|Home | About | Journals | Submit | Contact Us | Français|
Studies of active immunization against Helicobacter pylori indicate that antibodies play a minor role in immunity. There is also evidence, however, that the translocation of antibodies in the stomach may be insufficient to achieve functional antibody levels in the gastric lumen. We have used a suckling mouse model of passive immunity to determine if perorally delivered antibodies can protect against infection with H. pylori. Female C57BL/6 mice were immunized parenterally with formalin-fixed cells of three clinical isolates of H. pylori (3HP) or the mouse-adapted H. pylori strain SS1 before mating. Their pups were challenged with the SS1 strain at 4 days of age and left to suckle before determination of bacterial loads 14 days later. Compared to age-matched controls, pups suckled by 3HP-vaccinated dams were significantly protected against infection (>95% reduction in median bacterial load; P < 0.0001). Pups suckled by SS1-vaccinated dams were also significantly protected in terms of both median bacterial load (>99.5% reduction; P < 0.0001) and the number of culture-negative pups (28% versus 2% for immune and nonimmune cohorts, respectively; P < 0.0001). Similar results were obtained with pups suckled by dams immunized with a urease-deficient mutant of SS1. Fostering experiments demonstrated that protection was entirely attributable to suckling from an immunized dam, and antibody isotype analysis suggested that protection was mediated by the immunoglobulin G fraction of immune milk. Analysis of the bacterial loads in pups sampled before and after weaning confirmed that infection had been prevented in culture-negative animals. These data indicate that antibodies can prevent colonization by H. pylori and suppress the bacterial loads in animals that are colonized.
Helicobacter pylori colonizes the gastric mucosa of humans and commandeers host defenses to establish chronic active gastritis while increasing the host's susceptibility to gastroduodenal ulceration or certain gastric malignancies (37). Although H. pylori induces profound systemic and mucosal immune responses, clearance of infection is infrequent, and there is no protection against reinfection following eradication by antimicrobial chemotherapy (15). Consequently, there are no obvious parameters of natural immunity on which to base effective vaccination strategies.
Vaccination studies of animal models have suggested that antibody development is not necessary for protective immunity to H. pylori (19) and may even enhance colonization (5, 6). Conversely, cellular immunity, possibly in concert with innate immune factors, such as defensins (59), elicits protection or eradication by exaggerating the gastric inflammatory response induced by H. pylori, thus interrupting colonization without a need to interact with the bacteria directly (3). The importance of the inflammatory response for protection against H. pylori is supported by the association of postimmunization gastritis with vaccine efficacy (6, 23). Nevertheless, the failure of antibody to limit H. pylori colonization is yet to be fully explained. One reason for this failure may be the relatively low level of antibodies in the gastric lumen due to the apparent inability of the mucosal immune system to translocate sufficient quantities of antibody across the gastric mucosa.
Although well characterized in the intestine, relatively little is known about antibody secretion into the stomach. Some studies of H. pylori infection have reported that levels of immunoglobulin A (IgA) in gastric juice are significantly lower than those found in the saliva or intestinal contents (33, 34). Evidence that these low levels of IgA are due to inadequate antibody secretion in the stomach includes the following: (i) H. pylori-specific antibodies in gastric juice of infected individuals are predominantly nonsecretory IgA (10); (ii) equivalent amounts of IgG and IgA in the stomach suggest that IgA may leak across the mucosa rather than being actively secreted (14, 18); and (iii) much of the secretory IgA (sIgA) in the stomach is derived from swallowed saliva (17, 54). In addition, compared to the small intestine, the normal mammalian stomach has barely detectable secretory component (SC), suggesting a limited capacity for translocation of polymeric IgA across the gastric mucosa (8, 33). Moreover, despite considerable upregulation of SC by gamma interferon following the development of gastritis, there is no corresponding increase in the concentration of sIgA in gastric juice (4). Consequently, the concentration of sIgA in the stomach is unlikely to be sufficient to prevent or eradicate colonization by H. pylori.
On the other hand, there is evidence that passive immunization with antibodies delivered perorally may reduce the extent of gastric colonization by Helicobacter species. This therapeutic approach has shown some promise in adult mice given monoclonal IgA or hyperimmune bovine colostrum against Helicobacter felis (14, 41) or urease-specific, chicken-derived IgY against H. pylori (44). In addition, reports of delayed acquisition of H. pylori by Gambian infants that corresponded to their mothers’ levels of breast milk IgA specific for H. pylori (58) and the protection of infant mice against full colonization by H. felis while suckling from immunized dams (13) suggest that orally delivered antibodies may be beneficial in controlling gastric Helicobacter infections. Despite these favorable reports, there are no tightly controlled studies that conclusively show prevention of H. pylori infection by orally delivered immune antibodies in the absence of additional factors, such as famotidine (44). Moreover, no studies have investigated the refinement of vaccine preparations for use in the production of anti-H. pylori polyclonal antibody products.
In this study, we used a suckling mouse model of infection to investigate whether H. pylori-specific antibodies delivered during lactation to the gastric lumen of infant mice can protect against H. pylori infection. The route and adjuvant used to immunize the dams were selected to evoke an immune response similar to that required for the production of commercial quantities of polyclonal monomeric antibodies, such as from hyperimmune bovine colostrum. The model allowed us for the first time to quantify the contribution of passively acquired H. pylori-specific antibodies to protection against infection and provided an opportunity to examine different vaccine preparations for their ability to elicit these antibodies.
H. pylori clinical isolates CHP1, CHP2, and CHP3 (all VacA m1/s1a, CagA+), and the mouse-adapted H. pylori strain SS1, were routinely maintained under microaerophilic conditions on Dent plates or in brain heart infusion broth (BHIB; Oxoid, Basingstoke, United Kingdom) supplemented with 5% (vol/vol) fetal bovine serum (JHR Biosciences, Lenexa, KS) and Dent selective supplement (Oxoid), as described previously (22). A urease deletion mutant of SS1 (SS1Δure), SS1ΔureAB::aphA3, was created by replacing a region spanning ureB, ureA, and their promoter with a kanamycin resistance gene, aphA3, using the Apo1 sites corresponding to bases 76077 and 78150 of the genome of H. pylori 26695 (GenBank accession no. AE000511) (22). SS1Δure was routinely cultured in the presence of 20 μg/ml kanamycin sulfate (Roche Diagnostics, Mannheim, Germany). For construction of SS1Δure, the region of the H. pylori 26695 genome encompassing the genes encoding urease subunits A and B and their promoter (bases 75518 to 78904, inclusively) was cloned into pGEM-T Easy (Promega, Madison, WI). This construct, pURE, expressed substantial levels of both urease subunits when transformed into Escherichia coli DH5α and was used as a source of recombinant antigen for serological studies. E. coli DH5α containing pURE or empty pGEM was cultured in Luria broth or agar supplemented with 100 μg/ml ampicillin (CSL, Parkville, Victoria, Australia).
C57BL/6 mice were bred and housed in the Department of Microbiology and Immunology animal facility certified as specific pathogen free for all major murine pathogens but positive for Helicobacter bilis. In addition to routine testing for mouse pathogens (Cerberus Sciences, Thebarton, South Australia, Australia), specific testing for Helicobacter species using the protocol of Bohr et al. (11) confirmed the presence of H. bilis and showed mice in this facility to be free of other Helicobacter species. Adult mice were maintained on Barastoc laboratory rat and mouse pellets (Ridley AgriProducts, Pakenham, Victoria, Australia) and acidulated water to minimize transmission of SS1. Approval for all of the animal studies was obtained from the University of Melbourne Animal Experimentation Ethics Committee.
For the passive immunization model reported here, female mice were hyperimmunized with formalin-fixed whole cells of various H. pylori strains prior to mating. Vaccines were administered subcutaneously with a CpG/alum adjuvant that was selected for the following two reasons: (i) unlike Freund's complete adjuvant, CpG adjuvants have no toxic side effects that may interfere with reproduction, and (ii) when delivered subcutaneously, CpG-based adjuvants are potent stimulators of B-cell activation and mimic the strong IgG response required for an antibody-based product derived from bovine colostrum.
Offspring obtained after mating these vaccinated mice were challenged with H. pylori SS1 while receiving orally delivered H. pylori-specific antibody in their mothers’ milk. The suckling pups’ colonization levels were determined 2 weeks after infection and compared to those of pups suckling from nonimmune dams that had been infected with SS1 at the same time. Immune litters born on the same day shared a single age-matched control litter and all received the same SS1 cell suspension. Protection was investigated by comparing the median bacterial load of individual or combined test litters to that of control litters and also by comparing the number of protected pups in each litter. In this study, an individual pup was defined as protected if its bacterial load was less than 1% of the median bacterial load of the matched control litter.
For immunization of mice, plate-cultured CHP1, CHP2, and CHP3 (equivalent numbers of CFU for each strain) or BHIB-grown H. pylori SS1 or SS1Δure were washed twice with phosphate-buffered saline (PBS), fixed with 4% formalin for 3 days at 4°C, washed three times, and resuspended in PBS. The protein concentration of each antigen preparation was determined before formalin fixation by using the bicinchoninic acid assay (Sigma-Aldrich, St. Louis, MO). Four-week-old virgin female C57BL/6 mice received three subcutaneous injections of antigen (20 μg protein) at 14-day intervals together with ImmunEasy CpG mouse adjuvant (Qiagen Pty. Ltd., Clifton Hill, Australia) in final volumes of 50 μl, as recommended in the ImmunEasy protocol. Immunized mice were mated within 7 days after the third dose. Several serum samples were collected from the dams to monitor seroconversion/prevaccination, 10 days postboost, immediately postpartum (for most dams), at weaning, and 2 to 4 weeks after weaning. Only the results for serum samples collected at weaning are presented in this study. Nonimmune control mice, matched for age, gravidity, and parturition date, were either simultaneously sham-immunized mice or naïve animals from the animal facility breeding program. Preliminary studies showed that there were no differences in the bacterial loads of pups suckling from sham-immunized or unimmunized dams.
Suspensions of H. pylori SS1 used to infect mouse pups were prepared by inoculating 20 ml prewarmed brucella broth (Becton Dickinson, Sparks, MD), containing 5% fetal bovine serum and Dent selective supplement, with a freshly thawed aliquot of approximately 5 × 107 CFU of H. pylori SS1. After 48 h of incubation with shaking under microaerophilic conditions at 37°C, the cells were pelleted and resuspended at a concentration of 108 CFU/ml in antibiotic- and serum-free BHIB containing 0.003% (wt/vol) Evans blue (Sigma-Aldrich) as a tracer dye. Previous experiments have shown that Evans blue at this concentration does not affect the viability of H. pylori. Pups born to vaccinated and control dams were inoculated 4 days after birth with a 75-μl bacterial cell suspension, containing approximately 7.5 × 106 CFU, by gavaging, using a stainless steel neonatal mouse feeding needle (22 gauge, 1.5 inch straight; Cole-Parmer, Vernon Hills, IL), and then returned to their dams to suckle freely. Suckling mice were killed 14 days later by asphyxiation with CO2/O2. The number of litters and pups used in each experiment are shown in Table Table11.
The stomach of each mouse challenged with H. pylori was dissected along the border of the nonglandular region. The complete gastric contents from both the glandular and nonglandular regions were recovered into 0.5 ml PBS containing 0.05% (vol/vol) Tween 20 (PBS-T0.05) and used to determine the antibody content and specificity in maternal breast milk. The glandular region was opened along the greater curvature, washed twice in saline, blotted on sterile filter paper, weighed, and homogenized in 2 ml physiological saline for 5 s using a Polytron PT 10-35 homogenizer fitted with a PTA 10TS probe (Kinematica, Inc., Lucerne, Switzerland). Between each sample used, the probe was rinsed, disinfected with 70% ethanol, and chilled with sterile saline. Bacteria were enumerated by spreading 200 μl of undiluted and serial 10-fold dilutions of each homogenate on Dent plates supplemented with 20 μg/ml bacitracin (Sigma-Aldrich). For some cohorts, the stomach was further bisected along the lesser curvature before being washed, and half of the glandular region was saved for formalin fixation. For graphical presentation in this study, bacterial counts are provided as the number of CFU per tissue sample (half or whole stomach) to allow visualization of the detection limit for each experiment. Samples which showed no growth of H. pylori were assigned an arbitrary value of 1 CFU less than the detection limit before adjustment for weight, and comparative statistical analyses were conducted on weight-adjusted data (number of CFU/g tissue) that had been log10 transformed. The only exception was the weaning experiment, in which culture-negative animals were excluded to focus the analysis on the change in the infectious burden of infected animals before and after weaning. As such, the detection limit was irrelevant, and all data are presented as the weight-adjusted bacterial load. Significant reduction in the median bacterial load of individual litters, or all litters combined, was determined by the Mann-Whitney rank sum test.
Gastric culture, as described for suckling mice, was also performed on some dams (H1 to H3 and S1 to S5). Of these animals, only dams H2 and S2 were positive for H. pylori. Both of these mice showed IgA seroconversion in serum samples collected 2 to 4 weeks after weaning compared to those at weaning, i.e., above those resulting from hyperimmunization. As infection correlated directly with IgA seroconversion, the remaining dams were examined serologically for acquisition of SS1 infection from their pups, but no other seroconversions were identified.
Tubes containing gastric contents from infant mice in preweighed PBS-T0.05 were weighed and mixed on a shaking platform at 200 rpm at 4°C for 3 h with occasional vortexing. Particulate debris was removed by centrifugation (5 min at 13,000 × g), and the supernatants were stored at −20°C. Total IgA and IgG levels in the gastric contents of infant mice and sera from their dams were determined by using mouse IgA or IgG enzyme immunoassay (EIA) kits (Bethyl Laboratories, Montgomery, TX). Briefly, goat anti-mouse IgA or IgG bound to microtiter plates (F96 MaxiSorp; Nunc, Denmark) was used to capture antibody from gastric contents or serum samples at 37°C for 3 h, after which goat anti-mouse IgA or IgG conjugated to horseradish peroxidase was used to detect bound antibody. Antibody concentrations were read off a standard curve generated by using pooled mouse sera calibrated for each antibody isotype (Bethyl Laboratories).
Milk and serum samples were assayed for H. pylori antibodies by EIA, as follows. Milk samples were diluted in dilution buffer (1.2% casein in PBS-T0.05) either 1 in 4 or 1 in 10 for preliminary assays or standardized to 100 ng/ml IgA or IgG for measuring specific antibody isotypes. Microtiter plates were coated overnight with carbonate buffer (pH 9.6) containing 0.5 μg/ml whole-cell lysate of H. pylori (prepared by freeze-thawing washed cells of H. pylori SS1 three times and storing them in single-use aliquots), or 10 μg/ml recombinant urease antigen prepared from clarified supernatants of overnight cultures of sonicated E. coli DH5α(pURE). Plates were then blocked with 100 μl of 1.2% casein in PBS for 1 h at ambient temperature, after which diluted test samples, 50 and 100 μl for milk and sera, respectively, were added to duplicate wells (37°C for 3 h). After being washed, the wells were further incubated with peroxidase-conjugated goat anti-mouse IgG (whole IgG) or IgA (Fcα) in dilution buffer (1 in 5,000 and 1 in 1,500, respectively; KPL, Gaithersburg, MD) at 37°C for 1 h. Antigen-specific IgG or IgA was detected with TMB substrate (KPL). The resulting optical densities (ODs) were converted as follows: for preliminary analysis, the OD was adjusted for the dilution factor and, in the case of gastric contents, bolus weight. For isotype-standardized assays, the ODs representing EIA units per 5 ng of IgG or IgA were used directly or converted to EIA units per stomach per gram of gastric contents for comparative analysis. Dams’ sera were also assayed by EIA for IgG directed against outer membrane antigens by coating microtiter plates with outer membrane vesicles (OMV) recovered from 0.22-μm-filtered, spent SS1 culture medium, ultracentrifuged at 150,000 × g for 3 h.
Blood samples were collected from the portal vein of pups suckled by SS1-vaccinated and matched control dams at the time of death. The EIA was essentially the same as that used for the gastric contents, except that all volumes were doubled, the concentration of casein was 1% throughout, and peroxidase-conjugated goat anti-mouse IgG (Fcγ) was used (1 in 100,000; KPL). A standard curve using pooled serum samples from terminal bleeds of five dams (S1 to S5) was generated for each assay. This serum, which had an end-point titer of 1 in 64,000 (the next dilution after OD at 450/620 nm [OD450/620] was 0.100), was assigned an arbitrary value of 1,000 units and used to allocate units of reactivity to the unknown samples. Serum samples were doubly diluted (from 1 in 1,000 for immune-suckled and 1 in 200 for weaned or nonimmune-suckled pups), and ODs falling within the linear range of the curve were assigned units which subsequently were adjusted for their dilution factor and averaged for each sample. For negative samples (where the highest OD was below the lower linear limit of the curve), units were assigned from the nonlinear region.
Recombinant urease and E. coli control antigen for this assay were prepared from clarified supernatants of overnight cultures of sonicated E. coli DH5α(pURE) or DH5α(pGEM), respectively, and used with 10 μg/ml protein to coat microtiter plates. The assay protocol was essentially the same as that for the serum SS1-specific IgG assay, with the following modifications: (i) in addition to duplicate DH5α(pURE) wells, a well containing DH5α(pGEM) was included for every sample and control, thus allowing subtraction of background signals; (ii) the pooled sera (S1 to S5) used to generate the standard curve had an end-point, urease-specific titer of 1 in 6,400 and were assigned a value of 200 units; (iii) only two dilutions of each test serum sample were used (1 in 200 and 1 in 1,000); and (iv) a positive control of the standard pooled sera was included in each assay. This control serum sample gave, on average, 195 ± 13 units (mean result ± 95% confidence interval [95% CI]) in eight independent assays.
Total membrane and soluble fractions of H. pylori SS1Δure were prepared by ultracentrifugation of sonicated, washed whole cells. Membrane and soluble fractions were separated independently using Novex 4 to 12% Bis-Tris gels (Invitrogen, Carlsbad, CA) and then transferred onto polyvinyl-difluoride membranes, according to standard NuPage protocols (Invitrogen). Strips of membrane were blocked with 10% skim milk powder in PBS and incubated at room temperature for 2 h and then at 4°C overnight with pooled gastric contents from infant mice (1.5 ml, diluted 1 in 25; 0.26 μg total IgG and 0.59 μg total IgA) or pooled dams’ sera (3 ml, diluted 1 in 500; 14.26 μg total IgG and 1.28 μg total IgA). The antibody diluent used was 5% skim milk powder in PBS-T0.01. Blots were washed with PBS-T0.01 and then incubated with peroxidase-conjugated goat anti-mouse IgA (diluted 1 in 10,000 in antibody diluent; KPL) at room temperature for 1 h, washed again, and incubated with ECL substrate (Amersham, Buckinghamshire, United Kingdom) before exposure to Hyperfilm (Amersham). After various exposure times were used, the same blots were washed, incubated with peroxidase-conjugated goat anti-mouse IgG (1 in 25,000; KPL) as done for the anti-IgA conjugate, and washed, and the signal was captured using ECL substrate and Hyperfilm.
Statistical analysis of quantitative data was performed using Prism version 4.02 or InStat version 3.06 (GraphPad Software, Inc., San Diego, CA). P values of <0.05 were considered significant.
As there is no clearly recognizable antibody-mediated immunity following H. pylori infection, there are no specific antigens that stand out as essential inclusions in antibody-based treatments. For this reason, we randomly selected three clinical isolates from patients with symptomatic H. pylori infection for hyperimmunization of the initial cohort of mice. Three female mice (H1 to H3) were immunized with a cocktail of the three low-passage clinical isolates of H. pylori (3HP) and mated within a week of the final immunization. To determine if immunized dams could protect their offspring against colonization with H. pylori, their 4-day-old pups were challenged with H. pylori SS1 and examined for determination of their bacterial loads 14 days later. Significantly fewer bacteria were recovered from the stomachs of pups suckled by dams immunized with heterotypic H. pylori strains than from pups suckled by nonimmune dams (Fig. (Fig.1A),1A), with weight-adjusted median bacterial loads of 104.8 (interquartile range [IQR], 103.3 to 105.3) and 106.3 (IQR, 105.9 to 106.6) CFU/g, respectively (P < 0.0001; Mann-Whitney rank sum test, two tailed). However, the number of pups protected against full colonization by SS1 was not significantly different from that of the controls (relative risk [RR], 0.68; 95% CI, 0.47 to 0.99; P, 0.1; Fisher's exact test, two tailed) (Table (Table11).
To determine to what extent immunized dams could protect their offspring against colonization with a homotypic strain of H. pylori, five female mice (S1 to S5) were immunized with a vaccine prepared from cultures of H. pylori SS1, and their 4-day-old pups were challenged with H. pylori SS1 and investigated for determination of bacterial loads 14 days later. Age-matched litters from four nonimmune dams were used as controls.
Immunization with SS1 whole cells provided highly effective protection against colonization by the homotypic strain, with median weight-adjusted bacterial loads of 103.6 (IQR, 102.5 to 105.5) and 106.2 (IQR, 105.7 to 106.4) CFU/g in the pups of immune and nonimmune dams, respectively (P < 0.0001; Mann-Whitney rank sum test) (Fig. (Fig.1B).1B). Importantly, a significant number of pups suckling from immune dams had no detectable colonization compared with pups suckling from nonimmune dams (RR, 0.61; 95% CI, 0.44 to 0.82; P = 0.001; Fisher's exact test) (Table (Table1).1). When pups with bacterial loads of less than 1% of those of the matched control mice were included in this comparison, the homotypic vaccine had an overall protective efficacy of 61% (95% CI, 36 to 74%), with 19 of 31 immune pups protected compared with 1 of 25 nonimmune pups (RR, 0.38; 95% CI, 0.24 to 0.62; P < 0.0001; Fisher's exact test) (Table (Table11).
To compare the contribution of maternally derived breast milk antibodies and transplacental antibodies to passive immunity against H. pylori, pups were fostered between immune and nonimmune dams in a variation of the experimental protocol. For this study, two female mice (S6 and S7) were immunized with formalin-fixed SS1 whole cells and mated. Each resulting litter was divided at 2 days of age so that half the pups remained with the immune dam and half were fostered onto a nonimmune dam matched for parturition date. Pups from each corresponding nonimmune dam were reciprocally fostered in the same manner onto the immune dams so that each of the four dams (two immune and nonimmune each) suckled a mixture of biological and adopted offspring. At 4 days of age, i.e., after suckling for 2 days from their treatment dam, all pups were challenged with SS1 and left to suckle a further 14 days before they were killed for determination of bacterial loads.
Overall, pups suckled by immune dams had significantly reduced median bacterial loads compared to those suckled by nonimmune dams (103.4 [IQR, 103.2 to 104.4] versus 105.5 [IQR, 104.9 to 106.2] CFU/g; P < 0.0001; Mann-Whitney rank sum test), and this was also significant after consideration of the immune statuses of their birth mothers (overall, P of <0.0001; two-way analysis of variance [ANOVA]) (Fig. (Fig.2).2). The immune status of the birth dam did not influence bacterial load (P of 0.4; two-way ANOVA), and comparison of biological and adopted offspring showed no statistical interaction (P of 0.25; two-way ANOVA). Overall, vaccinated dams protected their adopted and biological pups to a degree similar to those of 6 of 9 adopted pups and 5 of 9 biological pups (11 of 18 in total) being protected against normal infection (Table (Table1).1). Conversely, all pups suckled by nonimmune dams were normally colonized, and pups born to immune dams lost any protective advantage once fostered onto nonimmune dams.
One of the immune dams included in the fostering experiment (S6) delivered a second litter of pups. These pups remained with the immune dam for the entire suckling period and were challenged with SS1 at 4 days of age. Seven of the nine pups in the second litter were protected (Table (Table1)1) to an extent that was indistinguishable from the protection she conveyed to her first litter containing both biological and adopted offspring (8 of 10 pups protected).
We found a substantial difference in the degree of protection conveyed by dams immunized with homotypic H. pylori strains compared with that of heterotypic H. pylori strains. In addition, we observed that protected pups appeared to cluster with certain dams. More specifically, all 13 culture-negative pups in the SS1-immune cohort belonged to only three of five litters from SS1-immunized dams (S1, S2, and S4). When pups with significantly reduced bacterial load were included, two of these dams (S1 and S4) protected 100% of their pups. Similar observations were made with the 3HP cohort, when five of the seven pups suckled by dam H1 were protected against full colonization, including two that were culture negative. This was significantly greater protection than the protection conveyed by the remaining two 3HP immune dams (2 of 12 pups protected; P of 0.045, Fisher's exact test). These findings suggested that discernible differences in the antibody response of vaccinated dams may correlate with protective efficacy. Accordingly, we quantified the dams’ serum antibody response to antigens from SS1 whole cells, SS1 OMV, and recombinant urease A and B subunits.
Total SS1-specific, SS1 OMV-specific, and urease-specific serum IgG levels were substantially higher in SS1-vaccinated dams than in 3HP-vaccinated dams (Fig. (Fig.3A).3A). Moreover, urease-specific EIA suggested that a large proportion of the OMV-reactive IgG from 3HP mice recognized urease within this antigen preparation, except for that of dam HP2, which also had substantial SS1 lipopolysaccharide (LPS)-reactive IgG as determined by immunoblot analysis (results not shown).
Anti-SS1-specific IgG in milk samples recovered from the pups at the time of death was analyzed in a similar manner (Fig. (Fig.3B).3B). In general, SS1-vaccinated dams secreted substantially more SS1-reactive IgG into their milk than 3HP-vaccinated dams, and this seemed to be associated with the protective efficacy of the milk from individual dams. However, while each of the 3HP dams significantly reduced the median bacterial load of their litters (HP1, <0.5% than that of matched controls, P value of 0.003; HP2 and HP3, <5% than that of matched controls, P values of 0.02 and 0.01, respectively; Mann-Whitney rank sum test), this was not the case for SS1-vaccinated dams S3 and S5. Therefore, it is likely that both the quantity and specificity of milk antibodies contributed to overall efficacy.
Preliminary analysis of breast milk antibodies suggested a correlation between levels of milk IgG and protection. However, the contribution of the predominant murine breast milk antibody isotype IgA was not included in this analysis. The most practical procedure for collecting milk samples in this study was to collect the bolus of gastric contents from each pup at the time of death. This is because collecting milk directly from the mammary glands of lactating dams requires an extended period of segregation from their pups which would have interrupted the pups’ treatment and carried the risk of them being rejected upon their reintroduction to their dams.
The use of gastric contents as a milk proxy, however, may have introduced some variables into the samples, including dissimilar bolus volumes and variable degrees of gastric protease activity, depending on when the pup had last suckled. Moreover, while we generally recovered a white, cheese-like clot of milk from the stomachs of pups sampled at 18 days of age, some animals showed evidence of also ingesting food. As the influence of these and various other factors on the quality of the milk samples was unpredictable, the total amounts of IgA and IgG in each sample were determined as reference points for standardization of the samples before reanalysis. IgA was the dominant immunoglobulin isotype in all samples and was, on average, twice as abundant as IgG (62 ± 5.2 and 28 ± 2.5 μg/g, respectively [mean result ± standard error of the mean result]). This proportion of IgA to IgG is in agreement with published data on breast milk from C57BL/6 mice, which suggested that these samples were suitable for comparison of the relative contribution of milk IgA and IgG to protection in this model (42).
All milk samples from pups suckled by SS1-vaccinated dams were standardized at 100 ng IgA or IgG/ml prior to EIA for anti-SS1 IgA or IgG. Antibody levels in similarly standardized serum samples from each immune dam were assayed in parallel for comparison. As shown in Fig. Fig.4,4, very little of the milk IgA was specific for H. pylori, and few samples had reactivity above background levels. This was despite two dams, S1 and S4, having moderate amounts of IgA specific for SS1 in their serum samples. Conversely, the same milk samples had abundant H. pylori-specific IgG at levels approximating those in the dams’ sera. Assays for urease-specific IgG and IgA in the gastric contents showed a similar relationship, with urease-specific IgA levels just above background contrasting with abundant urease-specific IgG in milk from dams S4, S3, and S2 (in order of abundance; data not shown). Milk from dams S1 and S5 had barely detectable and no urease-specific IgG, respectively, and neither had detectable urease-specific IgA.
To investigate the disparity between circulating and secreted levels of H. pylori-specific IgA further, serum and milk samples were analyzed by immunoblotting for IgA and IgG reactive for SS1 antigens other than urease. For this analysis, milk samples were pooled from two pups each from litters S2, S4, and S5 to achieve 30 and 15 times the amount of IgA and IgG used in each well of the EIA analysis, respectively. Serum samples from immune dams S2, S4, and S5 were also pooled and diluted to give around twice the total IgA and 50 times the total IgG of the pooled milk sample. SS1Δure-reactive IgG was readily detectable in both serum and milk samples and recognized a broad range of antigens (Fig. (Fig.5,5, lanes 1 and 2, respectively). However, SS1Δure-reactive IgA was detectable only in the serum samples, despite competing with the abundant IgG population, and was primarily directed against LPS (Fig. (Fig.5,5, lane 3). This specificity was confirmed by immunoblot analysis against purified LPS (not shown). Additional faintly IgA-reactive bands in both serum and milk samples did not increase in signal strength after overexposure of the chemiluminescent substrate, suggesting that they were nonspecific background signals. Immunoblot analysis of milk samples against soluble SS1Δure antigens also showed SS1Δure-specific IgG but no detectable reactive IgA (not shown). Taken together, these data suggest that despite IgA being the dominant immunoglobulin isotype in murine milk, these antibodies have limited specificity for H. pylori antigens following parenteral vaccination of the dams by the method we used. As such, it is most unlikely to be the effector of protection observed in this study. This contrasts with the lesser IgG population in the milk, which showed broad reactivity against H. pylori surface antigens and hence is likely to have contributed to protection against H. pylori infection.
While it is possible that antibody-mediated agglutination may have contributed to protection, agglutination may also have caused artifacts in culture-based determination of bacterial load. For this reason, we determined if the amount of antibody in the pups’ stomachs at the time of death influenced the bacterial load. Accordingly, IgG levels in all milk samples were corrected to give the amount of anti-SS1 IgG per stomach and per gram of milk and compared by correlation analysis against the bacterial load in each pup. Although there was no significant correlation between bacterial load and the total amount of specific IgG in the stomach at the time of death (r, −0.15; 95% CI, −0.49 to 0.23; P of >0.1; Spearman's rank order correlation), there was a significantly negative correlation with the proportion of specific IgG in the milk (r, −0.37; 95% CI, −0.65 to −0.004; P of 0.04; Spearman's rank order correlation). This suggests that while the total amount of SS1-specific antibody in the stomach at the time of death did not significantly influence the bacterial load data and conversely, the bacterial load did not influence the level of detectable SS1-specific IgG, the overall concentration of H. pylori-specific IgG in milk may reflect its protective efficacy.
Because IgG is absorbed from the small intestine of mice during the first 16 days of life, serum samples from infant mice contain the IgG profile of the milk ingested over the entire suckling period, as opposed to the analysis described above, which examined milk from day 18 of lactation. Accordingly, we assayed serum samples from SS1-challenged pups for SS1- and urease-specific IgG using standard curves generated from the pooled sera of immune dams. The comparative levels of SS1-specific IgG in pups’ sera were similar to those observed in gastric contents and were highest in sera of pups suckling from the highly protective dams S1 and S4 (Fig. (Fig.6).6). Interestingly, we found no correlation between levels of SS1-reactive IgG and levels of urease-specific IgG. Comparative levels of urease-specific IgG determined in pups’ sera were essentially the same as those detected in the milk samples but different from those detected in dams’ sera. This result verified the levels obtained by assay of gastric contents, suggesting that dams’ SS1-specific serum IgG was unequally secreted into the milk. All pups suckled by nonimmune dams had no detectable H. pylori-specific IgG.
Serum samples from 18-day-old pups fostered between immune and nonimmune dams showed SS1-specific IgG levels reflecting their period of nursing from an immune dam (Fig. (Fig.7A).7A). In general, pups suckled by the same immune dam from 2 days of age had similar SS1-specific antibody levels, regardless of whether their birth dam was immune or nonimmune, and pups born to and nursed by nonimmune dams had no detectable H. pylori-specific IgG. However, pups born to immune dams and nursed by nonimmune dams from 2 days of age still had detectable levels of specific IgG 16 days after removal from their birth dam. Despite this circulating H. pylori-reactive IgG, which would have been substantially greater at the time of challenge because maternal IgG has a half-life of around 7 days in mice (38), these pups were as susceptible to H. pylori as pups nursed solely by nonimmune dams. In addition, dams S6 and S7 were quite disparate in their ability to convey protection (80% versus 37% of pups protected, respectively), and although pups from dam S6 had slightly higher SS1-reactive IgG levels than those from dam S7, the opposite was true of urease-specific IgG (Fig. (Fig.7B).7B). Taken together, these results suggest that while overall H. pylori-specific IgG levels may correlate with protection, this was not the case for urease-specific antibody, which appeared to be of minor importance for protection against H. pylori infection by orally delivered antibodies.
To determine if urease-specific antibodies were necessary for passive immunity against H. pylori, five mice (U1 to U5) were immunized with a formalin-fixed mutant of H. pylori SS1 lacking both the A and B subunits of urease (SS1Δure). The absence of urease-specific antibodies in the blood samples and milk of mice immunized with this strain was confirmed by immunoblotting and EIA (data not shown). Age-matched litters from three nonimmune dams were used as controls for each challenge inoculum of SS1. Female mice immunized with SS1Δure conveyed a high degree of protection to their pups against colonization by wild-type H. pylori SS1 (Fig. (Fig.8).8). The overall median bacterial load of pups suckled by SS1Δure-vaccinated dams was >2 log lower than that of pups suckled by nonimmune dams (103.6 [IQR, 102.5 to 105.1] and 106.1 [IQR, 105.4 to 106.4] CFU/g, respectively; P of <0.0001; Mann-Whitney rank sum test). These results were indistinguishable from those of the cohort vaccinated with wild-type SS1 (Fig. (Fig.1B).1B). Overall, the SS1Δure vaccine was highly effective, with 58% of pups suckled by immune dams protected against full H. pylori colonization compared with none of the pups suckled by nonimmune dams protected (Table (Table1).1). While suckling from immunized dams, the pups had a significantly reduced risk of acquiring a full infection (RR, 0.42; 95% CI, 0.27 to 0.66; P of <0.0001; Fisher's exact test) or indeed any detectable infection (RR, 0.62; 95% CI, 0.45 to 0.83; P of 0.0007; Fisher's exact test). These levels of protection were indistinguishable from those in pups whose dams were immunized with wild-type SS1. As with the latter, the distribution of protected pups was uneven among the immune litters, in that dams U2 and U3 protected all nine of their pups, including seven that were culture negative.
Analysis of H. pylori-specific IgG and IgA in milk samples showed results comparable to those from mice immunized with wild-type SS1, insofar as IgG was readily detectable by EIA and immunoblotting against membrane and soluble fractions of SS1; IgA was LPS specific in dams’ sera and undetectable in milk (data not shown). In contrast to the wild-type cohort, however, there was an insignificant positive correlation between the proportion of H. pylori-specific IgG in milk and the bacterial loads of the pups (r, 0.21; 95% CI, −0.021 to 0.56; P of 0.31; Spearman's rank order correlation). The irrelevance of the overall antibody level was verified by assay of pups’ sera for H. pylori-specific maternal IgG, which showed that pups suckled by dams U2 and U3 had levels of IgG similar to those of pups suckled by less-protective dams (Fig. (Fig.9).9). Together, these data suggested (i) that urease-specific antibodies are not required for the efficacy of orally delivered antibodies against H. pylori and (ii) that there may be specific qualitative rather than general quantitative differences in the composition of milk antibodies that correlate with protection.
Three of the dams vaccinated with SS1Δure (U2, U3, and U4) had second litters which, along with three age-matched nonimmune litters, were challenged with SS1. In this experiment, approximately half the pups in each litter were killed at 18 days of age for determination of bacterial load (14 and 11 immune and nonimmune pups, respectively; detection limit of 10 CFU/entire glandular stomach). The remaining pups stayed with their dams until weaning 3 days later, after which they were maintained on standard food and autoclaved, nonacidulated water for an additional 11 days before being killed for determination of bacterial loads (28 days postinfection [dpi]; 12 and 13 immune and nonimmune pups, respectively). The bacterial loads of mice from the same litters killed before or after weaning from immune and nonimmune dams were then compared to determine the impact of treatment withdrawal on bacterial load.
Again, dam U3 was highly protective, and all of her nine pups were culture negative, regardless of whether they were killed before or after withdrawal of orally delivered antibody (five and four pups, respectively). This suggested that pups which were culture negative when sampled during the treatment period were indeed uninfected. In contrast, all nonimmune pups sampled during the suckling period were culture positive. In agreement with our earlier findings, the median bacterial load of infected pups suckled by immune dams was significantly reduced compared to that of pups suckled by nonimmune dams (104.6 [IQR, 104.1 to 105.3] versus 105.4 [IQR, 105.3 to 106.3], respectively; P of 0.008; Mann-Whitney rank sum test). However, the reduced infectious burden in pups suckled by immune dams was completely lost after weaning, i.e., there was no difference in bacterial load between pups from immune dams and those from nonimmune dams 11 days after weaning (P > 0.05) (Fig. (Fig.10).10). The increase in the bacterial loads of all pups after weaning was significant, regardless of the immune status of their dam (P < 0.0001). This suggests that a milk diet on its own may have a significant impact on H. pylori colonization of mice but is not sufficient on its own to prevent infection.
Serum samples from pups weaned from immune dams were compared to those of suckling pups killed while on an immune milk diet. Weaned pups had substantial amounts of SS1-specific IgG but at levels six- to eightfold lower than those of their siblings killed before weaning (Fig. (Fig.11),11), which in turn showed levels equivalent to suckling pups from the first litter of each dam. These findings suggest that orally ingested maternal IgG ceased to contribute to protective efficacy in this model once translocated into the pups’ sera. As expected, pups weaned from nonimmune dams had no detectable SS1-specific circulating IgG. Moreover, no pup, before or after weaning from immune or nonimmune dams, had any detectable urease-specific IgG (not shown). These data indicate that immunity in this model was entirely passive, as the infected pups showed no evidence of developing their own humoral response within the time periods examined in these experiments, even 4 weeks after infection.
In this study, we used a suckling mouse model to evaluate the efficacy of passive immunization against H. pylori. We found that 45 of 75 (60%) mice which had suckled from dams immunized with the wild type or a urease-deficient mutant of H. pylori SS1 were highly protected against infection with the same strain (bacterial loads less than 1% than that of control mice). Importantly, 24 (32%) of these pups showed no detectable colonization compared with 1 of 67 (1%) in the control group. Even heterotypic antibodies produced by immunizing mice with three randomly selected clinical isolates of H. pylori led to a highly significant reduction in the bacterial loads of SS1 in pups ingesting these antibodies.
The immunization protocol we used for this study was specifically designed to stimulate antibody populations similar to those that could be prepared from bovine colostrum or egg yolk in commercial quantities. Unlike IgA-rich milk from humans and rodents, bovine colostrum comprises mainly monomeric IgG. Although murine milk also contains relatively high concentrations of IgG, the majority of milk antibodies are IgA. Moreover, IgA levels in murine milk tend to rise during lactation such that the concentration of milk IgA on day 14 of lactation is approximately fourfold higher than that on day 4 (60). Throughout lactation, mice source milk IgA from distant mucosal sites. Immediately postpartum, this is achieved by seconding IgA at mucosal surfaces into the serum, which results in a rapid increase in circulating polymeric IgA that permits loading of IgA into the milk via the blood, such that all IgA in milk from 4-day lactating dams is serum derived (26, 27). By around day 8 of lactation, substantial numbers of mucosally derived IgA plasma cells have migrated to the mammary glands to establish local production, such that the proportion of serum-derived IgA decreases to below 25% of the total milk IgA (57, 60, 61). The density of IgA plasma cells in the mammary glands continues to increase up to day 18 of lactation in concert with increased expression of polymeric immunoglobulin receptor and SC (57, 60). Despite the preponderance of IgA in murine milk, analysis of various serum and milk samples from mice in this study indicated that the most likely protective factor in these animals was H. pylori-specific IgG in milk.
Neonatal mice receive maternal IgG via the following two pathways: across the placenta to a significant level by 15 days of gestation (7, 31) and via ingested breast milk, from which IgG is absorbed into the blood until around 16 days of age (28). As maternal IgG on the luminal and systemic sides of the gastric mucosa is intrinsically linked, it may be difficult to completely separate the contribution of ingested and circulating IgG to immunity. Nevertheless, fostering pups between immune and nonimmune dams allowed us to isolate the effect of transplacentally acquired antibodies and show that protection against infection with H. pylori was entirely attributable to suckling from an immunized dam. This was corroborated by our observation that in weanling mice, milk-derived maternal IgG ceased to contribute to protection once it was taken up into circulation.
The finding that IgA did not contribute to immunity in our model was unexpected. The fact that H. pylori-reactive IgA was readily detectable in IgG-rich serum but was barely detectable in the IgA-rich milk samples indicates that this result was not an artifact of the method we used to measure these antibodies. Our data regarding the protective role of milk IgG are in keeping with those of Abimiku and Dolby (1), who reported that passive protection of suckling mice against intestinal infection with Campylobacter jejuni is mediated by milk IgG. Moreover, these workers found that milk samples collected 4 and 8 days postpartum contained substantial levels of antigen-specific IgG but not IgA (1, 2). Passive protection against intestinal infection with Vibrio cholerae in the absence of specific milk IgA has also been confirmed for milk collected from the stomachs of 6- to 7-day-old pups suckled by dams immunized parenterally with V. cholerae OMV (52, 53). In contrast to these studies, however, we examined mature milk samples collected toward the end of lactation when the production and secretion of total IgA in the mammary gland are maximal. Although we were able to detect H. pylori-specific IgA in these late milk samples, the levels were negligible compared to those of specific IgG. Nevertheless, as our studies were not designed to investigate the capacity of human breast milk IgA from H. pylori-infected or immunized mothers to protect their nursing infants from infection, our data do not rule out the possibility that H. pylori-specific IgA in human milk could convey protection against infection similar to that mediated by IgG in mice.
Quantitative and qualitative analyses of serum- and milk-derived antibodies from dams vaccinated with SS1 whole cells and their pups showed a discrepancy between the proportion of urease-specific IgG circulating in sera of the dams and that of their pups, suggesting that serum IgG is unequally loaded into murine milk. In C57BL/6 mice, IgG in milk is predominantly IgG2b due to its abundance in serum, although IgG1 is also present in disproportionately high concentrations compared to those of serum (30, 42). The differential loading of immunoglobulin subclasses into milk may have been responsible for the discrepancy in urease-specific IgG levels in the sera from dams compared to those of their pups, but we did not confirm this by determining the subclasses of the antibodies we measured.
In this study, we also found that protected pups clustered with particular dams. Although differences in colonization by H. pylori in individual litters of normal suckling mice have been reported previously (43), we saw no evidence of this in 12 litters from nonimmune dams. We did observe, however, that protection may correlate with the concentration of H. pylori-specific antibodies ingested by the pups. Nevertheless, we also found that high levels of SS1-specific antibodies in milk from some dams did not necessarily cause a reduction in the bacterial loads of their litters. With respect to specific antibodies, high levels of urease-specific antibody produced by a nonprotective dam, S3, contrasted with low levels produced by two highly protective dams, S1 and S6, and suggested that antibodies to urease are neither sufficient nor essential for antibody-mediated protection in our model. The ability of urease to enhance vaccine efficacy (16, 45) and dominate the immune response (47) has been shown in studies using multivalent vaccines. There is also evidence that complex antigen preparations lacking urease are not always protective (29). While specific antigens other than urease have been shown to convey protective immunity to H. pylori, including catalase (12), VacA (39), CagA (40), HspA (16), HpaA (45, 55), and NapA (51), urease and whole-cell-derived preparations that are innately inclusive of urease have been the most reliable vaccines tested to date (20, 21, 46). As whole-cell-based vaccines have never been tested independently of urease, we are yet to determine the extent to which the achievable protection is attributable to urease-mediated immunity.
By using a urease deletion mutant of SS1 as a vaccine in the suckling mouse model, we showed that urease was dispensable for antibody-mediated passive immunity. This finding is in accordance with that of Corthesy-Theulaz et al. (13) in a similar model of passive immunity against H. felis using mucosally vaccinated dams. These workers found that while vaccination with H. felis whole-cell antigen conveyed protection against infection with H. felis in suckling pups, no such protection was observed in pups suckled by dams immunized with urease. Also in agreement with Corthesy-Theulaz et al. (13), we demonstrated by immunoblot analysis that protective milk antibodies recognized surface antigens of H. pylori, although by analyzing milk IgG and IgA separately, we found that this reactivity was restricted to IgG.
In contrast to previous studies, including that of Corthesy-Theulaz et al. (13), we showed for the first time that parenteral vaccination of dams can lead to the production of milk antibodies that prevent infection with H. pylori entirely. There are several caveats, however, regarding the use of the suckling mouse model that require consideration when examining our results. First, the pups were fed exclusively on hyperimmune milk before and after exposure to H. pylori, thereby creating a treatment regimen that is not practicable in humans. Second, the depth of the gastric mucous gel layer in infant rodents is less than that of adult rodents (56) and hence could be relatively less protective for H. pylori and more penetrable by antibodies than in adults. Moreover, the gastric pH of suckling mice is higher that that of adult mice (9), thus providing an environment conducive to antibody stability. Nevertheless, these scenarios are reminiscent of the weakened mucous layer (49, 50) and elevated gastric pH of symptomatic human carriers of H. pylori, particularly those receiving proton pump inhibitors (48), who as a result may be relatively more responsive to treatments involving the oral ingestion of antibodies.
The natural progression of infection with H. pylori also differs considerably between adult and infant mice. Only 70% of infant mice infected with SS1 (106.7 CFU) had detectable colonization when sampled at 4 dpi (8 days old), and those infected had a median bacterial load of 103.2 (IQR, 102.8 to 103.4) CFU per stomach (104.6 [IQR, 104.1 to 104.8] CFU/g). When sampled at 6 dpi, all mice had detectable colonization, but the median bacterial load remained low at 103.2 (IQR, 102.6 to 103.7) CFU/g (104.5 [IQR, 103.9 to 105.1] CFU/g). These findings are in keeping with a previous study reporting that mice infected with H. pylori at 8 days of age carry around 103.4 CFU per stomach when sampled after 24 h (25). The bacterial load we typically observed in infant mice at 14 dpi (18 days old) was between 104 and 105 CFU per stomach (106.0 [IQR, 105.5 to 106.3] CFU/g), which is less than that in similarly infected adult mice (typically >106 CFU/g at 3 dpi, increasing to >107 CFU/g by 7 dpi in our experience). While the differences in CFU/stomach between adult and infant mice may be partly due to differences in the sizes of the stomachs of mice of different ages, it is highly unlikely that size alone is responsible for the considerable differences in bacterial counts we observed in the stomachs of infant and adult mice.
Despite these various differences between adult and infant mice, the infant mouse model still affords an opportunity to examine the protective efficacy of antibodies against H. pylori in keeping with its infectious dose for humans, which is estimated to be less than 104 CFU (24). Another advantage of the infant mouse model is that the mechanisms by which IgG is loaded into murine milk are analogous to those by which antibody is loaded into bovine colostrum (35) and egg yolk (32, 36). Consequently, the model is well suited for evaluating and optimizing candidate vaccines for generating passive immunization strategies from such biological systems. For example, antibody samples from this current study have formed the basis of an immunoproteomics project to delineate correlates of protection mediated by passive immunity. Because of the relatively short duration of suckling by mice, however, the mouse model is less useful for determining whether orally ingested antibodies can function therapeutically in established infection with H. pylori, either alone or in conjunction with standard therapies. Humans, or adult animal models of H. pylori infection, are more appropriate for such studies.
Overall, our results indicate that ingestion of milk from hyperimmunized animals can prevent infection with H. pylori. Our results also showed that protection against infection was dictated by both the quantity and quality of antibody populations. The suckling mouse model we developed during the course of this work could be used to guide the development of improved vaccines for antibody production that may increase protective efficacy above that determined in this and other studies of passive immunity, which have used antibody populations generated by vaccination with whole cells of H. pylori. Moreover, the evidence that protection was mediated by the IgG component of murine milk indicates the potential usefulness of monomeric immunoglobulin-rich polyclonal egg yolk and bovine colostrum products in the prevention and possibly the treatment of infections with H. pylori.
This work was supported by Immuron Ltd. (formerly known as Anadis Ltd.) and the Australian National Health and Medical Research Council.
We are grateful to Susie Germano and Danijela Krmek for assistance with specimen recovery.
Editor: S. R. Blanke
Published ahead of print on 8 September 2009.