Vaccine-enhanced competition permits rational bacterial strain replacement in the gut

Abstract

Colonization of the intestinal lumen precedes invasive infection for a wide range of enteropathogenic and opportunistic pathogenic bacteria. Here we show that combining oral vaccination with engineered or selected niche-competitor strains permits pathogen exclusion and strain replacement in the mouse gut lumen. This approach can be applied both prophylactically to prevent invasion of non-typhoidal Salmonella strains, or therapeutically to displace an established Escherichia coli. Both intact adaptive immunity and metabolic niche competition are necessary for efficient vaccine-enhanced competition. Our findings imply that mucosal antibodies have evolved to work in the context of gut microbial ecology, by influencing the outcome of competition. This has broad implications for the elimination of pathogenic and antibiotic-resistant bacterial reservoirs, and for rational microbiota engineering.

Introduction

Drug-resistant infections with Escherichia coli (E. coli) and Salmonella spp. are increasing (1). Both species typically colonize the gut before initiating disease and can be carried asymptomatically in the gut lumen. There is a pressing need for control and prevention strategies that are independent of antibiotics, and which target not only disease but also gut pathogen reservoirs.

Most current vaccines focus on clearing infection from tissues, relying on serum antibodies and cellular immunity. However, protection from colonization of the gut lumen, topologically outside of the body, is fundamentally different. Invasion into this densely populated microbial ecosystem is a crucial step in enteropathogenic bacterial infections(2, 3). Correspondingly, the microbiota plays a role in recovering homeostasis and excluding colonizing opportunistic pathogens (4, 5). For example, fecal microbiota transplantation can be curative in recurrent Clostridioides difficile infections (6) and dietary shifts or antibiotic treatment break colonization resistance, permitting Salmonella infection (3). Consequently, protective intestinal immune mechanisms have evolved to work in the context of gut microbial ecology.

Whole-cell inactivated oral vaccines induce high-affinity T cell-dependent IgA responses against the Salmonella enterica subspecies enterica serovar Typhimurium (S.Tm) and E. coli surface, including the O-antigens of lipopolysaccharide (7, 8). This IgA response aggregates S.Tm in the gut lumen via enchained growth, increasing bacterial clearance due to rapid flushing of aggregates, and exerting a selective pressure (7, 9). Consequently, oral vaccination can alter the outcome of competition between O-antigen variants of Salmonella (10). Generalizing this concept, we expect high-affinity IgA to generate a fitness disadvantage for any targeted strain. Combining oral vaccination with oral supplementation of a live bacterial niche competitor could therefore drive competitive exclusion of the pathogen. An ideal niche competitor would be a non-pathogenic strain with complete metabolic niche overlap, a faster growth-rate, and an absence of surface antigen cross-reactivity to the pathogen of interest. We here define the combination of a whole-cell inactivated oral vaccine with a live niche competitor as “vaccine-enhanced competition”.

To explore the quantitative limits of vaccine-enhanced competition, we generated a model. This is based on exponential growth of wildtype Salmonella, a competitor strain and the microbiota (modelled as a single entity) in the presence of finite shared and private niches, and clearance due to flow and death. Kinetic parameters were estimated from in vivo competition assays, combined with direct quantification of bacterial generation numbers using the unstable plasmid pAM34 (11) (Supplementary Model). Finite niches were defined to be (i) available to the microbiota only, (ii) available to Salmonella and its competitor only, or (iii) competitively used by both Salmonella and the microbiota (see Supplementary Model Text). The rate of elimination of the target is expected to be related to (i) the magnitude of IgA-driven increase in clearance rate of the target and (ii) the extent of metabolic competition between the competitor and target strains, which suppresses the available niche for the pathogen (Supplementary model text). The model predicts clearance of virulent Salmonella within a few days if a typical IgA response is combined with a fast-replicating competitor. Neither IgA alone, nor competitor alone led to complete clearance. The model predicts complete robustness to the timing of competitor introduction, if some competitor bacterium is present at the time of challenge.

Using these timings as a guide, we explored vaccine-enhanced competition in the context of two in vivo models. Firstly, we applied vaccine-enhanced competition to prevent disease and eliminate colonization in the murine model of non-typhoidal Salmonellosis in resistant (Nramp1^+/+) mice (12, 13). Secondly, we demonstrated that vaccine-enhanced competition can be applied therapeutically to eliminate E. coli from the gut lumen of C57BL/6 mice. Vaccine-enhanced competition has broad potential to manipulate enterobacteriaceal colonization and disease.

Salmonella-based niche competitors enhance vaccine protection

As a proof-of-concept, a niche-competitor with complete metabolic niche overlap was constructed by engineering the pathogen itself i.e., S.Tm SL1344. To improve competitive fitness and abolish virulence, we deleted the master regulator of Salmonella Pathogenicity Island 1 (SPI-1) genes, hilD, and the Salmonella Pathogenicity Island 2 (SPI-2) component ssaV (14, 15). Antibody cross-reactivity was reduced by inactivating the abequose O-acetylase oafA (16), converting the wildtype serovar O:4[5],12 to O:4,12 serovar S.Tm^Comp (fig. S1). It should be noted that the O-antigen densely carpets the surface of Salmonella, and O-antigen modification alone is therefore sufficient to dramatically decrease live-cell recognition by antibodies specific for the unmodified O-antigen (10).

We first tested vaccine-enhanced competition based on oral vaccination against the wildtype S.Tm, combined with pre-colonization with S.Tm^Comp at 3 days or 19 days prior to challenge in the streptomycin non-typhoidal Salmonellosis model (12) (Fig. 1A, fig. S2). Antibody titres at endpoint were not altered by the presence of S.Tm^Comp (Fig. 1B). In control mice, S.Tm^WT robustly colonized the gut lumen to day 10 post-infection (Fig 1C+D, fig. S2). Niche competitor (S.Tm^Comp) or oral vaccination alone mildly suppressed colonization (Fig. 1C+D, fig. S2). In mice that were both vaccinated and colonized with S.Tm^Comp, S.Tm^WT initially colonized poorly (Fig. 1D, fig. S2) and was rapidly cleared. S.Tm^WT was undetectable in the cecum content in 7 of 13 mice in this group at day 10 of infection (Fig. 1C+D, fig. S2), and was suppressed more than 100’000-fold in the remaining animals. As S.Tm^Comp persists at low levels for several weeks in our SPF mouse colony (fig. S1), the time of pre-colonization with S.Tm^Comp did not significantly influence clearance efficacy (fig. S2, fig. S3). This is practically important as it indicates that a niche competitor can be fed considerably before an infection occurs. This highlights the need for rigorous safety testing, as these organisms need to colonize the gut long-term or to be delivered frequently.

Figure 1. Vaccine-enhanced competition based on an inactivated whole-cell oral vaccination and S.Tm^Comp can eliminate S.Tm^WT from the gut and prevent S.Tm^WT transmission.

(A) Experimental procedure. PBS (blue) or PA-S.Tm-vaccinated (pink) 129S6/SvEv mice were pretreated with streptomycin and infected with 1·10⁶ S.Tm^WT. Indicated groups were pre-colonized with 5·10³ S.Tm^Comp at day -3 (filled symbols). (B) S.Tm^WT-specific intestinal IgA titres at endpoint. S.Tm CFUs in cecum content (C), feces (D) MLN (E) and spleen (F). S.Tm^WT-free mice in (D) refers to Vacc.+S.Tm^Comp group. (G-I) Intestinal inflammation determined by fecal lipocalin-2 (G) and cecum histopathological scoring (H). (I) Representative images of H&E-stained cecum. Arrowheads indicate goblet cells. Scale: 100 µm. (J-K) Streptomycin-pretreated naïve 129S6/SvEv mice received a FMT with feces collected on day 9 from A-K. S.Tm^WT and S.Tm^Comp CFUs in feces (J) and cecum content (K). The pink open triangle depicts a donor from the Vacc.+S.Tm^Comp group in which S.Tm^WT transmission occurred.

Pooled data from two independent experiments with switched antibiotic resistances (n=5-8 mice/group (A-I) or n=8 mice/group (J-K)).

Solid lines: median. error bars: interquartile range. Dotted lines: detection limit. Shaded area: detection limit range. Open triangles: (n=1) euthanized prematurely due severe disease. One-way ANOVA (H) performed on log-normalized data (B-C, E-F) or area under the curve (AUC) (D, G). Unpaired two-tailed t-test performed on log-normalized data (B-C, E-F). DF, dilution factor; Lu., lumen; MFI, median fluorescence intensity; S.E., submucosal edema.

Wherever S.Tm^WT was rapidly eliminated from the gut lumen, systemic sites were also sterile (Fig. 1E+F, fig. S2). Pathology, as measured longitudinally by fecal Lipocalin-2 (LCN2) and at endpoint by histopathology (Fig. 1G-I) mirrored intestinal colonization. Pathology was prevented by the vaccine-enhanced competition regimen, but only partially prevented by each intervention alone.

To further corroborate our findings of sterilizing immunity, we performed fecal microbiota transplants (FMT) from infected vaccine-enhanced competition-treated mice at day 9 into naïve streptomycin pretreated mice (Fig. 1A). Transfer of 1 to 10 CFUs is sufficient to cause full-blown disease in this model (15, 17). Despite the high-dose fecal transfer, only one animal of eight showed transmission of S.Tm^WT (Fig. 1J+K). Intriguingly, this donor did not have the highest fecal S.Tm^WT counts at the time of transfer, but the lowest competitor:wildtype ratio of all donors (40-fold excess, compared to over 1000-fold excess). This suggests an additional benefit of the approach: a competitor can prevent transmission even in cases where sterilizing immunity in the gut lumen is incomplete. In contrast, all mice receiving feces from the untreated control group of S.Tm-infected mice became infected (Fig. 1J+K). Therefore vaccine-enhanced competition permitted clearance of high-dose S.Tm^WT challenge from the gut and prevention of invasion into all examined sites. This also largely prevented transmission to naïve hosts, providing herd immunity.

Metabolic niche overlap is necessary for vaccine-enhanced competition

To investigate whether vaccine-mediated clearance from the gut lumen required complete metabolic niche overlap, we compared the functionality of S.Tm^Comp to a galactitol-utilization mutant S.Tm^{Comp ΔgatABC} (Fig. 2A) (18, 19). Lack of overlap for a single sugar (galactitol) in the competitor resulted in significantly slower competitive exclusion of S.Tm^WT in vaccinated mice, as compared to a full niche overlap (Fig. 2B). The mechanism therefore requires metabolic niche overlap.

Figure 2. Metabolic niche overlap of S.Tm^WT and S.Tm^Comp favours vaccine-enhanced competition.

(A) Experimental procedure. PBS or PA-S.Tm-vaccinated 129S6/SvEv mice were pre-colonized with 5·10³ S.Tm^Comp or S.Tm^CompΔgatABC 3 days before infection and infected with a 1:1 ratio of 1·10⁶ S.Tm^WT and S.Tm^ΔgatABC. (B) S.Tm CFUs in feces (n=5 mice/group). Solid lines: median, error bars: interquartile range. Dotted lines/shading: detection limit and range. Two-way ANOVA on log-normalized data between S.Tm^WT and S.Tm^ΔgatABC and one-way ANOVA on AUC comparing S.Tm^WT between treatment groups was performed.

We next investigated the ability of a more distantly related mouse commensal B2 E. coli 8178 (Ec⁸¹⁷⁸) to act as a niche competitor against S.Tm. This strain has a partial niche overlap with S.Tm in vivo (20, 21). A benefit of using a more distantly related probiotic is that this E. coli produces a completely unrelated O-antigen structure, allowing us to use the “evolutionary trap” version of our S.Tm vaccine (10) (Fig. 3A) i.e., a version of the vaccine covering all common S.Tm O-antigen variations. Colonization with Ec⁸¹⁷⁸ did not induce detectable Ec⁸¹⁷⁸-binding intestinal IgA nor serum IgG (fig. S4). Vaccine-enhanced competition based on combining the S.Tm^WT-vaccination with Ec⁸¹⁷⁸ could decrease initial S.Tm^WT expansion about 1000-fold and completely prevented gut inflammation (Fig. 3B-H) (7, 22).

Figure 3. Vaccine-enhanced competition with an imperfect niche competitor limits S.Tm^WT colonization and prevents inflammation.

(A) Experimental procedure. PBS (blue) or EvoVax vaccinated (pink) 129S6/SvEv mice were pretreated with ampicillin and infected with 1·10⁶ S.Tm^WT. Two groups were pre-colonized with 5·10³ E. coli 8178 3 days before infection. (B-E) CFUs in feces (B), cecum content (C), liver (D) and spleen (E). Intestinal inflammation determined by fecal lipocalin-2 (F) and histopathological scoring of cecum (G). (H) Representative images of H&E-stained cecum. Arrowheads show goblet cells. Scale bars: 100 µm.

Pooled data from two independent experiments (n=6-8 mice/group). Solid lines: median, error bars: interquartile range. Dotted lines and shading: detection limit and range (when detection limit depends on sample weight). Open triangles: mice (n=3) euthanized due to excessive pathology. One-way ANOVA (E) on log-normalized data (C) or area under the curve (AUC) (B-D). Unpaired two-tailed t-test on log-normalized data for comparing 2 groups (C).

These experiments were then repeated using a completely unrelated probiotic strain (Lactobacillus casei), that has minimal niche overlap with Salmonella (23). Despite robust colonization, L. casei had no significant effect on vaccine-mediated protection (fig. S5).

A final aspect of niche competition tested is whether a more intact microbiome may contribute to niche competition. High-fat diet feeding induces mild and transient microbiota disruption (21). In this model, oral vaccination alone and niche-competitor alone were effective in preventing Salmonella colonization. However, vaccine-enhanced competition performed slightly better in preventing disease, suggesting that combining oral vaccination and niche competition has benefits also in the presence of natural competitors (fig. S6). In the murine typhoid model, mice are orally infected with a high dose of Salmonella without pretreatment or pre-existing damage to the microbiota (13). Vaccine-enhanced competition also provides robust protection from colonization and gut inflammation in this model. Of note, as the high infection dose permits immediate invasion of Salmonella into Peyer’s patches, protection of systemic sites was weak in this model (fig. S7) (7, 24–26).

Vaccine-enhanced competition can displace an E. coli strain from the gut microbiota

To discover whether vaccine-enhanced competition can eliminate a strain already present in the microbiota we used the non-encapsulated commensal E. coli strain HS as a target (27). These experiments were performed in C57BL/6 mice carrying a low-complexity microbiota that permits continuous E. coli colonization at up to 10⁸ CFU/g feces without antibiotic pretreatment (28). After pre-colonization with E. coli HS, mice were fed oral vaccine on days 3, 13 and 20 and the niche-competitor cocktail on days 10 and 17. E. coli HS and niche competitor levels were monitored in feces until day 23 (Fig. 4A). Whole-cell inactivated oral vaccines for E. coli HS induced an IgA response against the surface of live E. coli HS, but not against the three E. coli strains making up the niche-competitor probiotic (Fig. 4B). Vaccine-enhanced competition resulted in complete displacement of the targeted E. coli strain in 50% of the treated animals, with strongly suppressed colonization seen in the remaining animals (Fig. 4C+D). Each treatment alone had only a very mild effect on E. coli HS colonization levels (Fig. 4C+D).

Figure 4. Vaccine-enhanced competition can be used to therapeutically replace a gut E. coli strain.

(A) Experimental procedure. C57BL/6 low-complexity microbiota mice were colonized with the commensal E. coli HS strain. 3 days later, vaccination with either vehicle alone or PA-E. coli HS was started and a cocktail of competitor E. coli was introduced orally on day 10 and 17. (B) Intestinal IgA titres specific for E. coli HS and the competitor strains at endpoint. (C) Fecal and (D) cecal CFUs. Pooled data from two independent experiments (n=6-10 mice/group). Solid lines: median, error bars: interquartile range. Dotted lines and shading: detection limit and range. One-way ANOVA on log-normalized data (B, D) or area under the curve (AUC) (C).

Therefore, the concept of oral vaccine-driven strain replacement is generalizable to non-encapsulated E. coli and can eradicate a bacterium already present in the gut microbiota.

Vaccine-enhanced competition requires intact adaptive immune system

Finally, as immune stimulation could potentially induce antibody-independent effects contributing to pathogen clearance (29), we tested the role of adaptive immunity in vaccine-enhanced competition. As genetically immunodeficient lines were not easily available, we used antibodies to depleted more than 90% of all CD4⁺ T cells and B cells from the spleen, mesenteric lymph nodes and blood of treated mice (Fig. 5A+B, fig. S8). CD4⁺ T cell and B cells depletion during the vaccination period strongly suppressed induction of vaccine-specific IgA in both SPF 129SJL mice and gnotobiotic C57BL/6 mice (Fig. 5C, fig. S8). In the therapeutic clearance of E. coli, suppression of T-dependent antibody responses completely prevented E. coli HS elimination, while isotype-treated mice robustly cleared E. coli HS. This is consistent with a major contribution of vaccine-induced adaptive immunity (Fig. 5D+E). In prevention of Salmonella using Ec⁸¹⁷⁸ as a competitor, only the isotype-control-treated mice receiving the vaccination-enhanced protection treatment remained healthy throughout the experiment, with no bloom of Salmonella and effective prevention of systemic spread (fig. S8), again confirming the role of vaccine-induced adaptive immunity. Conversely, depleting CD4⁺ and CD8⁺ T cells only after vaccination, i.e. after induction of a T cell-dependent antibody responses, had no impact on vaccine-enhanced competition (fig. S9). As the only adaptive immune component present in non-inflamed gut are secretory antibodies, these findings are consistent with a role of T-dependent IgA in vaccine-enhanced competition.

Figure 5. Therapeutic clearance of E. coli is antibody dependent.

(A) Experimental procedure. C57BL/6 “low-complexity microbiota” mice were colonized with E. coli HS. 3 days later, vaccination with either vehicle alone or PA-E. coli HS was started, and a cocktail of competitor E. coli was introduced orally on day 10 and 17. Mice were treated with anti-CD4 and anti-CD20 or isotype controls one day before vaccination. (B) CD4⁺ T cells and B cells in blood at endpoint. (C) E. coli HS-specific intestinal IgA titres at endpoint. (D) Fecal and (E) cecal CFUs. Pooled data from two independent experiments (n=4-5 mice/group). Solid lines: median. error bars: interquartile range. Dotted lines and shading: detection limit and range. One-way ANOVA (B) on log-normalized data (C, E) or area under the curve (AUC) (D).

Our high-dose inactivated oral vaccines are not potent inducers of effector T cell responses, in contrast to live-attenuated vaccines (24). Correspondingly, live-attenuated vaccines are poor at preventing gut colonization but do generate T cell-dependent protection of deep tissues (fig. S9). This highlights a dichotomy in protective mechanisms between the gut lumen and systemic sites and between vaccine-enhanced competition and live-attenuated Salmonella vaccines.

Discussion

The concept that gut microbes contribute to prevention of pathogen colonization is well accepted (2, 3, 30, 31). We know secretory IgA increases the clearance rate of intestinal Salmonella and E. coli via enchained growth (7), and that this exerts selective pressure on IgA-targeted bacteria (10). Here, we show that these processes work best together. This represents a fundamental shift in approach to designing vaccines for enteric bacteria. Published nutrient blocking and rationally designed microbiome engineering approaches can supress Salmonella loads to around 10⁵ CFU/g feces (32, 33), while our combined approach can generate more than a 10⁹-fold reduction in colonization, down to undetectable levels.

Vaccine-enhanced competition requires both rational oral vaccine design and optimal niche competitor selection. Earlier work on evolutionary trap vaccines provided a starting point for oral vaccine design targeting the O-antigen of non-encapsulated Enterobacteriaceae (10). Here, we explored two ends of the spectrum for niche competitor selection/design, first by rationally modifying the target strain itself to generate a “perfect” competitor and second by testing commensal E. coli strains and a distantly related Lactobacillus strain (21, 27). There has been considerable progress in the genomic prediction of community function and metabolic niche overlap(34), recently defined as nutrient blocking (32), indicating improved approaches are likely to involve designed niche competitor consortia.

We can imagine several scenarios where this type of prophylaxis could be relevant. Prevention of enteropathogenic bacterial infections during travel could be achieved by vaccine-enhanced competition, given shortly before and potentially during travel. Orally administered vaccines and probiotics are ideal for self-administration. Additionally, we have shown that we can replace a strain already present in the gut, which allows elimination of opportunistic pathogen reservoirs, for example in patients scheduled for high-risk interventions (35, 36). This could be game-changing in preventing invasive, and increasingly multi-drug-resistant disease in vulnerable human patients. Another interesting hypothesis is that IgA-mediated strain replacement may occur frequently without our intervention, i.e. is an evolved function of secretory IgA. This is consistent with observations of instability of E. coli at the strain level, but not the species level, over time in healthy volunteers (37).

Our observations may also explain some controversies in the existing Salmonella literature. For example, the extent of protection from non-typhoidal Salmonellosis obtained with different types of oral vaccines in different laboratories varies extensively (38). Our data indicate that the variable efficacy of Salmonella niche competitors in the microbiota of mice is a critical determinant of protective efficacy. Microbiota composition also varies extensively between humans and over time (39). The vaccine-enhanced competition approach should remove the lottery of natural niche-competitor abundance and allow robust protection or therapy in most treated individuals. A limitation we would like to highlight here is that, due to practical and ethical restrictions, we have worked over relatively short timescales. We have not addressed the longevity of vaccine-induced responses, nor potential within-host evolution of competitor strains that might be observed over months or years. In this context, we cannot exclude that very small, below-detection limit reservoirs of infection may re-emerge at very late time-points post-treatment, even in this very robust prophylaxis system.

Currently, we have focused on one strain of pathogenic Salmonella and one commensal E. coli strain. Given the extensive strain-level variation in antigenicity, pathogenicity and metabolism between gut bacterial pathogens (40), the identification of the most relevant vaccine compositions and niche competitors may not be a case of simple extrapolation. Beyond non-encapsulated Enterobacteriaceae, there are open challenges in determining what the most relevant bacterial surface antigens are for IgA targeting and/or for designing oral vaccine strategies that optimally induce such responses. For example, C. difficile produces an S-layer with highly polymorphic exposed epitopes which is challenging for vaccine design (41). Klebsiella species that have been implicated in exacerbating inflammatory bowel disease symptoms (42) are heavily encapsulated, likely requiring glycoconjugate vaccines for induction of relevant antibody responses (43). Further important future directions therefore include (i) improving the breadth and affinity/avidity of IgA responses induced and (ii) optimizing the design of niche competitors/competitor consortia.

In conclusion, we have identified a mechanism to prevent Salmonella colonization and achieve E. coli strain replacement in the gut. These results lay the foundation for targeted treatments for bacterial infections in mammals, promising significant advancements in medical science.

One sentence summary.

Combining oral vaccination with a bacterial niche competitor allows complete replacement of wildtype Salmonella or Escherichia coli with non-pathogenic strains in the gut lumen.

Data and Materials availability

Original code and raw data have been deposited at ETH Zurich Research Collection https://doi.org/10.3929/ethz-b-000717590(44). Newly generated bacterial strains and all other materials are available from Emma Slack (emma.slack@hest.ethz.ch) and Médéric Diard (mederic.diard@biozentrum.unibas.ch).