For use as a mucosal adjuvant for human vaccines, a simple method has been developed for the affinity purification of recombinant cholera toxin B subunit which had been expressed in a safe host, Bacillus brevis. Recombinant cholera toxin B subunit, adsorbed quantitatively to a d-galactose-agarose column, was eluted with an 0.1–0.4 M d-galactose gradient with a yield of > 90%. The cholera toxin B subunit preparation was similar to the native cholera toxin B subunit with respect to GM1 binding ability, remarkable stability of the pentamer, and the dissociation-reassociation property by shifting pHs. Cross-linking experiments with glutaraldehyde demonstrated that the pentameric form was predominant; tetrameric, trimeric, dimeric and monomeric forms were detected to a lesser extent, and additionally 10- and 15-mers were observed depending on the concentration of the cholera toxin B subunit.
Mucosal vaccination has received increasing attention with respect not only to anti-infectious but also anti-inflammatory purposes. Attractive features of mucosal immunization are: protective immune responses at the mucosal surfaces that may be inducible for the majority of pathogens which infect via mucosal surfaces; immune responses inducible not only at the inductive sites, but also in distant mucosal effector sites; mucosal immunization that can also induce systemic immune responses including humoral and cellular immunities. Mucosal immunization also has the potential to regulate systemic tolerance against a variety of allergens associated with immunological disorders including allergic, inflammatory and auto-immune diseases . The other advantages of administration of mucosal vaccines may be that they are simpler and safer compared with injectable vaccines. However, as most protein antigens are poor immunogens when given mucosally, good adjuvants are needed for development of effective mucosal vaccines.
Cholera toxin (CT) is one of the most effective adjuvants yet described for experimental use for inducing mocosal immunity. It enhances the immunogenicity of relatively poor mucosal immunogens when mixed or conjugated together and given orally or nasally (reviewed by Elson ). CT is composed of an A subunit (28 kDa) which is involved in the ADP-ribosylation of the stimulatory Gs protein of adenylate cyclase, and a B oligomer (CTB) which is a homopentamer of noncovalently associated subunits (11.6 kDa each) that form a ring-like structure and which binds with high affinity to its receptor ganglioside GM1 present on most mammalian cells [3,4]. From a practical standpoint, CT cannot be used in humans because of its potential toxicity; thus, many studies using CTB, which retains the binding capability but does not have the toxic activity [5,6], revealed that CTB also has significant adjuvanticity . However, since commercial CTB preparations containing a trace amount of CT (< 0.5%) were used in most experiments, the contribution of the CTB alone on adjuvanticity has become unclear. Induction of adjuvanticity by CTB to co-administered antigen also depends on the antigens, their forms (admixture or conjugate with CTB), route of administration (oral, intranasal, rectal or vaginal), dose, genetic background, and animal species . Additionally, CTB has been shown to enhance oral tolerance to the conjugated antigens , in contrast with CT which abrogates tolerance .
To examine substantial adjuvant activity and tolerogenic activity and to determine conditions for inducing proper mucosal immune responses, especially for human vaccine use, a large amount of recombinant CTB (rCTB) which is inherently uncontaminated with the holotoxin is required. Some recombinant systems for overexpression of rCTB have been developed using Vibrio cholerae[9,10] and Escherichia coli. In our laboratory, a very efficient expression-secretion strain for mature CTB was constructed using a Bacillus brevis host-vector system ; the rCTB gene introduced was from V. cholerae O1 strain 569B, classical biotype, Inaba serotype and its amino acid sequence was identical to that reported by Takao et al. .
In this paper, we have established a simple purification method for rCTB and describe some properties including the fact that the preparation contained mainly a stable pentameric form of the subunits.
Materials and methods
Bacterial strain and growth conditions
B. brevis HPD31 (pNU212-CTB)  was cultured on T2M  -EM agar containing 1% polypeptone (Nihon Pharmaceutical), 0.5% fish meat extract (Wako Pure Chemical), 0.2% yeast extract (Difco Labs.), 1% glucose, 0.001% FeSO4·7H2O, 0.001% MnSO4·4H2O, 0.0001% ZnSO4·7H2O, pH7.0, 10 µg ml−1 erythromycin (EM) and 1.5% agar at 30°C for 2 days. Secretion of rCTB was confirmed by colony immunoassay on a nitrocellulose sheet as described in immunoblotting (see below). Bacteria secreting a sufficient amount of rCTB were then subcultured at 30°C to mid-exponential phase in 4 ml of enriched 5YC  -EM broth containing 4% polypeptone P1 (Nihon Pharmaceutical), 0.5% yeast extract, 3% glucose, 0.01% CaCl2·2H2O, 0.01% MgSO4·7H2O, 0.001% FeSO4·7H2O, 0.001% MnSO4·4H2O, 0.0001% ZnSO4·7H2O, pH 7.2 and 10 µg ml−1 EM on a reciprocal shaker (Model NR-3, Taitec Instruments) at 150 rpm. A fermenter (Bio Reactor Model CTB-1, Taitec Instruments) containing 1 l of fresh enriched 5YC-EM medium and 0.1 ml antifoam (Antifoam 289, Sigma Chemical) was inoculated with a total of 20 ml subculture and incubated at 30°C with high aeration (2.5 l min−1) and stirring (500 rpm) for 5 days; on the third day 3% glucose was added.
Preparation of d-galactose-agarose
For affinity chromatography, the immobilized galactose matrix was prepared by coupling d-galactose (Wako Pure Chemical) and divinyl sulfone-activated agarose (0.2 g d-galactose per g matrix, Mini-Leak High, Kem-En-Tec) in 1 M sodium carbonate-sodium bicarbonate, pH 11.0 at 25°C for 20 h with gentle shaking, followed by blocking of excess active groups with 0.1 M ethanolamine-NaOH, pH 11.0 for 2 h. The matrix was washed repeatedly with phosphate-buffered saline, pH 7.2 (PBS) and 0.05 M sodium citrate buffer-0.2 M NaCl, pH 2.8 and finally equilibrated with PBS.
Secreted proteins in 1 l of the culture supernatant of B. brevis were precipitated with 70% saturated ammonium sulfate and allowed to stand overnight at 4°C. The precipitate dissolved in PBS was dialyzed against PBS overnight at 4°C in Spectra/Por cellulose tube (MW cut-off 6000–8000, Spectrum). After loading the dialysate (ca. 150 ml) on the d-galactose-agarose column (bed volume, 180 ml) and washing extensively with PBS, rCTB was eluted in a gradient of 0–1 M d-galactose in PBS, followed by 1 M d-galactose in PBS and finally with 0.05 M sodium citrate buffer-0.2 M NaCl, pH 2.8; the last eluate was immediately neutralized by the addition of 2 N NaOH. The citrate buffer concentration was based on suitable conditions for holotoxin to be eluted from the GM1 affinity column . The rCTB fractions eluted with 0.1–0.4 M d-galactose were pooled, dialyzed against PBS in the Spectra/Por cellulose tube and condensed by dehydration with polyethyleneglycol 20,000 powder (Katayama Chemical) outside the dialysis tube at 4°C. The condensate was dialyzed extensively and loaded on a Bio-Gel P6DG (Bio-Rad Labs.) column in PBS to remove any residual amount of galactose; otherwise galactose may abrogate the binding ability of the mucosally administered CTB to the GM1 . Fractions eluted with 1 M galactose and the citrate buffer were treated in the same manner as above. The purified rCTB preparations were stored at −20°C.
GM1 enzyme-linked immunosorbent assay
GM1 binding ability of the rCTB was measured by a GM1 enzyme-linked immunosorbent assay (GM1 ELISA). Microtiter plates (Nunc-Immunoplate/MaxiSorp) were coated with 50 µl per well of 2 µg ml−1 monosialoganglioside-GM1 (Sigma Chemical) dissolved in methanol. Methanol was evaporated at room temperature overnight. Nonspecific protein binding sites were blocked with 1% bovine serum albumin (BSA) in PBS (200 µl per well) at room temperature for 2 h. Samples appropriately diluted with PBS (100 µl per well) were added and the plates were incubated at 30°C for 1 h. After washing with PBS, plates were incubated with rabbit polyclonal anti-CTB antiserum diluted with PBS (50 µl per well, 1:1000 dilution) at room temperature for 1 h; the antiserum was prepared by several subcutaneous immunizations of the rabbit against commercial CTB (Sigma Chemical) with Freund's complete and incomplete adjuvants (Difco Labs.). Plates were again washed with PBS and bound antibody was detected by incubation with horseradish peroxidase-conjugated protein A (50 µl per well; 1:1000 dilution, EY Labs.) at room temperature for 1 h. Finally, plates were washed with PBS and a chromogenic substrate, 3,3′5,5′-tetramethylbenzidine-H2O2 solution (50 µl per well, Bio-Rad Labs.) was added. Reaction was carried out at room temperature under vibration using a microplate mixer (Model MX-4, Sanko Junyaku). Color development was stopped after 30 min by addition of 1 N H2SO4 (50 µl per well) and the absorbance at 450 nm was measured with an automatic microplate reader (Model MPR-A4i, Tosoh Corporation). A CTB preparation purchased from Sigma Chemical was used as a standard.
Sodium dodecylsulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting
Recombinant CTB samples mixed 1:1 with a sample buffer (50 mM Tris-HCl, pH 6.8, 2% SDS, 2% 2-mercaptoethanol, 10% glycerol, and 0.01% bromophenol blue) were applied to a gel after boiling for 10 min unless otherwise stated. Electrophoresis was performed in a 10–20% gradient acrylamide slab gel (Ready Gel J, Nippon Bio-Rad Labs.) with a constant current of 30 mA per gel for 45 min. Proteins were detected by Coomassie brilliant blue staining (CBB-R250, Daiichi Pure Chemicals) or silver staining (Daiichi Pure Chemicals).
For immunoblotting, proteins separated by SDS-PAGE were transblotted from the gel into a PVDF membrane (Immobilon-P, Millipore) using a semidry electroblotter (Marysol). After blocking with 1% BSA-PBS at room temperature for 1 h, the sheet was incubated with rabbit anti-CTB antiserum diluted 1:1000 in 1% BSA-PBS at 30°C for 2 h, and washed three times with 0.25% BSA-PBS for 10 min each. The sheet was incubated with horseradish peroxidase-protein A diluted 1:2000 in 0.05% Tween 20-PBS at room temperature for 1 h, washed with 0.05% Tween 20-PBS three times and with water once, and finally stained with the precipitating peroxidase substrate, TMBlue (TM101, TSI-CDP). After color development, the sheet was washed with water and dried.
Cross-linking of CTB molecules
To investigate oligomerization of the CTB subunits, cross-linking of the protein was carried out. The rCTB (10–500 µg ml−1) in PBS was incubated with 0.05% freshly diluted glutaraldehyde (EM grade, Nisshin EM) in PBS at room temperature for 30 min, followed by addition of 1 M l-lysine·HCl to block excess aldehyde groups and dialysis against PBS using a dialysis cup (MW cut-off 3500, Bio-Tech International) for 2 h.
Estimation of rCTB
The amount of purified rCTB was estimated from OD1% at 280 nm in a 1-cm cell at a coefficient of 9.56  and that in crude preparations was estimated roughly by GM1 ELISA. Protein concentration was determined by the Bradford method (Bio-Rad Protein Assay) using BSA as a standard protein.
Purification of rCTB
A typical elution profile of the d-galactose-agarose column is shown in Fig. 1. After a large amount of unadsorbed proteins flowed through the column, adsorbed proteins with GM1 binding ability were eluted with 0.1–0.4 M galactose, 1 M galactose and citrate buffer, pH 2.8; the first eluate contained usually over 90% of the total protein in the eluate. The SDS-PAGE and immunoblot analyses of the peak fractions (P1, P2, P3 and P4 in Fig. 1) showed a distinct band corresponding to the standard CTB monomer with a molecular mass of 12 kDa and often another faint band corresponding to the dimer of 24 kDa in each peak fraction (Fig. 2), indicating that rCTB was obtained in pure form by single-step affinity chromatography.
Specific GM1 binding abilities of rCTB in fractions P1, P2 and purified preparation were almost the same as that of the standard CTB, and those in fractions P3 and P4 appeared a little lower (Fig. 3).
Oligomerization of rCTB
In spite of the high purity of the preparation examined by SDS-PAGE and immunoblotting (Fig. 2), heterogeneity with respect to the elution profile appeared (Fig. 1). To examine whether the apparent heterogeneity was due to oligomerization or nonspecific aggregation, cross-linking of the rCTB molecules with glutaraldehyde was done for fractions P1 and P2. The SDS-PAGE patterns after cross-linking demonstrated that a pentamer was found predominantly, and tetramer, trimer, dimer and monomer were detected to a lesser extent (Fig. 4, lanes 1 and 2). These results indicate that the CTB molecules were present mainly as the pentameric form, but not as nonspecific aggregated forms; practically no difference in band pattern was found between the two fractions.
It is interesting to note that bands of higher oligomers which were estimated to be 10-mer and 15-mer from the migration distances were additionally present in the purified rCTB (Fig. 4, lane 3). These oligomers were observed when the concentration of rCTB was high, such as 100 µg ml−1. These facts suggest that rCTB subunits spontaneously assemble to form pentamers, and the pentamer as a unit interacts with another pentamer to form higher oligomers depending on the concentration.
Properties of rCTB pentamer
Some of the characteristic features of native CTB, such as the remarkable stability of the pentamer and the dissociation at acidic pH and reassociation at neutral pH [19,20], were investigated for our rCTB preparation. Without cross-linking and boiling treatment of the loading samples, stable pentamer was observed on SDS-PAGE and it migrated a little faster than the cross-linked denatured pentamer (Fig. 5), the results being consistent with the report that the native CTB pentamer is stable in solutions containing up to 1% SDS at pH> 5.5 and migrates as discrete oligomeric proteins . Dissociation-reassociation properties were examined by gel filtration on Sephadex G-75. At neutral pH, a broad major peak at void volume (MW> 20 kDa) was separated from a minor peak with the MW expected for the monomer, showing the presence of a high population in oligomeric form. At acidic pH, such as in the citrate buffer, pH 2.8 or in 5% formic acid which was the subunit-dissociating condition for CT [6,16,19,21], no detectable peak at void volume was observed but a single peak at the monomer position. After neutralization and dialysis against PBS, the acid-dissociated monomer fraction shifted the elution position to the void volume, showing reassociation. The reassociated oligomers retained GM1 binding ability and migrated as a stable pentamer (Fig. 5, lane 5), while the acid-dissociated form migrated as a monomer on SDS-PAGE (Fig. 5, lane 4).
The purified rCTB was stable in storage at −20°C, and after several freezing-thawing treatments practically no loss in binding ability to GM1 was observed.
A simple and inexpensive purification method using affinity chromatography with d-galactose-agarose was developed for rCTB produced by B. brevis (pNU212-CTB). Apparent heterogeneity in elution profile (Fig. 1) may be due to different binding affinity to the galactose moiety caused by conformational heterogeneity around the binding site, but not the sites for subunit-subunit interaction. The rCTB preparation mainly contained a pentameric form determined by cross-linking with glutaraldehyde and SDS-PAGE analyses. The results were consistent with those using native CTB and dimethyl suberimidate as a cross-linking reagent . Although whether the quaternary structure of rCTB is identical with the native doughnut-shaped ring shown by crystallographic studies of CTB  is not known yet, GM1 binding ability and the characteristic stability of the pentamer in the presence of SDS were similar to the native CTB. In addition to the pentameric form, 10- and 15-mers were detected in the concentrated preparation, showing possible pentamer-pentamer interaction, although the exact conformation remained unclear.
Uesaka et al.  reported previously that by using immobilized d-galactose (Pierce), holotoxin forms of recombinant CT (derived from V. cholerae GP14) and CT-related E. coli heat-labile enterotoxin (LT, derived from E. coli WT-1) could be purified by elution with 1 M and 0.3 M d-galactose and the yields were about 34% and 59%, respectively. In the present study, rCTB was mainly eluted with 0.1–0.4 M d-galactose (Fig. 1), the concentration of galactose being much less than that for the CT holotoxin and similar to that for the LT holotoxin. Therefore, at least through the terminal galactose moiety of GM1 ganglioside, CT holotoxin may bind with different affinity from its subunit CTB. These differences in affinity may be due to conformational differences around the receptor binding region between CT and CTB. In fact, the presence and absence of an A subunit affect binding affinity, because the carboxyl terminus of the A subunit comes into contact with the receptor in the case of LT .
Various purification methods for CT, CTB, and LT have been developed, including affinity chromatography using Agarose A5m , Spherosil-DEAE-dextran-lysoGM1 beads , and immobilized d-galactose , as well as absorption with Controlled Pore Glass  and high-pressure liquid chromatography on a DEAE Mem-Sep column . For large-scale purification of CT, affinity chromatography on Spherosil-DEAE-dextran-lysoGM1 beads is excellent with high yield . However, the preparation procedure of the beads appears not simple . The preparation of the d-galactose-agarose used in this study was easy, and the matrix may be useful for purification of other CTB derivatives, including conjugates, fusion proteins and peptides which have GM1 binding ability associated with mucosal adjuvanticity .
Additionally, our recombinant system for CTB production has a significant advantage especially for human vaccine use, since the host used was the Gram-positive bacterium B. brevis which is non-pathogenic, and the possibility of contamination of endotoxin and other virulence factors was excluded.
Using this rCTB preparation with unrelated antigens, mucosal immunization experiments in mice are under investigation. Our results showed that the rCTB revealed significant adjuvant effects on the production of systemic IgG and mucosal IgA against a model antigen BSA  and against tetanus toxoid . Furthermore, the rCTB may be applicable for various human mucosal vaccines as well as the whole cell plus recombinant B subunit (WC/rBS) oral cholera vaccine which has been proven to be safe and effective in field trials in Peru .
We would like to thank Dr. Roy H. Doi, Section of Molecular and Cellular Biology, University of California, Davis for his critical reading of the manuscript.