2720 M. Tomasi et al. Eur. J. Immunol. 1997.27: 2720-2725

Maurizio Tomasi’, Mark T. Dertzbaugh’, Thomas H e m 3 , Robert L. Hunter4 and Charles 0. Elson’

Division of Gastroenterology and Hepatology, University of Alabama at Birmingham, Birmingham, USA USAMRIID, Frederick, USA Center for Disease Control, Atlanta, USA Department of Pathology, Emory University, Atlanta, USA

1 Introduction

Strong mucosal adjuvanticity of cholera toxin within lipid particles of a new multiple emulsion delivery system for oral immunization

Cholera toxin (CT) is an effective mucosal adjuvant but causes significant intes- tinal secretion which limits its usefulness. In the present study we developed a new multiple emulsion (ME) delivery system into which antigen and CT could be incorporated and asked whether CT would retain its mucosal adjuvanticity when sequestered within emulsion particles. ME were selectively taken up into Peyer’s patches, and those containing antigen plus CT generated intestinal secre- tory IgA and serum IgG antibody responses in mice comparable quantitatively and qualitatively to those occurring after oral immunization with soluble antigen plus CT. The ME particles containing CT did not cause intestinal secretion. The adjuvanticity of CT within ME was due to the CT present in the inner aqueous phase of the ME and was lost if CT binding was blocked by pre-incubation with GM1 ganglioside. Proteins incorporated in ME were protected from external acid, protease, and bile. We conclude that CT sequestered in ME, although unable to bind to the epithelium and thus stimulate intestinal secretion, still retains its mucosal adjuvanticity. Thus, the ability of CT to bind to enterocytes is not obligatory for its mucosal adjuvanticity.

Protection of mucosal surfaces against pathogens requires delivery of antigen to and activation of the mucosal immune system; parenteral immunizations are generally ineffective [ 1, 21. Successful mucosal immunization will likely require the use of adjuvants such as cholera toxin (CT) or Escherichiu coli heat-labile toxin (LT) [3], which have mucosal adjuvanticity; however, the induction of secretory diarrhea [4] makes their potential use in humans problematic. CT binds diffusely to the epithelium, induc- ing cytokine production [5] and increasing epithelial per- meability [6], both of which have postulated to be impor- tant in its mucosal adjuvanticity. It is uncertain whether CT would remain an effective mucosal adjuvant if rendered non-toxic by sequestration within a particle. To test this question, we developed a new delivery system for oral antigen delivery utilizing water-in-oil-in-water or multiple emulsion (ME) [7], composed of water droplets within squalene oil particles which are formed in the pres- ence of nonionic block copolymers as emulsifiers and potential adjuvants [8,9]. The antigen used was a recombi- nant chimeric protein of bacterial origin, which, although immunogenic when injected parenterally, did not induce mucosal immunity even when milligram amounts were fed repeatedly to mice [lo]. [I 169051

Received April 14, 1997; in revised form July 18, 1997; accepted July 21, 1997.

Present address: M. Tomasi, Laboratorio di Biologia Cellulare, Istituto Superiore di Sanita, Rome, Italy.

Correspondence: Charles 0. Elson, Division of Gastroenterology and Hepatology, School of Medicine, University of Alabama at Birmingham, Birmingham, AL 35294-0007, USA Fax: + 1-205-934-8493

Abbreviations: CT: Cholera toxin ME: Multiple emulsion PP: Peyer’s patches

Key words: Mucosal adjuvanticity I Cholera toxin I Emulsion

S-IgA: Secretory immunoglobulin A

2 Materials and methods

2.1 Animals

Inbred female mice of the CB6F1 strain were purchased from Jackson Laboratory, Bar Harbor, ME. The mice were between 6 and 12 weeks of age when immunizations were begun. Outbred female ICR mice were obtained from Harlan Sprague-Dawley, Birmingham, AL.

2.2 Antigens

A recombinant chimeric protein, GtfB.l:: PhoA, obtained by a genetic fusion to the 5’ end of the PhoA gene of E. coli of an oligonucleotide encoding amino acid residues 345 to 359 of the GtfB enzyme of Streptococcus mutant GS-5, was prepared as described [lo]. Purified CT was obtained from List Biological Labs Inc., Campbell, CA.

2.3 Reagents

GM1, p-nitrophenyl phosphate, squalene, pilocarpine, bile extract, trypsin, chymotrypsin and soybean trypsin inhibitor were obtained from Sigma Chemical Co., St. Louis, MO. Goat anti-CT-B was purchased from Calbiochem-Behring, La Jolla, CA. Affinity-purified anti- body to mouse secretory IgA (S-IgA), total IgG, IgG1, IgG2a, either biotin labeled and unlabeled, streptavidin, and tetramethylbenzidine (TMB) peroxidase substrate were obtained from Kirkegaard and Perry Laboratories Inc. Gaithersburg, MD. Mouse IgA and IgG, used as quantitative standards, were purchased from Southern Biotechnology Associates Inc., Birmingham, AL. Tween- 80 and Tween-20 were from Fisher Scientific, Pittsburg, PA. Span 80 and P123 were obtained from Emulsion Engi- neering, Inc., Sanford, FL. The nonionic block copolymer Ll80, P1005 and P1003 were obtained from Cytrx Cor- poration, Norcross, GA.

0014-2980/97/1010-2720$17.50 + .SO10 0 WILEY-VCH Verlag GmbH, D-69451 Weinheim, 1997

Eur. J. Immunol. 1997.27: 2720-2725 Multiple emulsion for oral immunization 2721

2.4 Preparation of multiple emulsions

To immunize a group of five mice (0.5 mudose), 3.0 ml ME was routinely prepared. Sodium chloride solution (0.5 ml, 0.9 YO) containing 10 % nonionic block copolymer L180 or P1005 was incubated for 20 min at 4°C with the antigen dissolved in 0.3 ml 0.1 M NaHC03. Then 0.2-ml aliquots of this mixture were added and mixed with 0.7 ml 90 YO squalene oil, 10 '30 Span 80 in an apparatus consisting of two luer-lock single-use syringes without rubber gasket (Henke-Sass Wolf GmbH, Tuttlingen, Germany) con- nected by a three-way stopcock. When the emulsion was ready, as denoted by a sharp increase in the viscosity, 1.5 ml of 0.9 YO NaCl solution containing 0.25 YO Ween-80 and 0.5 YO block copolymer P1003 was added and mixed to homogeneity. In some experiments CT was blocked by GM1 ganglioside by preincubation of 60 pg CT with 200 pg GM1 ganglioside, previously dissolved in 200 pl PBS, for 30 min at room temperature, prior to incorpora- tion with antigen into ME.

2.5 Uptake of multiple emulsions in vivo

Outbred female ICR mice (Harlan Sprague-Dawley, Bir- mingham, AL) were infused intragastrically (i.g.) using blunt-tipped feeding needles (Popper & Sons, Inc., New Hyde Park, NY) every other day for 12 days with 0.5 ml of ME containing 5 mg nonionic block copolymer P1005 plus 35 mg R-900 titanium dioxide (TiO,) particles (E. I . du Pont de Nemours & Co., Inc., Wilmington, DE) in the inner phase. Two days after the last of the six doses, mice were killed and the Peyer's patches (PP) removed from the small intestine. The stained tissues were examined by light and dark field microscopy using a Leitz Laborlux micro- scope (Leica, Inc., Deerfield, IL).

2.6 Immunization and sample collection

For oral immunization the mice were given 0.5 ml ME i.g. using a 22-gauge feeding needle (Popper & Sons). Antigen given in soluble form (not in ME) as a control was sus- pended in 0.1 N NaHCO, pH 8.2 and given in a similar volume. i.g. infusions were performed on day 0, 14 and 28. Samples of serum were obtained on days 13,27 and 58. For i.p. injections, 0.1 ml of antigen suspended in alum (10 pg in 1 ml of 20% Maalox) was administered. Sera were obtained by collecting the blood with Natelson heparin- ized capillary tubes at the retro-orbital plexus under keta- mine anesthesia. Intestinal secretions were collected using a method as previously described [l l] .

2.7 Separation of the inner and outer water phase of multiple emulsion

A 1.5-ml Eppendorf tube with 1.0 ml of ME suspension was centrifuged (Eppendorf microfuge Mod. 5412 C) at 6°C for 10 min at 5520 x g. The milk-like emulsion (oil particles with the inner water phase) floated on the top of the tube above the external water phase. The tube bottom was then pierced with a needle and the external water phase collected into a recipient tube. The puncture in the original tube was tightly sealed with a Parafilm strip and

the milky emulsion resuspended in buffer. After washing in buffer, the emulsion was dissolved by mixing and vortex- ing it with 1.0 ml of 10 % (v/v) Triton X-100 followed by incubation for 2 h at 22 "C.

2.8 Fluid accumulation assay

To monitor the toxicity of CT in vivo we used the method developed by Richardson et al. [12]. Mice were infused i.g. with different combinations of CT in buffer or in the ME, and after a 6-h interval, they were killed and the entire small intestine was weighed. The data were expressed by the formula: FA = G/(B - G) X 1O00, where FA is the fluid accumulation, B = body weight in grams; G = gut weight in grams.

2.9 Measurement of antibody

The enzyme-linked immunosorbent assay (ELISA) was used to determine the concentration of antigen-specific antibody and for isotype distribution as previously described [lo]. Two methods of quantitation were used. In the first, serial twofold dilutions of serum or secretions were tested to determine the maximum dilution yielding an absorbance reading >0.1. The reciprocal of that dilu- tion was used for that sample in the calculations and sta- tistics. In the second approach, the specific antibody level was determined by reference to a standard curve obtained in a separate ELISA using known concentrations of mye- loma IgG or IgA [13].

2.10 Statistical analysis

The data were converted to mean and SD for comparison of groups. The significance of differences among means was determined using non-parametric Wilcoxon rank sum test.

3 Results

3.1 Uptake of multiple emulsion particles in vivo

The ability to deliver antigen into PP appears to be required for effective oral immunization. ME particles loaded with small 200 nm particles of TiO, were identified within the dome region of the PP after gavage but none were seen entering the epithelial layer or within the lamina propria (Fig. 1).

3.2 Oral delivery of antigen incorporated within multiple emulsions induced both secretory and systemic immunity

The recombinant chimeric bacterial protein GtfB.l:: PhoA was incorporated in ME with and without the addition of CT as a mucosal adjuvant. As shown in Fig. 2, a substan- tial S-IgA anti-GtfB.l:: PhoA titer was detected in the secretions of mice fed ME containing GtfB.l:: PhoA plus CT (p = 0.008, group 4 vs. group 3). The titer in this group was not significantly different from that in the posi-

2722 M. Tomasi et al.

a b

Eur. J. Immunol. 1997.27: 2720-2725

Figure 1. Uptake of ME containingTi02 into a mouse Peyer’s Patch after oral administration. Six feedings of ME containingTi02 were administered every other day and the mouse killed 2 days later. (A) TiOz particles are seen in the subepithelium of the dome but few are present in the deep cortex. Some SOz appears to be associated with intraepithelial lymphocytes (hematoxylin-eosin stain, Dark Field, 2 5 0 ~ ) . (B) The same field illuminated to show cell morphology (hematoxylin-eosin stain, 2 5 0 ~ ) .

tive control group fed soluble GtfB.l:: PhoA plus CT. The titer of S-IgA anti-CT was about 20-fold higher than S-IgA anti-GtfB.l:: PhoA (data not shown).

Mice from the groups with a significant S-IgA anti- GtfB .1:: PhoA response in secretions also had substantial serum IgG anti-GtfEi.l:: PhoA titers (Fig. 2), Again, the titers of serum antibody in the group receiving ME con- taining GtfB.l:: PhoA plus CT were significantly greater than in the group receiving ME containing GtfB.1::PhoA alone, and comparable to those achieved with soluble GtfB.1::PhoA plus CT. The subclass of IgG after feeding ME containing GtfB.1:: PhoA plus CT was predominately IgG1, indicating that delivery of antigen in ME did not alter the quality of the immune response as compared to that found after feeding soluble GtfB.1::PhoA plus CT (data not shown and [14]).

tem. To establish whether this amount of external CT might have adjuvanticity when given with ME, 0.5 pg of CT was added to the external water phase of ME contain- ing only GtfB.1::PhoA. Mice fed with this material did not produce any detectable amount of specific anti- GtfEX.l:: PhoA either in secretions or serum (Fig. 3, group 4), although small amounts of S-IgA and serum anti-CT IgG were produced (data not shown). This result indicates that the CT entrapped within the inner aqueous phase of the oil particles is responsible for the adjuvant activity.

A second question addressed was the number of doses of CT that are required. A group receiving ME containing GtfB.l:: PhoA plus CT for the first feeding but ME con- taining GtfB.l:: PhoA alone for the second and third feed- ings showed a response (Fig. 3, group 2), but that response was not as strong as that in mice fed three times with ME containing GtfEX.1::PhoA and CT (Fig. 3, group 3).

3.3 Importance of CT localization within ME particles for adjuvanticity 3.4 The ability of CT to bind to GM1 ganglioside is

required for its adjuvanticity within ME When the distribution of CT was measured, 5 YO was recov- ered from the external water phase of the ME, indicating that 95 % of the CT and likely also of GtfS.l:: PhoA was incorporated into the inner aqueous phase of the ME sys-

Because CT delivered in ME is sequestered from the epi- thelium (see below) and is delivered into the PP where it can be taken up by antigen-presenting cells, the ability to

Anti- Ad iuvant

I GtfB.1::PhOA - -

2 GttB.1::PhoA CT -

3 GtfB.1::PhoA - +

4 GtfB.1::PhoA CT +

5 - CT +

Figure 2. Intestinal S-IgA and serum IgG responses to GtfJ3.1: : PhoA in groups of mice orally immunized with different regimens. Groups of five female CB6F1 mice were gavaged on days 0, 14 and 28 with the combina- tions shown. The dose of GtfB.1::PhoA was 0.5 mg except for group 2 which received 1 mg/dose. The dose of CT used was 10 pg. Intestinal secretions were collected on day 28 and 58 and the titer of specific S-IRA anti- body measured by ELISA. Group 4 vs. group 3, p = 0.008; group 4 vs. group lgG anti GtfB.1::PhoA

( titer x 103 )

S-lgA anti GtfB.l::PhoA ( titer x 10-2 1 2, p > 0.05.

Eur. J. Immunol. 1997.27: 2720-2725 Multiple emulsion for oral immunization 2723

dntiaen Adiuvant .h!E

1 GttB.1::PhoA - +

2 GttB.l::PhoA CTx 1 +

3 GttB.1::PhoA C T x 3 +

4 GttB.l::PhoA ExtCT +

5 Gm.1::PhoA CT-GM1 + ,

IgG anti CtfB.1::PhoA (I%/ m l )

S-IgA anti CtfB.1 ::PhoA (ng I ml )

Figure 3. Intestinal S-IgA and serum IgG response to GtfB.l:: PhoA in mice orally immunized with GtfB.l:: PhoA in ME with or with- out CT as adjuvant. Groups of five female CB6F1 mice were gavaged with GtfB.1:: PhoA in ME on days 0,14 and 28 as shown. The dose of GtfB.1::PhoA was 0.5 mg and for CT was 10 pg except for group 4 which contained 0.5 pg CT in the external water phase only (Ext CT). CT x 1, CTpresent in ME only for the day 0 gavage; CT-GM1, CTpreincubated with GM1 ganglioside (at a 1 : 3.3 wlw ratio) before incorporation into ME. Intestinal secretions and serum were collected on day 28 and 58 and specific S-IgA antibody measured by quantitative ELISA as described in Sect. 2.9.

bind to GM1 ganglioside on cells might not be required. Therefore the binding sites of CT were blocked with GM1 ganglioside prior to incorporation into ME along with GtfB A:: PhoA antigen. Mice in this group had no antibody response to GtfEi.1::PhoA or to CTin either secretions or serum (Fig. 3, group 5 and data not shown).

mera was specific for GtfE3.1 peptide, which roughly corre- sponds to the contribution of the peptide to the overall chi- meric protein, i.e. about 5% of the amino acid composi- tion of the molecule.

3.6 Delivery of CT within multiple emulsions does not stimulate intestinal secretion

3.5 Response to GtfB.l chimeric subunit peptide

Previous studies found that the feeding of soluble GtfB.1::PhoA did not induce antibodies to the GtfEi.l peptide [lo]. Serum from mice fed with Gtfl3.1::PhoA plus CTin ME (Fig. 3, group 3) were pooled and analyzed for the quantity of antibody to the GtfB.1::PhoA chimera and to GtfB.l peptide alone. Serum IgG antibody to GtfE3.1::PhoA was 21.8 pg/ml, and to GtfB.1 peptide was 0.38 pg/ml. Thus, about 2 % of the antibody to the chi-

CT binds to intestinal epithelial cells and induces a potent fluid release [15] in a dose-dependent fashion [12]. Fluid accumulation after soluble CT was given i.g. was 6 after 2.5 pg/ml, 17 after 5 pg and 42 after 10 pg. In contrast, when CT was incorporated in ME, no fluid accumulation occurred, even with 50 pg, a fivefold higher amount of CT than that which stimulated immunity. A 200-pg dose of CT in ME resulted in fluid accumulation of 12, which is less than that resulting from 5 pg soluble CT.

Table 1. Protection of antigen within multiple emulsions during sequential treatment with HCI, trypsin and bile extract

Treatment Antigen recovered from

phase (%) phase (%) Outer water Antigen Total recovery (%) Inner water

Control CT 96 97 3 Control GtfB. 1 :: PhoA 88 92 8 HCI"' CT 47 100 0 HCI GtfB. 1 :: PhoA 55 100 0 HCI Trypsinb' CT 44 94 6 HCI Trypsin GtfB.1: : PhoA 38 100 0 HCl Trypsin Bile') CT 44 88 12 HCl Trypsin Bile GtfB.1::PhoA 26 100 0

a) ME (0.5 ml), prepared as described in Sect. 2.4, containing 100 pg C T and 500 kg of GtfB.l::PhoA, were adjusted to pH 2.2 by adding 50 mlO.1 N HCI, incubated 30 min at 37"C, then neutralized with 50 mlO.1 N NaOH. The antigen content in the inner and outer water phase was analyzed as described in Sect. 2.7.

b) Trypsin (10 mg) plus 10 mg chymotrypsin dissolved at the moment of use in 60 ml of 0.05 M NaHC03 was added to a sample treated with HCI as above and incubated for 30 min at 37°C. The enzymatic reaction was stopped by addition of 20 mg of soybean trypsin inhibitor. The antigen content in the inner and outer water phase was analyzed as described in Sect. 2.7.

c) Of 10 % bile extract (porcine) in PBS 70 ml was added to a sample treated with HCI, then trypsidchymotrypsin was added as above and incubated for 30 min at 37°C. The antigen content in the inner and outer water phase was analyzed as described in Sect. 2.7.

2724 M. Tomasi et al. Eur. J. Immunol. 1997.27: 2720-2725

3.7 Incorporation within ME protected proteins from

To determine whether incorporation within ME could pro- tect protein antigen and CT from digestion in the intestine, a series of sequential incubations was done in vitro to mimic the environment encountered in the intestine. After these incubations in acid, trypsin, and bile, the amount of immunoreactive Gtf13.1:: PhoA and CT was measured by ELISA in both the outer and inner aqueous phase (Ta- ble 1). Although the total recovery decreased somewhat with time and treatment, most of the immunoreactive GtfB.1::PhoA and CT within the inner aqueous phase remained detectable.

acid, trypsin and bile salts

4 Discussion

When CT was incorporated along with antigen into ME, strong secretory and serum antibody to the antigen was induced. The levels of response, quantitatively and quali- tatively, were similar to those found by gavaging soluble antigen plus CT as adjuvant in a buffer, which has been previously reported to induce good humoral immune responses at mucosal surfaces [16, 171. Thus, CT was able to serve as a mucosal adjuvant even when incorporated within these lipid particles. The exact mechanisms by which CT exerts its mucosal adjuvanticity remain unde- fined, although multiple mechanisms have been proposed [3]. One theory is based on the diffuse binding of CT to its ligand on surface epithelial cells when given in a soluble form. Such binding has resulted in production of certain cytokines in vitro [5] and in vivo [18] and an increased mucosal permeability which might facilitate immunization [6]. Under the conditions used in the present experiments, CT was unable to bind to epithelium as shown by the lack of fluid secretion after CT incorporated into ME was admi- nistered. Yet CT within ME was effective as a mucosal adjuvant. It follows, therefore, that the interaction of CT with the epithelium is not obligatory for its mucosal adju- vanticity.

An area of continuing interest and some controversy is the role of the two subunits in the adjuvanticity of CT. In the present studies, blocking of the binding sites on the CT-B subunit by GM1 ganglioside abolished the mucosal adju- vanticity of CT incorporated within multiple emulsions. Interestingly, the immunogenicity of CT was also lost (data not shown), suggesting that the adjuvanticity and immuno- genicity are linked. We have similar data, using mutant CT molecules, which contain a genetically inactivated B sub- unit but a wild-type A subunit [19]. This is a somewhat surprising result in that the CT delivered to PP within ME particles should be phagocytosed and internalized by APC in gut-associated lymphoid tissue (GALT), subserving what is thought to be the major function of the B subunit, i.e. delivering the A subunit into cells. These results sug- gest that CT incorporated in ME is released within GALT where it must be able to bind to the cell surfaces of APC and other lymphoid cells to exert its adjuvant effects.

In the last decade many strategies have been utilized for the development of effective oral vaccines, including the incorporation of antigen into liposomes [20], microspheres [21] and ISCOMS [22]. Some of these strategies require

passage of antigen through denaturing organic solvents [21] or chemical lipidation of the protein antigen [22]. Each of these strategies involves placing antigen in small particles that can be taken up by the follicle-associated epi- thelium of the mucosal immune system. Multiple emul- sions have several features that make them good candi- dates for an antigen delivery system for mucosal immun- ization. These include the ease and simplicity of prepara- tion, a high incorporation rate of antigen (>90 %), which is not dependent on the characteristics of the antigenic molecule, the absence of any organic solvents or denatura- tion, and components that are both non-toxic and biodeg- radable [9,23]. The oil phase of the ME system used here consists mainly of squalene, a lipid precursor of choles- terol that is nontoxic. A second major component was nonionic block copolymers which improve not only the stability of the emulsion, but also may contribute to adju- vanticity [24]. Lastly, as shown here, incorporation into ME is able to protect at least partially protein antigens and adjuvants from the harsh gut environment. The results obtained in the present study with ME containing CTcom- pare favorably to other strategies that have been developed for oral immunization. Delivery of antigen plus CT in ME reduced the toxicity of CT while maintaining its adjuvanticity. The immune responses that resulted were remarkably uniform. They were qualitatively and quantit- atively similar to that obtained previously with soluble CT. Lastly, ME are easy to prepare, inexpensive, stable, and will be able to incorporate a diverse variety of hydrophilic and hydrophobic antigens and immunomodulators in their native forms.

We gratefully acknowledge H. Bowdon for her expert technical assistance. M. T. is recipient of a fellowship from the Istituto Super- iore di Sanita’ “Progetto finalizzato alla lotta allHlDS”. This work is supported by NIH grants 2UOI AI 33231 and DK44240.

5 References

1 Mestecky, J., McGhee, J. R. and Elson, C. O., Immunol. Allergy Clin. North. Am. 1988. 8: 349.

2 McGhee, D. W., Elson, C. 0. and McGhee, J. R., Infect. Immun. 1993. 61: 4637.

3 Elson, C. 0. and Dertzbaugh, M. T., in Ogra, P. L., Mestecky, J., Larnm, M. E. , Strober, W., McGhee, J. R. and Bienen- stock, J. (Eds.), Handbook of Mucosal Immunology, Aca- demic Press, San Diego 1994, p. 391.

4 Levine, M. M., Kaper, J. B., Black, R. E . and Clernents, M. L., Microbiol. Rev. 1983. 47: 510.

5 McGee, D. W., Elson, C. 0. and McGhee, J. R., Infect. Immun. 1993. 61: 4637.

6 Lycke, N., Karlsson, U., Sjolander, A. and Magnusson, K. E. , Scand. J . Immunol. 1991. 33: 691.

7 Herbert, W. J., Lancet 1965. 2: 771. 8 Hunter, R. L., Olsen, M. and Buynitzky, S., Vaccine 1991. 9:

250. 9 Hunter, R. L., Olsen, M. R. and Bennett, B., in Stewart-Xll,

D. E. S. (Ed.), Theory and Practical Applications of Adju- vants, John Wiley & Sons, New York 1995, p. 52.

10 Dertzbaugh, M. T. and Elson, C. O., Infect. Immun. 1993. 61: 48.

11 Elson, C. O., Ealding, W. and Lefkowitz, J., J. Immunol. Methods 1984. 67: 101.

12 Richardson, S., Giles, J. C. and Kruger, K. S., Infect. Immun. 1984. 43: 482.

13 Russell, M. W., Brown, T. A., Radl, J., Haajman, J. J. and Mestecky, J., J. Immunol. Methods 1986. 87: 87.

Eur. J. Immunol. 1997.27: 2720-2725

14 Xu-Amano, J., Jackson, R. J., Staats, H. F., Fujihashi, K., Kiyono, H., Burrows, P. D. , Elson, C. O., Pillai, S. and McGhee, J. R., J. Exp. Med. 1993. 178: 1309.

15 Finkelstein, R., Crit. Rev. Microbiol. 1973. 2: 553. 16 Elson, C. 0. and Ealding, W., .I. Immunol. 1984. 133: 2892. 17 Lycke, N. and Holmgren, J., Immunology 1986.59: 301. 18 Klimpel, G. R., Asuncion, M., Haithcoat, J. and Niesel, D.

W., Infect. Immun. 1995. 63: 1134. 19 Elson, C. O., Tomasi, M., Chang, T.-T., Jobling, M. G. and

Holmes, R. K., FASEB J. 1995. 9: A290. 20 Michalek, S. M., Eldridge, J. H., Curtis 111, R. and Rosen-

thal, K. L., in Ogra, P. L., Mestecky, J., Lamm, M. E., Stro- ber, W., McGhee, J. R. and Bienenstock, J. (Eds.), Handbook of Mucosal ImmunoIogy, Academic Press, San Diego 1994, p. 391.

Multiple emulsion for oral immunization 2725

21 Eldridge, J. H., Staas, J. K., Meulbroek, J. A., McGhee, J. R., Tice, T. R. and Gilley, R. M., Mol. Immunol. 1991. 28: 287.

22 Mowat, A. M., Donachie, A. M., Reid, G. and Jarrett, O., Immunology 1991. 72: 317.

23 Hunter, R. L., Strickland, F. and Kezdy, F., J. Immunol. 1981. 127: 1244.

24 Hunter, R. L., Kidd, M. R., Olsen, M. R., Patterson, P. S. and Lal, A. A., J. Immunol. 1995. 154: 1762.