnaJ and Hsp25. In order to further increase the yield of secreted proteins, cosecretion of ATP-independent chaperones was tested. In a first experiment, the cosecretion of DnaJ was analyzed. This protein belongs to the Hsp70 (DnaK) system of E. coli and is known to suppress aggregation of nonnative proteins also in the absence of Hsp70 (7, 24, 31).

View larger version (29K): [in this window] [in a new window]   FIG. 3.   Limited proteolysis for determination of native DnaJ. Spheroplasts were incubated with 25 µg of trypsin per ml and purified DnaJ was incubated with 6 µg of trypsin per ml for the indicated times. Proteolysis products of DnaJ and its fragments were detected with a rabbit anti-DnaJ antibody and subsequently with a donkey anti-rabbit immunoglobulin G horseradish peroxidase conjugate (Amersham Pharmacia Biotech).

Yields of native rPA and proinsulin in the periplasm of E. coli upon cosecretion of DnaJ and Hsp25. Cosecretion of DnaJ yielded a fivefold increase of functional rPA in periplasmic extracts compared to what was observed with the clone without cosecretion. Upon addition of fibrinogen fragments, protease activity was stimulated 35-fold, indicating the native conformation of the secreted rPA. Under optimal expression conditions (0.4 M L-arginine and 5 mM GSH), the yield increased 170-fold (Table 1).

View larger version (37K): [in this window] [in a new window]   FIG. 4.   Effects of cosecreted chaperones and scFv-ox (control) on the yields of native rPA at different concentrations of L-arginine. Cells were grown in the presence of 5 mM GSH. Active rPA was determined as described previously (38). Mean values of at least three shake flask culture experiments and standard deviations are indicated.

View larger version (29K): [in this window] [in a new window]   FIG. 5.   Yields of proinsulin after cosecretion of DnaJ, Hsp25, and scFv-ox (control). Proinsulin was determined by ELISA (see Materials and Methods). Values represent mean values of at least three shake flask culture experiments. Standard deviations are indicated.

Influence of low-molecular-size additives on the yield of secreted rPA. In in vitro refolding experiments, several low-molecular-size additives, especially derivatives of formamide or urea, proved useful for increasing the yield of native rPA (29). We therefore examined the effects of formamide, methylformamide, acetamide, methylurea, and ethylurea on the yield of native rPA. Bacteria were able to grow in media which contained concentrations of up to 1 M formamide or acetamide but only up to 0.6 M methylformamide, methylurea, or ethylurea. The yield of native rPA was tested with the strain E. coli BL21(DE3)/pUBS520-dnaJ/pET20b(+)-rPA cosecreting DnaJ upon cultivation in the presence of these additives and 5 mM GSH. Although L-arginine, which was used for comparison, proved to be the most effective additive, acetamide or ethylurea also had significant beneficial effects on the yield of rPA (Fig. 6).

View larger version (37K): [in this window] [in a new window]   FIG. 6.   Yields of native rPA in E. coli secreting DnaJ in the presence of 5 mM GSH and low-molecular-size additives at the indicated concentrations. Active rPA was determined as described previously (38).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Native expression of disulfide-bridged proteins in prokaryotic host cells remains a scientific challenge (32). Though approaches have been taken to change the cytosolic milieu of E. coli to more oxidizing conditions to allow intracellular formation of disulfide bonds (4, 35), expression of disulfide-bridged proteins in the periplasmic space is an alternative strategy that has not been fully exploited. A major disadvantage of the periplasm as a folding compartment for proteins with multiple disulfide bonds is the strong oxidant DsbA. DsbA has been shown to introduce disulfide bonds into translocating polypeptides as soon as two cysteines have emerged into the periplasm (20, 33). Although disulfide isomerases exist in the periplasm, their function is obviously insufficient to correct wrongly paired cysteines of proteins containing multiple disulfide bonds of nonlinear connectivities (3, 26). The consequence is usually inclusion body formation of these misfolded proteins in the periplasm (15).

Our approach to overcome these problems was to suppress inclusion body formation in the periplasm by adding disulfide-reshuffling reagents and substances known to stabilize folding intermediates to the cultivation medium. Also, the effects of cosecreted ATP-independent molecular chaperones DnaJ and Hsp25, which have been shown to suppress aggregation of nonnative proteins in vitro (11, 31), were analyzed.

We were able to increase the yield of native rPA in the periplasm of E. coli up to 170-fold upon cosecretion of DnaJ and 125-fold upon cosecretion of Hsp25. This huge increase is, to our understanding, mainly due to a synergistic effect of the respective cosecreted chaperone and medium additives on the folding of rPA, as cosecretion of DnaJ or Hsp25 in the absence of medium additives gave rise to ca. fivefold or twofold increases, respectively, of rPA (Table 1). Improvement of the periplasm as an expression compartment for disulfide-bridged proteins has been reported earlier (27). For example, overexpression of DsbC considerably increased the yield of full-length tPA (27). Unfortunately, the yields of functional proteins published in reference 27 and those of our studies cannot be compared, as a variant of tPA has been used in the latter; furthermore, the data of the former study result from high cell density fermentations, whereas here shake flask cultures were used.

The fact that 5 mM GSH was optimal for the folding of rPA under almost all tested conditions confirms previous results that demonstrate that addition of GSH improved folding of an -amylase-trypsin inhibitor in the periplasm (39).

With proinsulin, a 37-fold increase in the yield of native protein was obtained by cosecretion of DnaJ. Proinsulin secreted to the periplasm has been reported to be degraded by E. coli proteases (36). The presence of DnaJ may prevent the action of proteases and promote native structure formation. Cosecretion of Hsp25 or addition of L-arginine, however, did not improve the yield of native protein.

The increases of native rPA and proinsulin upon cosecretion of DnaJ and Hsp25 are likely to be due to the specific chaperoning activities of these proteins. Our interpretation that we are dealing with specific chaperone functions in cases where we observe increased amounts of folded proteins upon cosecretion of the chaperones is supported by the following observations. (i) If cosecretion of a heterologous protein should unspecifically enhance folding of rPA and proinsulin, cosecreted scFv-ox should have increased the yield. Furthermore, cosecretion of scFv-ox did not result in efficient stimulation of rPA activity by fibrinogen fragments by factors known for the completely folded protease. (ii) The chaperone requirement of a given protein is known to be relatively specific. In accordance with this notion, cosecretion of Hsp25 proved not to be effective in the case of proinsulin and neither DnaJ nor Hsp25 increased the yield of scFv-TSH. Thus, we propose that the beneficial effects of the secreted chaperones reflect the folding activities of DnaJ and Hsp25.

Besides L-arginine, a series of low-molecular-size reagents added to the cultivation medium increased the yield of active rPA. We limited our investigations to additives for which effects on the in vitro refolding of full-length tissue-type plasminogen activator had been demonstrated (29). Although L-arginine was the most effective compound in the case of rPA, other low-molecular-size additives may prove most efficient with different proteins. The effects of the tested additives on the yield of rPA are comparable to those obtained by in vitro refolding experiments (29). Therefore, we consider the effects to be due to the folding-enhancing activities of the compounds and not to their secondary osmolytic effects on cells.

Our results demonstrate that cosecretion of ATP-independent chaperones and the use of low-molecular-size medium additives to the culture medium can dramatically increase the yield of native eukaryotic proteins with complex disulfide patterns in the periplasm of E. coli. The mechanism by which the chaperones act in the periplasm remains unclear and needs further investigation. Still, this study may open new avenues for the production of disulfide-bridged proteins in their native conformation in prokaryotic organisms.