Compartmentalization of mammalian proteins produced in Escherichia coli

Export and folding of signal-sequenceless Bacillus licheniformis beta-lactamase in Escherichia coli

Two genetically engineered variants of the Bacillus licheniformis beta-lactamase gene were expressed in Escherichia coli. One variant coded for the exo-small mature enzyme without the signal peptide. The other coded for the exo-large mature enzyme preceded by 10, mostly polar, residues from an incomplete heterologous signal. As observed following the extraction by a lysozyme-EDTA treatment, the signal-less variant was exported to the periplasm with nearly 20% efficiency, whereas the variant with the N-terminal extension was translocated to a lesser degree; interestingly, nearly all of the former and half of the latter were extracted by osmotic shock, which may be of importance for our understanding of cellular compartments. The fact that a signal-less protein is translocated with substantial yields raises questions about the essential role of signal peptides for protein export. As folding and export are related processes, we investigated the folding in vitro of the two variants. No differences were found between them. In the absence of denaturant, they are completely folded, fully active and have a large DeltaG of unfolding. Under partially denaturing conditions they populate several partially folded states. The absence of significant amounts of a non-native state under native conditions makes a thermodynamic partitioning between folding and export less likely. In addition, kinetic measurements indicated that these B. licheniformis lactamases fold much faster than E. coli beta-lactamase. This behavior suggests that they are exported by a kinetically controlled process, mediated by one or more still unidentified interactions that slow folding and allow a folding intermediate to enter the export pathway.

A single amino acid deletion in the carboxy terminal of apolipoprotein A-I impairs lipid binding and cellular interaction

The carboxy-terminal region of apolipoprotein (apo) A-I has been shown by mutagenesis or synthetic peptides to play an important role in lipid binding. However, the precise functional domain of the C-terminal remains to be defined. In this study, apoA-I Nichinan, a naturally occurring human apoA-I variant with a deletion of glutamic acid 235, was expressed in Escherichia coli to examine the effect of this mutation on the functional domain of apoA-I for lipid binding and related consequences. A dimyristoyl phosphatidylcholine binding study with recombinant (r-) proapoA-I Nichinan showed a significantly slow initial rate of lipid binding. On preincubation with human plasma lipoprotein fractions (d<1.225 g/mL) at 37 degrees C for 1 hour, (125)I-labeled normal r-proapoA-I was chromatographed as a single peak at the high density lipoprotein (HDL) fraction, whereas (125)I-labeled r-proapoA-I Nichinan was chromatographed into the HDL fraction as well as the free r-proapoA-I fraction (23% of radioactivity). Circular dichroism measurements showed that the alpha-helix content of lipid-bound r-proapoA-I Nichinan was reduced, being 62% (versus 73%) of normal r-proapoA-I. Nondenaturing gradient gel electrophoresis of reconstituted HDL particles assembled with r-proapoA-I Nichinan and normal r-proapoA-I showed similar particle size. To study cholesterol efflux, human skin fibroblasts were labeled with [(3)H]cholesterol, followed by incubation with either lipid-free r-proapoA-I or DMPC/r-proapoA-I complex. Fractional cholesterol efflux from [(3)H]cholesterol-labeled fibroblasts to lipid-free r-proapoA-I Nichinan or DMPC/r-proapoA-I Nichinan complexes was significantly reduced relative to that of normal r-proapoA-I or DMPC/r-proapoA-I during the 6-hour incubation. Binding assays of human skin fibroblasts by lipid-free r-proapoA-I showed that r-proapoA-I Nichinan was 32% less bound to fibroblasts than was normal r-proapoA-I. Our data demonstrate that the deletion of glutamic acid 235 at the C-terminus substantially reduces the lipid-binding properties of r-proapoA-I Nichinan, which may cause a reduction in its capacity to interact with plasma membranes as well as to promote cholesterol efflux from cultured fibroblasts.

Leaderless polypeptides efficiently extracted from whole cells by osmotic shock

Three molecular foldases, DsbA, DsbC, and rotamase (ppiA), exhibited the unusual property of accumulating in an osmotically sensitive cellular compartment of Escherichia coli when their signal sequences were precisely removed by mutation. A mammalian protein, interleukin-1 (IL-1) receptor antagonist, behaved in a similar fashion in E. coli when its native signal sequence was deleted. These leaderless mutants (but not two control proteins overexpressed in the same system) were quantitatively extractable from whole cells by a variety of methods generally employed in the recovery of periplasmic proteins. A series of biochemical and genetic experiments showed that (i) leaderless DsbA (but not the wild type) was retained in a nonperiplasmic location; (ii) beta-galactosidase fusions to leaderless DsbA (but not to the wild type) exhibited efficient alpha complementation; (iii) none of the leaderless mutant proteins were substantially associated with cell membranes, even when they were overexpressed in cells; and (iv) leaderless DsbA was not transported to an osmotically sensitive compartment via a secA- or ftsZ-dependent mechanism. The observation that these proteins transit to some privileged cellular location by a previously undescribed mechanism(s)--absent their normal mode of (signal sequence-dependent) translocation--was unexpected. DsbA, rotamase, and IL-1, whose tertiary structures are known, appear to be structurally unrelated proteins. Despite a lack of obvious homologies, these proteins apparently have a common mechanism for intracellular localization. As this (putative) bacterial mechanism efficiently recognizes proteins of mammalian origin, it must be well conserved across evolutionary boundaries.

Strategies for achieving high-level expression of genes in Escherichia coli

Progress in our understanding of several biological processes promises to broaden the usefulness of Escherichia coli as a tool for gene expression. There is an expanding choice of tightly regulated prokaryotic promoters suitable for achieving high-level gene expression. New host strains facilitate the formation of disulfide bonds in the reducing environment of the cytoplasm and offer higher protein yields by minimizing proteolytic degradation. Insights into the process of protein translocation across the bacterial membranes may eventually make it possible to achieve robust secretion of specific proteins into the culture medium. Studies involving molecular chaperones have shown that in specific cases, chaperones can be very effective for improved protein folding, solubility, and membrane transport. Negative results derived from such studies are also instructive in formulating different strategies. The remarkable increase in the availability of fusion partners offers a wide range of tools for improved protein folding, solubility, protection from proteases, yield, and secretion into the culture medium, as well as for detection and purification of recombinant proteins. Codon usage is known to present a potential impediment to high-level gene expression in E. coli. Although we still do not understand all the rules governing this phenomenon, it is apparent that "rare" codons, depending on their frequency and context, can have an adverse effect on protein levels. Usually, this problem can be alleviated by modification of the relevant codons or by coexpression of the cognate tRNA genes. Finally, the elucidation of specific determinants of protein degradation, a plethora of protease-deficient host strains, and methods to stabilize proteins afford new strategies to minimize proteolytic susceptibility of recombinant proteins in E. coli.

Response of interleukin-6-deficient mice to tumor necrosis factor-induced metabolic changes and lethality

Whether interleukin (IL)-6 contributes to tumor necrosis factor (TNF)-induced lethal shock or whether, on the contrary, it is part of a protective feedback system, remains unresolved. Here, we report experiments with IL-6 gene-disrupted mice (IL-6(0/0)). We have tested the susceptibility of these to TNF-induced metabolic changes and lethality in different models, and compared the results with those obtained with IL-6+/+ wild-type mice. We studied the response to TNF in three different models: (i) murine TNF administration; (ii) TNF in galactosamine (GalN)-sensitized mice; (iii) TNF in Bacillus Calmette-Guérin-sensitized mice. We observed no significant difference between the two types of mice in any of the three models. Furthermore, IL-6(0/0) mice could be equally well desensitized (by IL-1) to TNF/GalN-induced lethality and tolerized to TNF-induced shock as IL-6+/+ mice. We also observed that, in response to turpentine, TNF or IL-1, IL-6(0/0) mice produced significantly less acute phase proteins (APP) than IL-6+/+ mice. In IL-6(0/0) mice, less corticosterone was induced by TNF than in the control mice, while the response to adrenocorticotropic hormone was the same. The results indicate that IL-6 is not contributing in a major way to the pathogenesis leading to TNF-induced shock, and that neither IL-6 nor the APP studied are essential for a protective feedback system.

Cytoplasmic and periplasmic production of human apolipoprotein E in Escherichia coli using natural and bacterial signal peptides

To understand the toxicity of high levels of heterologous human serum apolipoprotein E (ApoE) in Escherichia coli, as well as to prepare a system for producing the structural domains of this protein, plasmids were constructed in which the coding sequence of the N-terminal domain or all of ApoE followed E. coli or human apolipoprotein signal peptides (SP) or the N-terminal eleven amino acids (f10) of the gene 10-encoded protein of phage T7. High levels of production of the 22-kDa N-terminal domain (22K) of ApoE were obtained either as an f10::22K fusion protein, or using the natural SP, or SP derived from the periplasmic protein, alkaline phosphatase (PhoA), or from the outer membrane protein A (OmpA). Microsequencing showed that the SP of sPhoA::22K and sOmpA::22K, but not sApoE::22K, were correctly processed and, in the former cases, the protein could be released from the cells by osmotic shock. The extent of maturation of sPhoA::22K depended upon the host strain; with JM109, about 50% of the protein was not processed. Microsequencing of the f10::22K fusion protein, which could easily be purified following lysis of the cells, showed that the N-terminal methionine had been removed in agreement with the length parameter rule. Although considerable levels of the f10::ApoE fusion protein could be produced in the cytoplasm, production was markedly less using the PhoA signal peptide and the protein was not easily isolated following osmotic shock. The recombinant protein was biologically active after reconstitution with lipids in spite of the N-terminal modifications introduced.(ABSTRACT TRUNCATED AT 250 WORDS)

A thioredoxin gene fusion expression system that circumvents inclusion body formation in the E. coli cytoplasm

We have developed a versatile Escherichia coli expression system based on the use of E. coli thioredoxin (trxA) as a gene fusion partner. The broad utility of the system is illustrated by the production of a variety of mammalian cytokines and growth factors as thioredoxin fusion proteins. Although many of these cytokines previously have been produced in E. coli as insoluble aggregates or "inclusion bodies", we show here that as thioredoxin fusions they can be made in soluble forms that are biologically active. In general we find that linkage to thioredoxin dramatically increases the solubility of heterologous proteins synthesized in the E. coli cytoplasm, and that thioredoxin fusion proteins usually accumulate to high levels. Two additional properties of E. coli thioredoxin, its ability to be specifically released from the E. coli cytoplasm by osmotic shock or freeze/thaw treatments and its intrinsic thermal stability, are retained by some fusions and provide convenient purification steps. We also find that the active-site loop of E. coli thioredoxin can be used as a general site for small peptide insertions, allowing for the high level production of soluble peptides in the E. coli cytoplasm.

The expression of murine protein disulfide isomerase in Escherichia coli

Protein disulfide isomerase (PDI), a luminal enzyme of the endoplasmic reticulum (ER), is thought to be involved in the process that assures that the correct disulfide bonds form as a newly synthesized protein folds into its appropriate three-dimensional structure (Freeman, 1984). In recent years, the ER has been shown to have at least two additional, distinct PDI-related luminal proteins (Bennett et al., 1988; Mazzarella et al., 1990). As a potential first step toward an investigation of the structure and function of PDI and of the PDI-related proteins as well, we have developed a bacterial expression system in Escherichia coli capable of synthesizing significant levels of enzymatically active PDI under the control of the inducible tac promoter. We have observed that the use of this bacterial expression system is complicated by the fact that there is a significant amount of internal initiation of protein synthesis within the PDI coding sequence and the fact that all of the PDI-related expression products are found equally distributed between the cytoplasmic and periplasmic fractions due to a single peptide-independent mechanism. Our studies with this system have demonstrated that at least some truncated PDI molecules containing the carboxy-terminal most active site have significant PDI activity.

Abnormal fractionation of beta-lactamase in Escherichia coli: evidence for an interaction with the inner membrane in the absence of a leader peptide

beta-Lactamase with the -20 to -1 region of the leader peptide deleted (almost complete deletion of the leader peptide) [delta(-20,-1) beta-lactamase] was released from Escherichia coli cells by osmotic shock. Fractionation of the cells by conversion to spheroplasts and protease accessibility experiments further indicated that a portion of the protein may be bound to the cytoplasmic membrane and be partially exposed in the periplasmic space. Expression of delta(-20,-1) beta-lactamase conferred a 25-fold increase in the 50% lethal dose for ampicillin relative to that for controls, thus confirming that a small amount (about 2%) of the active protein is completely exported from the cytoplasm. These results suggest that even in the absence of a leader peptide, mature beta-lactamase is able to interact with the cytoplasmic membrane and be translocated into the periplasmic space, albeit with a low efficiency.

Expression of recombinant human apolipoprotein A-I in Chinese hamster ovary cells and Escherichia coli

Human recombinant apolipoprotein (apo) A-I was produced by Chinese hamster ovary (CHO) cells and Escherichia coli with expression vectors containing cDNAs encoding preproapoA-I or apoA-I, respectively. The apoA-I from CHO cells was purified from the culture medium by ammonium sulfate precipitation, phenyl-Sepharose chromatography, and affinity purification on anti-apoA-I immunoabsorber. Human apoA-I was produced in E. coli as a fusion protein with glutathione S-transferase. A four amino acid linker, which separated the two proteins, was specifically recognized and cut by Factor Xa. The purification was accomplished by chromatography of E. coli extracts on glutathione-Sepharose and digestion with Factor Xa. The highest production level was found to be 0.5 micrograms/ml of culture medium per 48 h for a clone of stable transformant of CHO cells, whereas E. coli could produce as much as 20 micrograms/ml of bacterial culture. These apoA-I forms were compared in terms of molecular weight, isoelectric point, and expression of several epitopes. Recombinant apoA-I obtained from CHO cells appears intact and its isoelectric point is compatible with that of the mature form and the proform of apoA-I, whereas a part of the apoA-I produced by E. coli does not contain the COOH-terminus. Also, two of six epitopes are expressed to a greater extent in apoA-I obtained from E. coli than in apoA-I obtained from human plasma.