OUP user menu

Production of cholera toxin B subunit in Lactobacillus

Philippe Slos, Philippe Dutot, Jacqueline Reymund, Patricia Kleinpeter, Deborah Prozzi, Marie-Paule Kieny, Jean Delcour, Annick Mercenier, Pascal Hols
DOI: http://dx.doi.org/10.1111/j.1574-6968.1998.tb13295.x 29-36 First published online: 1 December 1998


The intracellular expression of the B subunit of cholera toxin (CTB) was first achieved in Lactobacillus paracasei LbTGS1.4 with an expression cassette including the P25 promoter of Streptococcus thermophilus combined with the translation initiation region from the strongly expressed L. pentosusd-lactate dehydrogenase gene (ldhD). Secretion of CTB was next attempted in L. paracasei LbTGS1.4 and L. plantarum NCIMB8826 with four different signal sequences from exported proteins of lactic acid bacteria (Lactococcus lactis Usp45 and PrtP, Enterococcus faecalis unknown protein and S. pyogenes M6 protein). Host-dependent secretion of CTB was clearly observed: whereas none of the secretion cassettes led to detectable CTB in the extracellular fraction of L. paracasei LbTGS1.4, secretion of CTB molecules was clearly achieved with three of the selected signal sequences in L. plantarum NCIMB8826.

Key words
  • Lactobacillus
  • Expression vector
  • Protein secretion
  • Cholera toxin B
  • Live vaccine

1 Introduction

Lactobacilli are part of the normal commensal microbial flora of healthy humans where they proliferate in body cavities such as the gastrointestinal (GI) and urogenital tracts [1]. Many pathogenic viruses and bacteria use the corresponding mucosal surfaces to infect their hosts. The use of non-pathogenic Gram-positive bacteria (Lactococcus lactis, Streptococcus gordonii, Staphylococcus xylosus and lactobacilli) as antigen delivery system has been widely reported in the literature (see [2] for a review). Lactobacilli differ from the other non-pathogenic Gram-positive bacteria used as potential live vaccines (with the exception of S. gordonii, a commensal from the oral cavity) by naturally occupying specific mucosal niches (i.e. GI tract and vagina) which could be used to stimulate a strong local immune response.

The exploitation of lactobacilli as potential live delivery vehicles requires the development of adequate antigen expression and secretion systems. We recently demonstrated secretion with different efficiencies of a model antigen consisting of the N-terminal part of the S. pyogenes M6 protein fused to a HIV gp41-epitope in L. plantarum NCIMB8826 (a strain of human buccal origin) and L. paracasei LbTGS1.4 (vaginal murine strain) [3]. In this paper, we focus on the expression and secretion of the cholera toxin B subunit (CTB) which is a major protective antigen from Vibrio cholerae[4]. The cholera toxin (CT) is composed of five identical non-toxic B subunits of 11.6 kDa (CTB) that bind to GM1 ganglioside receptors in intestinal cells, and of one toxic A subunit of 27 kDa (CTA). In the original host, CTB and CTA are exported into the periplasm after cleavage of the signal sequences. During export, an intramolecular disulfide bridge is formed in each CTB monomer and the mature polypeptides spontaneously form pentamers. Subsequently, the A subunit and the CTB pentamer assemble to form the holotoxin [4, 5].

We report here that intracellular production of CTB in L. paracasei LbTGS1.4 was successfully achieved but that targeting to the secretion pathway with different signal sequences did not lead to detectable CTB molecules either secreted in the external medium or intracellularly located. Conversely, upon transfer in L. plantarum NCIMB8826, most of the constructs led to secretion of CTB monomers. The secreted CTB subunits, however, did not bind the GM1 receptor probably due to the failure of pentamer assembly.

2 Materials and methods

2.1 Bacterial strains, media and growth conditions

Escherichia coli ED8739 and MC1061 were used as host strains for all the cloning experiments and E. coli NM522 was used for M13 bacteriophage propagation as previously described [3, 5]. L paracasei LbTGS1.4 was isolated from mouse vagina [3] whereas L. plantarum NCIMB8826 (NCIMB, National Collection of Industrial and Marine Bacteria, Terry Research Station, Aberdeen, UK) originated from human saliva. MRS medium was routinely used for the propagation of lactobacilli at 37°C under limited aeration [3]. When needed, chloramphenicol was added at the following concentrations: 20 μg ml−1 for E. coli and 10 μg ml−1 for lactobacilli.

2.2 Plasmid constructions and DNA manipulations

The detailed construction strategies and sequences of the 24 oligonucleotides used during this study can be obtained on request to the corresponding author. DNA fragments were isolated from agarose gels using the Geneclean kit (Bio 101). Vector production, DNA restriction and ligation were performed according to standard techniques [6]. E. coli, L. plantarum NCIMB8826 and L. paracasei LbTG1.4 transformation and plasmid DNA extraction were performed as described previously [3].

Three different ribosome binding sites (RBS) were inserted upstream of the ctxB gene cloned in M13TG130 bacteriophage [5]. These insertions were performed by in vitro mutagenesis using primers containing translation initiation signals from an unknown ORF (ORFIII) of L. helveticus[7], from the ldhD gene of L. pentosus[8] and from the gene encoding factor IIIlac of L. casei[9], respectively. Mutated DNA inserts were sequenced by the dideoxynucleotide method [6]. The SalI and XbaI fragments encoding the RBS-ctxB fusions were inserted (Fig. 1A) into the corresponding sites of the P25 promoter-carrying plasmid pTG292 [10]. A general expression vector, pTG2247, was constructed by inserting into pTG292 a synthetic DNA fragment encoding the ldhD RBS (Fig. 2A).

Figure 1

Features of the expression modules. A: Construction of the ctxB intracellular expression vectors pTG2217, pTG2218 and pTG2219 by insertion of SalI-XbaI fragments bearing different RBS-ctxB fusions into pTG292. The sequences of the RBS between the SalI site and the starting codon ATG are shown and their origins are given below. The SD sequences are boxed; the ΔG values (in kcal mol−1) for the different duplexes with the 16S rRNA are indicated above each box. The distance between the SD sequence and the start codon is indicated between brackets above each RBS (the reference nucleotide is G next to A in the SD sequence). The fragment encoding the mature CTB (′ctxB) is represented by a dashed box. The transcription module of the basal pTG292 vector ([10], lactococcal pSH71 derivative) includes the chromosomal P25 S. thermophilus promoter (boxed), a multicloning site containing six unique restriction sites (overlined) and the T1T2 transcriptional terminator (double vertical hairpins). B: Potential RNA secondary structures and their ΔG values (in kcal mol−1) observed in the TIRs (RNAFOLD software, PCGENE, Intelligenetics).

Figure 2

Features of the secretion modules. A: Nucleotide sequence between SalI and SacI in the general expression vector pTG2247 and its derivative pTG2251. Conventions concerning RBS, ′ctxB and the restriction sites are as in Fig. 1A. B: Schematic representation of the different protein fusions carried by the secretion vectors. The PCR fragments encoding the signal sequences were inserted in pTG2251. The origins of the different signal sequences are indicated on the left. The sequence and the number of positively charged aa of the ‘n’ region, the size of the hydrophobic ‘h’ region and the the sequence of the ‘c’ region near the potential signal sequence cleavage site are shown. The surrounded amino acids are encoded by the multicloning site of the vector.

The constructions of the secretion vectors were carried out following a two-step strategy. First, a BamHI site was created by in vitro mutagenesis upstream of the codon encoding the first amino acid of mature CTB using M13TG130 vector. The BamHI/KpnI-cleaved mutant ctxB was inserted into BamHI/KpnI-cleaved pTG2247 to generate plasmid pTG2251 (Fig. 2A). In a second step, a series of secretion leaders were obtained by polymerase chain reaction (PCR) amplification. Amplification was performed with 1 ng of target plasmid DNA with the following step programme (94°C/30 s; 55°C/45 s; 72°C/45 s; 25 cycles; final 10 min extension at 72°C). The primers used to amplify the secretion leaders of Usp45 [11], PrtP [12] or M6 [13] contained a XbaI site which resulted in the replacement of the corresponding second and third codons of the signal peptide by those encoding serine and arginine, respectively (Fig. 2B). These three secretion leader fragments were cleaved with XbaI and BamHI and inserted into XbaI/BamHI-cleaved restricted pTG2251(Fig. 2A). No change was brought to the amino-terminal sequence of the signal sequence encoded by plasmid pGIP212.4 since it was amplified with a 5′-end primer carrying a NdeI site. In this case, fragments were cleaved by NdeI and BamHI and inserted into pTG2251 cleaved with NdeI (partial) and BamHI. The ctxB gene was also inserted into the central part of the coding sequence of the M6 protein to give construction pTG2297 (Fig. 2B).

2.3 Preparation of protein extracts and supernatant concentration

The CTB exported in the periplasm E. coli MC1061 cells bearing the desired plasmids was recovered as described previously [5]. Cellular extracts of L. paracasei LbTGS1.4 were obtained with the same protocol except that these cells required a longer sonication time compared to that of E. coli. L. plantarum NCIMB8826 cells harbouring the CTB secretion constructs were harvested by centrifugation (12 000×g; 4°C; 5 min) after 7 h of growth at 37°C until absorbance at 600 nm reached 3.0. The supernatant from 3 ml of culture was concentrated 15-fold by filtration on Microsep Filtron 10K (Schleicher and Schuell, Gent, Belgium) and diluted in 5×Laemmli buffer [6]. Total cell extracts of L. plantarum NCIMB8826 were prepared by centrifugation of 1.5 ml of cells (12 000×g, 5 min, 4°C), followed by resuspending the cell pellet in 75 μl of PBS buffer before diluting in 5×Laemmli buffer [6].

2.4 SDS-PAGE and protein characterisation

Protein extracts or concentrated supernatants were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblots were performed with a rabbit antibody against CTB protein (primary antibody, a gift of C. Locht, Institut Pasteur, Lille, France) as described previously [3, 5]. Ganglioside immunosorbent assay (GM1-ELISA) was carried out as described by Svennerholm and Holmgren [14]. For non-denaturing electrophoresis, mercaptoacetic acid, SDS and heating for 3 min at 95°C were omitted. Rainbow markers (Amersham) served as standards.

3 Results and discussion

3.1 Intracellular expression of ctxB in L. paracasei LbTGS1.4

Since lactobacilli belong to the Clostridia group of Gram-positive organisms like Bacillus subtilis which have been shown to be translationally specific [15], it was anticipated that their ribosomes would initiate translation only from RBS of Gram-positive bacteria. In consequence, RBS from three sequenced genes of lactobacilli were inserted upstream of the ctxB gene. These fusions were inserted into pTG292 to yield the corresponding plasmids pTG2217 (FIIIlac), pTG2218 (LdhD) and pTG2219 (ORFIII) (Fig. 1A).

Intracellular expression of ctxB was first analysed in E. coli. A product of same electrophoretic motility as purified CTB from V. cholerae was indeed detected by immunoblots in the insoluble part of the cytoplasmic fraction [5] of all three E. coli recombinants (data not shown). A hierarchy of expression was established as follows: pTG2218>pTG2217≫pTG2219 (barely detectable). In L. paracasei LbTGS1.4, ctxB expression was detectable by immunoblots only for pTG2218 (Fig. 3, lane 4). As for E. coli, CTB was recovered exclusively in the insoluble fraction of the cellular extracts.

Figure 3

Immunoblots of soluble (Cs) or insoluble (Ci) cell extracts from L. paracasei LbTGS1.4 cells harbouring pTG2217, pTG2218 or pTG2219. Lane 1, untransformed LbTGS1.4; lane 2, control pCK17; lane 3, pTG2217; lane 4, pTG2218; lane 5, pTG2219; lane 6, 100 ng purified rCTB (arrow).

It is well established that Gram-positive bacteria are more stringent in translation initiation that Gram-negative bacteria. Their translation initiation regions (TIR) require an extended Shine-Dalgarno (SD) sequence and translation initiation is more sensitive to secondary structures [15]. The SD sequences of the three different expression cassettes show similar free energies of binding with the 3′-end of 16S rRNA, but the spacing between the SD sequence and the start codon is variable and especially short (6 nt) for pTG2217 (Fig. 1A). A search for potential RNA secondary structures in the TIR region revealed hairpin structures that could correlate with the variable level of expression in E. coli and the absence of functionality of the TIRs from pTG2117 and pTG2119 in L. paracasei LbTGS1.4 (Fig. 1B). The most dramatic situation is observed for pTG2119 where the SD sequence and the ATG start codon are both sequestered in an hairpin, a situation which has been reported to be very unfavourable for translation initiation [15]. The TIRs of pTG2217 and pTG2218 both displayed a secondary structure sequestering the SD sequence with ΔG values of −4.1 and −3.2 kcal mol−1, respectively (Fig. 1B). However, the spacer length (12 nt) in the TIR of pTG2218 is in the range of values (between 10 and 13 nt) observed in efficient translation signals while a 6-nt spacer, as observed for pTG2217, results generally in very poor translation [15].

Nevertheless, the fact that ctxB can be expressed intracellularly from pTG2218 in L. paracasei LbTGS1.4 provided evidence that there are no transcriptional and translational (e.g. codon usage) barriers to prevent expression in this host.

3.2 Construction of secretion vectors

Based on the results described above, the general expression vector pTG2247 designed to perform translation fusions was constructed by inserting the RBS of the ldhD gene including a NdeI site spanning the ATG start codon (Fig. 2A). To demonstrate that the addition of the NdeI site did not modify the expression level, the ctxB gene was inserted between the BamHI and KpnI sites to generate plasmid pTG2251 (Fig. 2A). L. paracasei LbTGS1.4 transformed with pTG2218 and with pTG2251 indeed produced similar amounts of CTB (data not shown).

In order to target CTB to the general secretion pathway of L. paracasei LbTGS1.4, four signal sequences known to be functional in lactobacilli (see below) were inserted upstream of the ctxB gene in pTG2251 (Fig. 2B). The first secretion construct (pTG2289) made use of the signal sequence from PrtP, a cell envelope-associated proteinase found in L. lactis SK11 and L. paracasei subsp. paracasei[12]. In the second construct (pTG2288), we inserted the usp45 signal sequence from L. lactis[11]. We also made use of plasmid pGIP212.4 containing the signal sequence of an unknown gene from E. faecalis which was reported to allow efficient α-amylase secretion in L. plantarum NCIMB8826 [16]. As signal sequence analysis had previously revealed four potential cleavage sites, four fragments of increasing lengths covering these potential cleavage sites were inserted in pTG2251 to generate constructions pTG2269, pTG2270, pTG2265 and pTG2267, respectively (Fig. 2B). The last two constructs were based upon the signal sequence from the S. gordonii M6 protein [13] which was previously shown to drive secretion of an M6-derived fusion protein carrying an HIV epitope in LbTGS1.4 [3]. No recombinant clones could be obtained by inserting the minimal length of the M6 signal sequence in pTG2251. In contrast, insertion of a fragment encoding the signal sequence and the first five amino acids of the mature part of the M6 protein was obtained (pTG2294, Fig. 2B). Furthermore, since the export initiation domain adjacent to the signal sequence may be critical for protein translocation [17], a translational fusion was also made between ctxB and the first 106 amino acids of the M6 protein to give construction pTG2297 (Fig. 2B).

3.3 Host-dependent secretion of CTB in lactobacilli

The capacity of these various secretion constructs to direct CTB export was first evaluated in E. coli. With the exception of pTG2297, all constructs enabled the corresponding recombinant E. coli strains to export CTB molecules that were closely related to rCTB [5] in the following aspects: molecular mass under denaturing conditions, pentamer formation under non-denaturing conditions and binding to plastic-coated GM1 ganglioside (data not shown). The recombinant protein from pTG2297 displayed a 30-kDa fusion protein as expected but was unable either to form pentamers nor to bind GM1 receptors probably due to the large (106 aa) N-terminal M6 extension.

The CTB secretion constructs were next introduced by electroporation into L. paracasei LbTGS1.4. Secretion of CTB was undetectable in the culture supernatants (4 ml TCA-precipitated) of the corresponding recombinant lactobacilli by the GM1 immunosorbent assay or by immunoblot. Moreover, no traces of CTB precursor nor of processed CTB could be detected in the soluble or insoluble intracellular fractions of any of the recombinant LbTGS1.4 clones by Western blot analysis. Nevertheless, specific mRNA molecules isolated from all the recombinant strains were detected by Northern blot analysis (data not shown), suggesting that a posttranscriptional event prevented CTB production in this host.

The capacity to secrete proteins may vary between different species belonging to the same genus. Previous studies have shown that secretion was much more efficient in L. plantarum NCIMB8826 as compared to L. paracasei LbTGS1.4 [3], which prompted us to introduce the CTB secretion constructs into this host as well. Fig. 4 shows an immunoblot performed on concentrated supernatants from cultures of L. plantarum transformants carrying the different secretion plasmids. A protein of the same size as the control mature monomeric CTB was detected for pTG2265, pTG2267, pTG2270, pTG2289, and pTG2294; a larger protein was observed for pTG2297 as expected (Fig. 4, lane 9) but no CTB molecules were detected for pTG2269 nor pTG2288 containing the shortest version of the signal sequence from E. faecalis and the Usp45 signal sequence, respectively (Fig. 4, lanes 4 and 6). However, none of the supernatant showed GM1 ganglioside binding activity in agreement with the observation that no pentamers were formed (Fig. 4).

Figure 4

Immunoblots of supernatants (180 μl concentrated) from L. plantarum NCIMB8826 harbouring the CTB secretion constructs. Samples were diluted at room temperature in 5×Laemmli buffer prepared without mercaptoacetic acid or SDS prior to migration. Lane 1, rCTB (26 ng); lane 2, pTG2265; lane 3, pTG2267; lane 4, pTG2269; lane 5, pTG2270; lane 6, pTG2288; lane 7, pTG2289; lane 8, pTG2294; lane 9, pTG2297; lane 10, pCK17. The upper and lower arrows indicate the running positions of the pentameric and monomeric form of CTB, respectively.

In order to establish that the absence of CTB-reactive material in the pTG2269 and pTG2288 supernatants was due to a secretion defect, we compared the supernatants and the total cell extracts from these two recombinant strains by Western blot analysis. As can be seen in Fig. 5A, this clearly demonstrated in each case the presence of a CTB band in the total cell extracts with a slightly lower mobility than the reference mature CTB, giving credit to the interpretation that the two corresponding precursors could not be properly processed. We also wanted to make sure that the CTB bands seen in the supernatants from the other cultures were not artefacts resulting from uncontrolled cell lysis. This was done by repeating the Western blot analysis on both the supernatants and total cell extracts from cultures pTG2265, pTG2267, pTG2270, pTG2289, and pTG2294. It can be seen in each case (Fig. 5B) that CTB bands from the two sources have exactly the same mobility as the reference (processed) rCTB (Fig. 4). The presence of cell-associated CTB displaying a similar size as the extracellular CTB suggests that the precursor has been processed by the secretion machinery and that a fraction of the mature CTB is entrapped in the cell wall. A similar hypothesis has already been proposed for the secretion of various proteins in Gram-positive bacteria including lactobacilli [3].

Figure 5

Immunoblots of supernatants (S, 180 μl concentrated) and total cell extracts (C, equivalent to 200 μl of culture) from L. plantarum NCIMB8826 harbouring the CTB secretion constructs. A: Cultures of L. plantarum harbouring pTG2269 (lane 2) and pTG2288 (lane 3). Purified rCTB (26 ng) is shown on lanes 1 and 4. Samples were boiled in 5×Laemmli buffer prior to migration. B: Cultures of L. plantarum harbouring pTG2265 (lane 1), pTG2267 (lane 2), pTG2270 (lane 3), pTG2289 (lane 4) and pTG2294 (lane 5). Samples were prepared as in A.

Thus, L. plantarum is able to drive secretion of CTB molecules using three different signal sequences containing ‘n’ regions with 3–7 positively charged amino acids (aa), hydrophobic ‘h’ regions comprising 16–24 aa and ‘c’ regions 4–7 aa in size (Fig. 2B). The absence of secretion with the shorter signal sequence could easily be explained by a too short ‘c’ region which does not allow a correct positioning of the potential cleavage site as already observed with other truncated signal sequences [17]. The absence of secretion with the Usp45 signal sequence correlated with the poor secretion efficiency (50%) observed in L. plantarum and L. paracasei using this cassette fused to the Bacillus stearothermophilusα-amylase (amyS) reporter [3]. Furthermore, one positively charged amino acid out of three has been lost during construction of pTG2288 which could negatively affect the functionality of the signal sequence by a modification of the N/C charge balance [17]. Finally, the signal sequence of Usp45 in its original configuration displayed a higher predicted propensity for β-turn configuration around the cleavage site than the other signal sequences in their genuine context (data not shown). Fusion of CTB to Usp45 dramatically decreased this β-turn propensity which is known to be an important determinant of cleavage site recognition by the signal-peptidase [17].

However, the absence of GM1 binding for these CTB molecules suggests that their folding was incorrect in L. plantarum. Knowing that oxidation of cysteine residues in the B subunits is a prerequisite for in vivo formation of CTB pentamers in V. cholerae[4], it is possible that L. plantarum NCIMB8826 is limited in its ability to catalyse the formation of disulfide bonds. Interestingly, a bdb gene coding for a thiol disulfide oxidoreductase shown to be enzymatically active in both reduction and oxidation of disulfide bonds in vitro has been cloned from Bacillus brevis[18]. The biological role and the true substrate for Bdb remain unidentified. It is intriguing that B. brevis is, to our knowledge, the only Gram-positive organism in which this type of activity has been found, and also in which high amounts of correctly folded CTB have been produced [19]. The co-expression of bdb and ctxB in lactobacilli is currently under investigation to test the possibility of producing correctly folded and biologically active CTB molecules in these hosts.


D. Villeval, A. Spindler, and O. Roch contributed expert technical assistance during the course of this work. We acknowledge W. de Vos and G. Pozzi for providing pNZ10α-1, pNZα-15 and pGP603, respectively. We thank C. Locht for the gift of the anti-CTB antibodies. This research was carried out in the framework of Community Research Programme (Contract BIOT-CT94-3055). The authors acknowledge the financial support of the French Agency for Aids Research (ANRS).

gastrointestinal tract
B subunit of cholera toxin
recombinant CTB
A subunit of cholera toxin
ribosome binding site
translation initiation region


View Abstract