Search

Production of recombinant antibody fragments in bacterial expression systems results in intentional or fortuitous differences compared to the original products prepared by hybridoma technology. These differences may have significant effects not only on antigen-binding properties but also on pharmacokinetic and tumor-seeking properties. Our major goal was to investigate some of these possible differences. We produced in Escherichia coli an rFab′ fragment containing only 1 cysteine residue in the hinge region; the fragment was derived from a mouse MAb (F6) specific for CEA. The rFab′ had a slightly lower m.w. and a higher isoelectric point relative to the corresponding nonrecombinant fragment (pFab′). This was explained by the absence of N-glycosylation on the Vκ domain of rFab′. Vκ glycosylation had no significant effect on antibody-binding affinity and kinetics. However, rFab′ was eliminated from the circulation much faster than pFab′, and the maximal dose accumulated in the tumor was reduced relative to pFab′. Thus, glycosylation appears to modify the targeting efficiency of antibody fragments. rF(ab′)2 fragments were obtained either spontaneously from the culture supernatant of E. coli or by chemical cross-linking [rcF(ab′)2]. We observed improved tumor targeting with rcF(ab′)2 compared to rF(ab′)2, which could be explained by the greater stability of the thioether compared to the disulfide linkage. These results demonstrate that a single cysteine residue in the hinge region of rFab′ is particularly well suited to prepare stable, chemically coupled, bivalent or bispecific antibodies, avoiding intrahinge disulfide bonding and thus achieving higher production yields. © 2002 Wiley-Liss, Inc.

To obtain an expression vector that will optimize secretion of proteins with disulfide bridges in Escherichia coli, we fused the phoA gene, encoding the bacterial alkaline phosphatase (PhoA), to the sequence encoding the pectate lyase B signal sequence (PelBSS). We used an extensively degenerate pelBSS with silent mutations to study their effects on the production level and activity of PhoA. 11 representative clones differed by a factor of five between the lowest and the highest level of activity, and by a factor greater than seven for the production levels. The efficiency of translocation seems to be the result of an equilibrium between production and secretion levels that favours the secretion of active PhoA according to the competence of the fusion protein being translocated. Free energy calculations and the predicted mRNA secondary structures of the translation initiation regions showed that the high stability of the secondary structure decreased production and secretion levels of PhoA and vice versa. A stem-loop encompassing the degenerate positions downstream from the AUG start codon appears to be responsible for the differences in the production levels.

International audience; Pore-forming colicins are soluble bacteriocins which form voltage-gated ion channels in the inner membrane of Escherichia coli. To reach their target, these colicins first bind to a receptor located on the outer membrane and then are translocated through the envelope. Colicins are subdivided into two groups according to the envelope proteins involved in their translocation: group A colicins use the Tol proteins; group B colicins use the proteins TonB, ExbB, and ExbD. We have previously shown that a double-cysteine colicin A mutant which possesses a disulfide bond in its pore-forming domain is translocated through the envelope but is unable to form a channel in the inner membrane (D. Duché, D. Baty, M. Chartier, and L. Letellier, J. Biol. Chem. 269:24820-24825, 1994). Measurements of colicin-induced K+ efflux reveal that preincubation of the cells with the double-cysteine mutant prevents binding of colicins of group A but not of group B. Moreover, we show that the mutant is still in contact with its receptor and import machinery when it interacts with the inner membrane. From these competition experiments, we conclude that each Escherichia coli cell contains approximately 400 and 1,000 colicin A receptors and translocation sites, respectively.

The use of the colicin A lysis protein to direct the extracellular release of a fusion protein from Escherichia coli was investigated as an approach for the preparation of recombinant animal toxins. Apamin, a bee venom neurotoxin, was used as the model toxin. It is reticulated by two disulfide bridges and interacts with small conductance Ca(2+)-activated K+ channels. Substantial amounts of free recombinant apamin were obtained by CNBr cleavage of the fusion protein [col-(1-171)-apa] and HPLC purification. It was recognized by conformation-dependent monoclonal antibodies with a K0.5 value close to that for natural apamin, indicating that folding was correct. In toxicity and binding experiments, the recombinant apamin displayed low activity. The recombinant and natural molecules differed by the amidation of the C-terminal histidine residue. Previous structure/activity relationship studies do not implicate this C-terminal residue in activity but the role of its amidation was not investigated. An apamin analog with a non-amidated C-terminal residue was then chemically synthesized. The biological properties of both recombinant and chemical molecules were determined. Amidation of the C-terminal alpha-carboxyl of apamin appears to be essential for full expression of its biological activity.

Two monoclonal antibodies (mAbs), 5A4 and 6D6, directed against cortisol, have been obtained; 6D6 is used in an assay kit for cortisol. The antibodies also recognize other, structurally related steroids present in the sample assayed. To improve the specificity of the assay, we aimed to minimize the recognition of non-cortisol steroids by the two anti-cortisol mAbs. Our strategy consisted in constructing an efficient expression vector in E. coli which produced the single-chain variable fragment (scFv) of the mAbs in the periplasmic space. We demonstrated that temperature and inducer concentration of the bacterial culture influenced dramatically the yield of active scFv. From the nucleotide sequence we constructed a three-dimensional model of the two variable fragments in order to understand why related steroids are, or are not recognized by the antibody. For both antibodies, we have identified chemical groups which are probably involved in the binding of the steroid haptens and the antibodies. The hydrophobic pocket formed by the antibody comprises two or three tryptophan residues which can interact with the steroid nucleus by stacking. The serine at position 35 of the heavy chain is buried in the back of the pocket and can form a hydrogen bond with the 20-keto group of the cortisol. The stacking interactions and the hydrogen bond orient the steroid in the pocket. This reactivity of the binding site is sustained by the analysis of the cross-reactions of related steroids with the mAbs.

International audience; Mutants of colicin A have been prepared in which the three tryptophan residues (Trp86, Trp130 and Trp140), localized in the C-terminal domain of the soluble wild-type protein, have been substituted by phenylalanine. The Trp140Phe single mutation had the effect of decreasing the percentage of protein that is expressed as insoluble aggregates. The created hydrophobic cavity decreased the stability of the protein during its folding, resulting in partial aggregation in the cytoplasm of the Escherichia coli-producing cells. Aggregation was increased when Trp140 was substituted by Lys, Leu or Cys, or if the Trp140 mutation was combined with the Trp86Phe and/or Trp130Phe mutations. A single mutation, Lys113Phe, however, was able to restore the solubility of the aggregated mutants in vivo. Detailed analysis of the 3-D structure of the C-terminal domain of colicin A suggests that filling of the hydrophobic cavity is responsible for this effect.

International audience; Antibody microarrays are among the novel class of rapidly emerging proteomic technologies that will allow us to efficiently perform specific diagnoses and proteomic analysis. Recombinant antibody fragments are especially suited for this approach but their stability is often a limiting factor. Camelids produce functional antibodies devoid of light chains (HCAbs) of which the single N-terminal domain is fully capable of antigen binding. When produced as an independent domain, these so-called single domain antibody fragments (sdAbs) have several advantages for biotechnological applications thanks to their unique properties of size (15 kDa), stability, solubility, and expression yield. These features should allow sdAbs to outperform other antibody formats in a number of applications, notably as capture molecules for antibody arrays. In this study, we have produced antibody microarrays using direct and oriented immobilization of sdAbs, produced in crude bacterial lysates, to generate a proof-of-principle of a high-throughput compatible array design. Several sdAb immobilization strategies have been explored. Immobilization of in vivo biotinylated sdAbs by direct spotting of bacterial lysate on streptavidin and sandwich detection was developed to achieve high sensitivity and specificity, whereas immobilization of "multi-tagged" sdAbs via anti-tag antibodies and a direct labeled sample detection strategy was optimized for the design of high-density antibody arrays for high-throughput proteomics and identification of potential biomarkers.

A construction was carried out to obtain a high level of expression in Escherichia coli of the gene celCCA, coding for the endoglucanase A from Clostridium cellulolyticum (EGCCA). The enzyme was purified in two forms with different molecular weights, 51,000 and 44,000. The smaller protein was probably the result of proteolysis, although great care was taken to prevent this process from occurring. Evidence was found for the loss of the conserved reiterated domains which are characteristic of C. thermocellum and C. cellulolyticum cellulases. The two forms were extensively studied, and it was demonstrated that although they had the same pH and temperature optima, they differed in their catalytic properties. The truncated protein gave the more efficient catalytic parameters on carboxymethyl cellulose and showed improved endoglucanase characteristics, whereas the intact enzyme showed truer cellulase characteristics. The possible role of clostridial reiterated domains in the hydrolytic activity toward crystalline cellulose is discussed.

DNA regions encoding the various domains of a protein can be expressed as separate entities by inserting at appropriate sites a 'STOP-Shine-Dalgarno-sequence-ATG' cassette encoding a termination codon, a Shine-Dalgano sequence and an initiation codon within the structural gene. This technique has been used to obtain a 137-amino-acid-residue pore-forming protein designated DA70C comprising the final 136-amino-acid-residue COOH-terminal of colicin A preceded by an NH2-terminal methionine. Da70C was correctly expressed but poorly released to the extracellular medium. Its purification involved, as a final step, a partition in Triton X-114 thus demonstrating that hydrophobic regions are exposed in this protein. The ability of DA70C to form ion channels in planar lipid bilayers was investigated and pore properties were analyzed. The results indicate that helices 1-3 of the 204-amino-acid-residue colicin pore-forming domain (containing 10 alpha-helices) are not involved in ion conduction through the channel. However, they are important in maintaining the stability of the soluble state of the COOH-terminal domain.

International audience; Pore-forming colicins exert their lethal effect on E coli through formation of a voltage-dependent channel in the inner (cytoplasmic-membrane) thus destroying the energy potential of sensitive cells. Their mode of action appears to involve 3 steps: i) binding to a specific receptor located in the outer membrane; ii) translocation across this membrane; iii) insertion into the inner membrane. Colicin A has been used as a prototype of pore-forming colicins. In this review, the 3 functional domains of colicin A respectively involved in receptor binding, translocation and pore formation, are defined. The components of sensitive cells implicated in colicin uptake and their interactions with the various colicin A domains are described. The 3-dimensional structure of the pore-forming domain of colicin A has been determined recently. This structure suggests a model of insertion into the cytoplasmic membrane which is supported by model membrane studies. The role of the membrane potential in channel functioning is also discussed.

Single-domain antibodies specific to methotrexate (MTX) were obtained after immunization of one llama (Llama glama). Specific VHH domains (V-D-J-REGION) were selected by panning from an immune-llama library using phage display technology. The antibody fragments specific to MTX were purified from Escherichia coli (C41 strain) periplasm by immobilized metal affinity chromatography with an expression level of around 10mg/L. A single band around 16,000Da corresponding to VHH fragments was found after analysis by SDS-PAGE and Western blotting, while competition ELISA demonstrated selective binding to soluble MTX. Surface plasmon resonance (SPR) analysis showed that anti-MTX VHH domains had affinities in the nanomolar range (29-515nM) to MTX-serum albumin conjugates. The genes encoding anti-MTX VHH were found by IMGT/V-QUEST to be similar to the previously reported llama and human IGHV germline genes. The V-D and D-J junction rearrangements in the seven anti-MTX CDR3 sequences indicate that they were originated from three distinct progenitor B cells. Our results demonstrate that camelid single-domain antibodies are capable of high affinity binding to low molecular weight hydrosoluble haptens. Furthermore, these anti-MTX VHH give new insights on how the antigen binding repertoire of llama single-domain antibody can provide combining sites to haptens in the absence of a VL. This type of single-domain antibodies offers advantages compared to murine recombinant antibodies in terms of production rate and sequence similarity to the human IGHV3 subgroup genes.

Production of recombinant antibody fragments in bacterial expression systems results in intentional or fortuitous differences compared to the original products prepared by hybridoma technology. These differences may have significant effects not only on antigen-binding properties but also on pharmacokinetic and tumor-seeking properties. Our major goal was to investigate some of these possible differences. We produced in Escherichia coli an rFab' fragment containing only 1 cysteine residue in the hinge region; the fragment was derived from a mouse MAb (F6) specific for CEA. The rFab' had a slightly lower m.w. and a higher isoelectric point relative to the corresponding nonrecombinant fragment (pFab'). This was explained by the absence of N-glycosylation on the V kappa domain of rFab'. V kappa glycosylation had no significant effect on antibody-binding affinity and kinetics. However, rFab' was eliminated from the circulation much faster than pFab', and the maximal dose accumulated in the tumor was reduced relative to pFab'. Thus, glycosylation appears to modify the targeting efficiency of antibody fragments. rF(ab')(2) fragments were obtained either spontaneously from the culture supernatant of E. coli or by chemical cross-linking [rcF(ab')(2)]. We observed improved tumor targeting with rcF(ab')(2) compared to rF(ab')(2), which could be explained by the greater stability of the thioether compared to the disulfide linkage. These results demonstrate that a single cysteine residue in the hinge region of rFab' is particularly well suited to prepare stable, chemically coupled, bivalent or bispecific antibodies, avoiding intrahinge disulfide bonding and thus achieving higher production yields.

The colicin A pore-forming domain (pfColA) was fused to a bacterial signal peptide (sp-pfColA). This was inserted into the Escherichia coli inner membrane in functional form and could be coimmunoprecipitated with epitope-tagged immunity protein (EpCai). We constructed a series of fusion proteins in which various numbers of sp-pfColA α-helices were fused to alkaline phosphatase (AP). We showed that a fusion protein made up of the hydrophobic α-helices 8 and 9 of sp-pfColA fused to AP was specifically coimmunoprecipitated with EpCai produced in the same cells. This is the first biochemical evidence that Cai recognizes and interacts with the colicin A hydrophobic helical hairpin.

Intracellularly expressed antibodies have been designed to bind and inactivate target molecules inside eukaryotic cells. Here we report that an antibody fragment can be used to probe the periplasmic localization of the colicin A N-terminal domain. Colicins form voltage-gated ion channels in the inner membrane of Escherichia coli . To reach their target, they bind to a receptor located on the outer membrane and then are translocated through the envelope. The N-terminal domain of colicins is involved in the translocation step and therefore is thought to interact with proteins of the translocation system. To compete with this system, a single-chain variable fragment (scFv) directed against the N-terminal domain of the colicin A was synthesized and exported into the periplasmic space of E. coli . The periplasmic scFv inhibited the lethal activity of colicin A and had no effect on the lethal activity of other colicins. Moreover, the scFv was able to specifically inactivate hybrid colicins possessing the colicin A N-terminal domain without affecting their receptor binding. Hence, the periplasmic scFv prevents the translocation of colicin A and probably its interaction with import machinery. This indicates that the N-terminal domain of the toxin is accessible in the periplasm. Moreover, we show that production of antibody fragments to interfere with a biological function can be applied to prokaryotic systems.

The 6D6 anti-cortisol scFv was prepared as fusion protein with maltose-binding protein (MBP) to increase the amount of soluble product. This fusion was almost completely insoluble when produced in a wild-type strain of Escherichia coli. However, when MBP-scFv fusion was produced in a tolR leaky strain, it was secreted into the culture medium as an active, soluble protein. Production of recombinant proteins in the tolR strain greatly enhances the recovery of active protein and may be a useful system to produce MBP fusion proteins that would normally aggregate when produced in wild-type bacterial strains. © 1997 Federation of European Biochemical Societies

International audience; Four colicin A double-cysteine mutants possessing a disulfide bond in their pore-forming domain were constructed to study the translocation and the pore formation of colicin A. The disulfide bonds connected alpha-helices 1 and 2, 2 and 10, 3 and 9, or 3 and 10 of the pore-forming domain. The disulfide bonds did not prevent the colicin A translocation through the Escherichia coli envelope. However, the mutated colicins were able to exert their in vivo channel activity only after reduction of their disulfide bonds. In vitro studies with brominated phospholipid vesicles and planar lipid bilayers revealed that the disulfide bond that connects the alpha-helices 2 and 10 prevented the colicin A membrane insertion, whereas the other double-cysteine mutants inserted into lipid vesicles. The disulfide bonds that connect either the alpha-helices 1 and 2 or 3 and 10 were unable to prevent the formation of a conducting channel in presence of membrane potential. These results indicate that alpha-helices 1, 2, 3, and 10 remain at the membrane surface after application of a membrane potential.

International audience; Four disulfide bonds were engineered into the pore-forming domain of colicin A to probe the conformational changes associated with its membrane insertion and channel formation. The soluble pore-forming domain consists of 10 alpha-helices with two outer layers (helices 1, 2, and 3-7, respectively) sandwiching a middle layer of three helices (8-10). Helices 8 and 9 form a hairpin which is completely buried and consists of hydrophobic and neutral residues only. This helical hairpin has been hypothesized to be the membrane anchor. Each double-cysteine mutant possessing an individual disulfide bond, cross-linking either helices 1 to 9 (H1/H9), 5 to 6 (H5/H6), 7 to 8 (H7/H8), or 9 to 10 (H9/H10), respectively, is unable to promote K+ efflux from sensitive Escherichia coli cells. Activity can be restored by addition of a reducing agent. In vitro studies with brominated lipid vesicles and planar lipid bilayers show that the disulfide bond which connects the helices 1 to 9 prevents colicin A membrane insertion, whereas the other disulfide bond mutants insert readily into lipid vesicles. All of the engineered bridges prevented the formation of a conducting channel in the presence of a membrane potential. This novel approach indicates that membrane insertion and channel formation are two separate steps. Moreover, the effects of the distance constraints introduced by the different disulfide bonds on colicin A activity indicate that the helical pair 1 and 2 moves away from the other helices upon membrane insertion. Helices 3-10 remain associated together. As a consequence, the results imply that the helical hairpin lies parallel to the membrane surface. In contrast, induction of the colicin channel by the membrane potential requires a profound reorganization of the helices association. These results are discussed in light of several proposed models of the membrane-bound colicin and channel structures.

International audience; The bacterial toxin colicin A binds spontaneously to the surfaces of negatively charged membranes. The surface-bound toxin must subsequently, however, become an acidic 'molten globule' before it can fully insert into the lipid bilayer. Clearly, electrostatic interactions must play a significant role in both events. The electrostatic field around the toxin in solution was calculated using the finite-difference Poisson-Boltzmann method of the Delphi programme and the known X-ray structure. A large positively charged surface was identified which could be involved in the binding of colicin to negatively charged membranes. The applicability of the result was tested by also calculating the fields around modelled structures of the closely related colicins B and N. Surprisingly, colicin N showed a similar charge distribution in spite of its isoelectric point of pI 10.20 (colicin A has pI 5.44). One reason for this is the strong conservation of certain negative charges in all colicins. There is a single highly conserved aspartate residue (Asp78) on the positively charged face which provides a small but discrete region of negative charge. This residue, Asp78, was replaced by asparagine in the mutant D78N. D78N binds faster to negatively charged vesicles but inserts only half as fast as the wild-type protein into the membrane core. This indicates that, first, the initial membrane binding has a significant electrostatic component and, second, that the isolated charge on Asp78 plays a role in the formation of the insertion intermediate.

International audience; Three double cysteine mutants, each possessing a disulfide bond in its pore-forming domain, were used to study the translocation of colicin A through the Escherichia coli envelope. These mutated colicins were able to exert their in vivo channel activity only after their disulfide bonds had been reduced by dithiothreitol. In solution, the reduction of the disulfide bonds by dithiothreitol was a slow process whose kinetics depended on the position of the disulfide bond (t1/2 varying between 35 and 100 s). This t1/2 was strongly decreased (t1/2 = 8-9 s) upon predenaturation of the mutated colicins with urea. The t1/2 values of reduction of the mutants bound to E. coli-sensitive cells were similar to those of predenatured colicins. This suggested that the interaction of the oxidized double cysteine mutants with the E. coli envelope triggered their unfolding. The disulfide bonds did not prevent but delayed the translocation of the colicins. The amplitude of the delay and the time at which it occurred during translocation depended on the position of the disulfide bond. We could discriminate between the delays accumulated during binding to the receptor and those during the translocation via OmpF and the Tol proteins.

The colicin A lysis protein (Cal) was used to direct the extracellular release of recombinant proteins produced in Escherichia coli. The cal gene, under the control of its inducible promoter, was introduced into an expression vector encoding the human growth hormone devoid of its signal sequence (Met-hGH). Cal and Met-hGH were simultaneously expressed at two different levels of Met-hGH induction. The results indicate that Cal causes the excretion of non-aggregated Met-hGH from the cytoplasm to the culture medium and that the Met-hGH is correctly folded since the released Met-hGH is antigenically indistinguishable from the authentic mature hGH and is biologically active in binding to specific receptor sites.

International audience; All bacterial organisms have an envelope which has the primary role to constitute a physical barrier between the cytoplasm and the extracellular medium thereby protecting cells from harmful compounds from this medium. However, exchanges with the latter are required since cells must take up nutrients useful for growth.

International audience; Colicins A and E1 are two pore-forming colicins sharing homology in their C-terminal domains but not in their N-terminal or central domains. Using site-directed mutagenesis, restriction sites were inserted at the proper locations to allow recombination of these domains. Six different constructs were obtained. All these proteins were expressed in Escherichia coli and properly recognized by monoclonal antibodies directed against epitopes located in different domains of colicin A. Out of the six hybrids, only two were released to the extracellular medium. Immunocytolocalization indicated that some of the hybrids aggregated within the cytoplasm. With some hybrids, the defect in release was related to a defect in synthesis of the lysis protein that normally promotes release.

International audience; Six different hybrid colicins were constructed by recombining various domains of the two pore-forming colicins A and E1. These hybrid colicins were purified and their properties were studied. All of them were active against sensitive cells, although to varying degrees. From the results, one can conclude that: (1) the binding site of OmpF is located in the N-terminal domain of colicin A; (2) the OmpF, TolB and TolR dependence for translocation is also located in this domain; (3) the TolC dependence for colicin E1 is located in the N-terminal domain of colicin E1; (4) the 183 N-terminal amino acid residues of colicin E1 are sufficient to promote E1AA uptake and thus probably colicin E1 uptake; (5) there is an interaction between the central domain and C-terminal domain of colicin A; (6) the individual functioning of different domains in various hybrids suggests that domain interactions can be reconstituted in hybrids that are fully active, whereas in others that are much less active, non-proper domain interactions may interfere with translocation; (7) there is a specific recognition of the C-terminal domains of colicin A and colicin E1 by their respective immunity proteins.

A plasmid (pColAF1), derived from pColA, and lacking the region encoding Cai (colicin A immunity protein) and Cal (colicin A lysis protein) has been constructed. The strains carrying pColAF1 produce normal amounts of colicin A which remains in the cell cytoplasm and does not result in loss of viability. Similar results have also been obtained for transposon insertion mutants lacking Cai. Structure prediction analysis indicates that four peptide regions of Cai might span the cytoplasmic membrane. Since the NH2- and COOH-terminal regions are charged, this analysis suggests a topology of the 178 residues polypeptide chain in which regions 38 to 70 and 124 to 143 might be exposed at the outer side of the cytoplasmic membrane. With mutants constructed using recombinant DNA techniques, we could demonstrate that the removal of a 30 residue COOH-terminal region, and mutations altering the surface exposed loop comprised of aminoacid residues 124-143 abolish the protecting function of Cai.

We have studied the synthesis, processing and export of human growth-hormone-releasing factor (hGRF) in Escherichia coli transformed with a plasmid constructed for the expression of hGRF as a hybrid protein. A DNA fragment containing the entire sequence of phosphate-binding protein gene (phoS) is fused to a modified hGRF-coding sequence (phoS-mhGRF). The hybrid protein, PhoS-mhGRF, was recovered in the supernatant fluid after spheroplasting treatment indicating correct export to the periplasmic space. Pulse-chase experiments demonstrated that the hybrid protein was similarly processed as the PhoS precursor.

Presented here is an experimental demonstration of our theoretical predictions on the role of the downstream context of unfavourable codons in a gene on its expression level. Six non clustered AGG codons were inserted in the chloramphenicol acetyltransferase (cat) gene of E. coli and the expression of this modified gene (cat4) was compared with that of a cat gene in which four clustered AGG codons were inserted (cat2 gene). As predicted, the rate of production of the corresponding CAT4 and CAT2 proteins is equal as long as the rate of transcription of the gene does not exceed a given limit. When this limit is exceeded, production of CAT4 continues to increase, whereas CAT2 production decreases dramatically. Various consequences and possible applications of this downstream context effect are discussed.

Ovalbumin is secreted by the tubular gland cells without cleavage of a signal sequence at the N-terminus. In Escherichia coli strains which produce a chicken ovalbumin-like protein (OLP) from a plasmid-cloned gene, the OLP is synthesized on membrane-bound polysomes and secreted without cleavage into the periplasmic space. In contrast, a deleted protein, which lacks 126 amino acids in the N-terminal half, is not secreted and is synthesized from free polysomes. Our results are compatible with the presence, in the N-terminal half of the molecule, of a signal sequence necessary for the transport across the membrane.

International audience; Colicins A, E1, E2 and E3 belong to the BtuB group of colicins. The NH2-terminal region of colicin A is required for translocation, and defects in this region cannot be overcome by osmotic shock of sensitive cells. In addition to BtuB, colicin A requires OmpF for efficient uptake by sensitive cells. The roles of BtuB and OmpF in translocation and binding to the receptor of the colicins A, E1, E2 and E3 were compared. The results suggest that for colicin A OmpF is used both as a receptor and for translocation across the outer membrane. In contrast, for colicin E1, OmpF is used neither as a receptor nor for translocation. For colicins E2 and E3, the situation is intermediate: only BtuB is used as a receptor but both BtuB and OmpF are involved in the translocation step.

The immunity protein to colicin A (Cai), which is constitutively expressed at a very low level in Escherichia coli strains, has been studied in recombinant plasmid constructs allowing expression of various immunity fusion proteins under the control of inducible promoters. The 13-amino acid NH2-terminal region of Cai was substituted by polypeptides from beta-galactosidase or from colicin A. Upon induction of the chimeric proteins, the rate of expression of the immunity protein could be correlated to the level of resistance to colicin A. The immunity protein has been "tagged" with an epitope from the colicin A protein for which a monoclonal antibody is available. Using this technique, we have directly demonstrated that the immunity protein is located in the cytoplasmic membrane. The results indicate that the NH2-terminal region of Cai is directed toward the cytoplasm and is probably not required for Cai insertion into the membrane or for its function.

DNA constructs coding for human growth hormone (hGH)-releasing factor (hGRF) preceded by the specific recognition sequence for the activated blood coagulation factor X (FXa), fused in frame to the N-terminal 172-amino acid residues of colicin A, have been expressed in Escherichia coli. The construct was placed under the control of the inducible caa promoter in an operon containing a downstream gene coding for the cell lysis protein, Cal. Induction resulted in excretion of only the processed colicin A fragment. Replacement of Cal by the terminator from phage fd resulted in high expression of the hybrid protein, which was recovered as cytoplasmic aggregates. Enzymatic cleavage of the purified and renatured hybrid protein using FXa allowed the recovery of authentic hGRF.

Treatment of Escherichia coli cells with procaine (0.55%, w/v) results in the accumulation of precursor in addition to mature forms of two periplasmic proteins, alkaline phosphatase and glutamine-binding protein. The precursor form of alkaline phosphatase has a higher molecular weight than the mature form by about 2600. An experimental technique is described to isolate and purify precursor forms of any presumably exported protein. After the membrane solubilization step in the presence of nonionic detergent, a peptidase is stimulated, resulting in partial cleavage of the precursors. The products of this cleavage have been identified as the mature protein and presumably the signal peptide in the case of alkaline phosphatase. The amino acid composition of this peptide, which is comprised of 25 residues, has been determined. Procaine (0.55%, w/v) causes an increase in molecular packing of lipid molecules in the membrane which might result in an alteration of membrane fluidity sufficient for selective inhibition of processing of precursors of exported proteins.

A large number of mutants introducing point mutations and deletions into the COOH-terminal domain of colicin A have been constructed by using site-directed mutagenesis. The COOH-terminal domain carries the channel activity. The effects of the alterations in the polypeptide chain on the secretion of colicin A by colicinogenic cells have been investigated. All deletions and some mutations were found to lead to protein aggregation in the cytoplasm, thereby preventing release into the medium. The mutated colicin A proteins have been purified, and their activity in vivo (on sensitive cells) and in vitro (in planar lipid bilayers) has been assayed. Deletions in the region containing putative helices 4, 5, and 6 (predicted to be involved in pore formation) and the transitions (Ala----Asp-492, Phe----Pro-493) in helix 4 abolished the activity. No correlation was observed between mutations leading to protein aggregation and those leading to loss of channel activity. Some mutations were found to alter characteristic properties of the single channels, such as stability, current-relaxation kinetics, voltage dependence, and pore conductance. Site-directed mutagenesis provides a powerful tool for studying structure-function relationships of voltage-sensitive ionic channels.

The ability of colicin A, a bacteriocin produced by some Enterobacteriaceae, to fuse phospholipid vesicles at acidic pH, was demonstrated by electron microscopy and resonance energy transfer. The fusion depends on protein concentration and on the nature of the phospholipids. Vesicles, prepared from Escherichia coli phospholipids, fused one or more rounds at pH 4.5 upon addition of stoichiometric amounts of colicin A. Fusion was not only induced by pore-forming colicins (E1, K) but also by colicins that contain nuclease activities (E2, E3). By recombinant DNA technology it is shown that the first glycine-rich 70 NH2-terminal amino acids and, most probably, the extreme COOH-terminal end of colicin A are involved in the fusion activity of the protein. The physiological relevance of this property of colicins is discussed.

An antibody was raised against the precursor form of alkaline phosphatase. The antiserum was depleted of the immunoglobulin G (IgG) directed against the mature alkaline phosphatase. A pure anti-(signal peptide) antibody was then obtained by affinity chromatography using a column of Sepharose linked to the signal peptide by the C-terminal portion. As expected from the size of the antigen, this antibody did not precipitate with the peptide suggesting that this peptide contains only one antigenic determinant. Thus, a double immunoprecipitation system, with an anti-rabbit IgG raised in goat, was used. In this system with the mature alkaline phosphatase very little precipitation occurred, showing that the antibody isolated was mostly directed against the signal peptide. Not only the signal peptide from the precursor form of the alkaline phosphatase but also that from the precursor form of another periplasmic protein: the leucyl-isoleucyl-valine binding protein, was recognized by the antibody. This suggests that conformational homologies exist between the signal peptides of precursors of periplasmic proteins. Aminopeptidase N, a protein located on the outer face of the cytoplasmic membrane, was precipitated by the anti-(signal peptide) antibody in the double immunoprecipitation system. This peripheral membrane protein could also inhibit the precipitation of the signal peptide. Because the signal sequences usually show no consistent homology, we propose that a conformation of signal peptides is recognized by the antibody isolated. This hypothesis is discussed.

Summary The topology of the integral membrane protein Cai (colicin A immunity protein), which is required to protect producing cells from the pore-forming colicin A, was analysed using fusions to alkaline phosphatase. The properties of these fusion proteins support the model for Cai topology previously proposed on theoretical grounds. The protein was found to contain four transmembrane sequences and its N-and C-terminal regions were found to be directed towards the cytoplasm. Oligonucleotide-directed mutagenesis and sequence comparisons between Cai, Cbi (colicin B immunity protein), and Cni (colicin N immunity protein) were carried out to determine the functional regions of Cai. The possible roles of the various regions of Cai in its protective function and in its topological organization are discussed.

Colicins are antibiotic proteins that kill sensitive Escherichia coli cells. The structure of the pore-forming fragment of colicin A has been solved to 2.5 A resolution using the techniques of X-ray crystallography and genetic engineering. Site-directed mutagenesis was used to construct a number of cysteine-containing mutant proteins, one of which yielded an excellent mercurial derivative. Our experiences suggest strategies for obtaining useful heavy-atom derivatives for protein crystallography using genetic engineering techniques.

In E. coli cells grown in the presence of procaine (0.55% w/v), precursor forms of alkaline phosphatase and of glutamine binding protein accumulate besides mature forms synthesized prior to procaine addition. An experimental technique, of general application, for isolation and purification of mature and precursor forms obtained under these conditions, is described.

The possible involvement of topogenic export sequences within the colicin A polypeptide chain has been investigated. Different constructs have been made using various techniques to introduce deletions in the central and NH2-terminal regions of colicin A. Together, these deletions span the region from amino acid 15 to the end of the protein. None of these regions was found to be required for extracellular release or had any effect on the efficiency of this process. By inserting a termination codon, a Shine-Dalgarno sequence and an initiation codon into the gene for colicin A, the NH2-terminal and central plus COOH-terminal domains could be demonstrated to be released to the same extent when produced as separate polypeptides as when produced as linked ones. The introduction into the COOH-terminal domain of mutations promoting cytoplasmic aggregation had no effect on the secretion of the NH2-terminal polypeptide. These results demonstrated that no specific interaction between the NH2- and COOH-terminal regions of the colicin A polypeptide chain is involved in the release of colicin A. We are led to conclude that there is no topogenic export signal in the polypeptide chain of colicin A involved in the release mechanism. Thus the process is non-specific with respect to the colicin itself and depends solely on the expression of the colicin A lysis protein (Cavard et al., 1985, 1987). The expression of the protein causes the release of not only the colicin but also many other cellular proteins, including beta-lactamase, EF-Tu, and chloramphenicol acetyltransferase.

A large number of mutations which introduce deletions in colicin A have been constructed. The partially deleted colicin A proteins were purified and their activity in vivo (on sensitive cells) and in vitro (in planar lipid bilayers) was assayed. The receptor-binding properties of each protein were also analysed. From these results, we suggest that the NH2-terminal region of colicin A (residues 1 to 172) is involved in the translocation step through the outer membrane. The central region of colicin A (residues 173 to 336) contains the receptor-binding domain. The COOH-terminal domain (residues 389 to 592) carries the pore-forming activity.

We have analysed by S1 nuclease mapping the in vivo termination sites of transcription of the caa-cal operon and cai gene. The termination region for caa mRNA (T1A terminator) features characteristics of a rho-independent terminator. This terminator is a convergent transcription terminator, its complementary secondary structure being present at the 3'-end of cai mRNA. The caa-cal mRNA terminator (T2A terminator) has a stable potential secondary structure and shows homology with rho-dependent terminators. In vitro transcription of caa-cal operon demonstrated that the two terminators T1A and T2A are efficient. The 3'-ends of the mRNAs which end at T1A and T2A were analysed by S1 mapping with total RNA purified from a mutant strain deficient in exoribonuclease activities, in particular RNase II. The results suggest that the potential secondary structures of T1A and T2A are sufficiently stable to prevent 3'-end degradation by RNase II. On the other hand, the T2A terminator should be efficient enough to stop transcription through the downstream DNA region involved in pColA replication.