Microbiome as a tool and a target in the effort to address antimicrobial resistance

Abstract

Reciprocal, intimate relationships between the human microbiome and the host immune system are shaped by past microbial encounters and prepare the host for future ones. Antibiotics and other antimicrobials leave their mark on both the microbiome and host immunity. Antimicrobials alter the structure of the microbiota, expand the host-specific pool of antimicrobial-resistance genes and organisms, degrade the protective effects of the microbiota against invasion by pathogens, and may impair vaccine efficacy. Through these effects on the microbiome they may affect immune responses. Vaccines that exert protective or therapeutic effects against pathogens may reduce the use of antimicrobials, the development and spread of antimicrobial resistance, and the harmful impacts of these drugs on the microbiome. Other strategies involving manipulation of the microbiome to deplete antibiotic-resistant organisms or to enhance immune responses to vaccines may prove valuable in addressing antimicrobial resistance as well. This article describes the intersections of immunity, microbiome and antimicrobial exposure, and the use of vaccines and other alternative strategies for the control and management of antimicrobial resistance.

The structure and function of the human microbiome help shape immune system development, nutritional status, and other processes in early postnatal life that prepare the host for future environmental exposures and microbial encounters. In turn, microbiome structure and function are shaped by and reflect a history of host immune system activity and prior microbial encounters as well as a history of other exposures, especially exposures to antimicrobials. These two-way relationships have important consequences for antimicrobial resistance (AMR) and for the possibility of vaccine use to control this growing problem. Antimicrobials select for drug-resistant strains, and their repeated use creates a host-specific reservoir of antimicrobial-resistance genes and organisms as well as risks for subsequent microbiome invasion by pathogens and subsequent disease. Vaccines that protect against pathogens may preempt the need for antimicrobials, prevent the development and spread of AMR, and avoid the harmful impacts of these drugs on the microbiome. The relationships between microbiome and host can work in the opposite direction as well, as developing evidence suggests that the state of the microbiome may affect the efficacy of a vaccine.

This Perspective article highlights some of the key aspects of the symbiosis between humans and their microbiome as they pertain to AMR and to vaccines as part of a strategy to address this global health crisis. It describes the intersections of immunity, microbiome, and antimicrobial (antibiotic) exposure and takes an ecological view. The concepts and findings discussed here have important implications in that they suggest complementary strategies for the control and management of AMR. Besides the use of vaccines per se, these strategies include other approaches for manipulation of the microbiome, such as deliberate delivery and installation of strains and synthetic communities to inhibit colonization by antibiotic-resistant organisms or the direct targeting of resistant organisms and genes.

Microbiome and Immune System Development

Long-standing evolutionary and mutualistic relationships between humans and their indigenous microbiota are managed in part by the host immune system. Others have reviewed this admittedly broad but compelling topic (1). Our purpose is to highlight the specificity and early-life maturation of these relationships and the implications for strategies to address AMR.

Host–Microbiome Specificity May Complicate Microbiota-Targeted Therapies and Offer Opportunities.

Germ-free animals are born with an immune system that displays features of incomplete development and differentiation; immune system development and differentiation can be completed by colonization of the animal with a normal microbiota from its own species. Work by Chung et al. (2) showed that human and rat microbiotas fail to correct the impaired immune maturation of the germ-free mouse to the same degree as achieved with mouse microbiota. The abundance of CD4⁺ and CD8⁺ T cells and dendritic cells in the small intestine and local secondary lymphoid organs and antimicrobial peptide expression were significantly lower in germ-free mice after administration of human or rat microbiotas than in those given a mouse microbiota, although roughly similar overall densities of gut colonization were achieved in each case. Lower levels of cell proliferative activity in secondary lymphoid organs were thought to explain the lower T cell numbers in the small intestine. These differential features, along with differences in the newly established microbiotas, were accompanied by functional deficiencies, as the mice given a human gut microbiota were more susceptible to Salmonella infection after oral inoculation. Thus, proper matching of a host species with its coevolved host species-specific microbiota appears to be important for postnatal immune function.

There is also specificity in the effects of certain members and components of the microbiota on immune system development and function. For example, specific clades of microbes, such as Clostridium species (3, 4) and segmented filamentous bacterium (5), have been proposed to play roles in T cell differentiation, proliferation, and function. At the same time, the net effect of the microbiota on immune system function is distinct from and more complex than the sum of the effects of the individual microbiota members (6). That is, the actions of individual microbes depend on their ecological and environmental setting. While the nature of these ecological and environmental interactions and the relative importance of individuality in the microbiome for immune system function are not entirely clear (6), these issues have important implications for the design of microbiome restoration strategies.

Microbiome and Host Immune Maturation Are Intertwined and May Affect Vaccine Efficacy.

Perinatal microbial exposures, proper timing of these exposures, subsequent assembly of the microbiota, and early-life milestones in mucosal and systemic immune system maturation are all intertwined (7, 8). These interdependencies have relevance for early childhood responses to vaccines. Microbial priming of the immune system probably begins before birth. Experiments in mice suggest that maternal gut colonization patterns during pregnancy affect intestinal innate lymphoid cells in the newborn, which in turn prevent intestinal inflammation in the pups (9). Transient colonization of the dams during gestation led to improved intestinal barrier integrity and systemic immune responsiveness. It was proposed that these effects were mediated by maternal microbiota products transferred to the offspring by antibodies.

Passive transfer of breast milk-associated IgA to the newborn provides early protection against commensals and pathogens until the newborn can develop its own protective responses, especially low-affinity, broadly reactive IgA and high-affinity, specific IgA (10). High-affinity IgAs are typically produced in Peyer’s patches with help from CD4⁺ T cells after antigen presentation and are directed against pathogens. Low-affinity IgAs are produced within the gut mucosa without a requirement for T cell help, are secreted into the gut lumen, and help restrict bacterial commensals from translocating across the gut epithelium as well as limit local and systemic immune responses to them (11, 12). They are directed at microbial antigens, induce changes in targeted microbes, and affect the composition of the microbiota in a host age-dependent manner (6, 10, 13). In turn, the maturation of the microbiota in early life regulates the development of IgA responses even during the period of milk-exclusive feeding (13). Although the mechanisms of IgA developmental regulation have not been thoroughly elucidated, the expression of IgA is influenced to some degree, and probably throughout life, by the short-chain fatty acids (SCFAs) produced by bacterial fermentation of complex carbohydrates in the colon. Acetate is a positive regulator of IgA production; its effects are mediated by the G protein-coupled receptor GPR43, aldehyde dehydrogenase, and retinoic acid (14). On the other hand, butyrate, which acts through GPR109a, may be a negative regulator (15).

Koch et al. (16) have reported that mice generate circulating IgG2b and IgG3 antibodies directed against their gut commensals that are also T cell independent. The specificity of these antibodies differs from that of IgAs, and their production depends upon Toll-like receptors 2 and 4 (TLR2, TLR4) on B cells. These anticommensal IgG2b and IgG3 antibodies are acquired from the mother in utero and later via breast milk. Maternally derived IgG2b and IgG3 antibodies in breast milk appear to diminish mucosal T helper cell activation and expansion in young pups and act together with maternal IgAs to decrease T helper cell responses to commensals. Thus, maternally acquired IgAs and IgGs may cooperate in the newborn to recognize commensals via different mechanisms and at different sites, control translocation, and limit potentially harmful immune responses.

Microbial cell-wall components and other TLR ligands (17, 18) including flagellin (19), surface polysaccharides such as Bacteroides fragilis polysaccharide A (20, 21), metabolites including those derived from tryptophan (22, 23), and other small molecules shape and regulate immune function as well as intestinal epithelial homeostasis. SCFAs in particular have a wide variety of effects on host immunity in addition to their regulatory effects on IgA production (24–26) and enhance differentiation of colonic T cells into T regulatory cells (27, 28). Although these microbial components and products are often studied in isolation, they occur together in the native ecosystem and exert combinatorial effects, sometimes additive or synergistic, sometimes synthetic, and sometimes mutually inhibitory or counterbalancing. For example, butyrate counteracts the proinflammatory effects of lipopolysaccharide on macrophages in the lamina propria (29).

Because the first year of life is a critical period for the establishment of a healthy and well-functioning microbiome (8) as well as for postnatal development and biasing of immune responses, strategies that can prevent infections and thus antibiotic treatment early in life may be disproportionately beneficial. For example, concern has been raised about the role of early antibiotic exposure in altered immune development and an increased risk for allergic and other immune-mediated disease in later childhood (30, 31). While a detailed understanding of the timing and agent-specific features of antibiotic risk in early childhood is lacking, it seems clear that, if they are required, very brief and infrequent courses of these drugs, and especially those with less durable effects on microbiota, are especially important during this period of life. This line of reasoning would also place higher value on vaccines, which can prevent early-life infections. On the other hand, while early-life immune development has consequences that play out well into adulthood (32), antibiotic use in adulthood also may jeopardize protective responses to infectious agents (33). Benoun et al. (33) have suggested that the therapeutic use of antibiotics may impair the formation of tissue-resident CD4 memory T cells by shortening the period of antigen presentation. Also, as we discuss later, antibiotics may have detrimental effects on vaccine responses (34). At present, it seems too early to say whether the value of vaccines in preventing antimicrobial use early in life should be considered greater than the value of preventing similar amounts of use later in life, but further work on the role of disruption at different ages may clarify this issue.

As we look forward, new strategies for addressing AMR should exploit these connections between microbiome and immune system. One might consider targeted manipulations of the microbiome that induce or reinforce protective host responses against clades of resistant organisms and enhance protective responses to vaccination, thereby minimizing the need for antibiotics. For example, the deliberate selection, design, and testing of probiotic sets of strains for use in early childhood to stimulate protective and nonallergic immune responses deserves a great deal more attention.

Microbiome as a Reservoir of Antibiotic-Resistance Genes, Resistant Organisms, and the Accumulated Effects of Past Antibiotic Exposures

Antibiotics are ancient molecules that have played roles in microbial sensing and signaling. Not surprisingly, AMR is also ancient (35) and has been a part of the human microbiome from its earliest evolutionary origins. The genetic basis of AMR has been detected in some of the oldest available specimens of the human microbiome (36). Current-day traditional human communities in relatively isolated regions of the world with little contact with modern communities also carry a multitude of antibiotic-resistance genes (ARGs), including in the oral microbiome (37), although gene and gene family diversity are lower in isolated communities than in modern ones (38). The function of ARGs in the healthy human microbiome has been confirmed by exploiting selection schemes with expression libraries (39). Here we provide examples of findings that underscore the diversity of ARGs within the microbiome and the impact of antibiotics on the microbiome and discuss the potential importance of such findings in the design and selection of strategies to mitigate AMR.

Among modern-day societies, patterns of ARG abundance and diversity in the gut microbiome vary with antibiotic use among populations. In a study of 252 fecal samples from 207 individuals living in the United States, Denmark, or Spain, resistance genes for antibiotics used in animal husbandry were more abundant than those for antibiotics not in animal use, and resistance genes were more abundant for antibiotics introduced into the commercial market earlier than for those commercialized in more recent years (40). ARG abundance also correlated with country-specific patterns of antibiotic consumption. At the level of the individual, resistomes (collections of ARGs within an individual) persisted for at least a year. Comparisons across culturally and genetically distinct populations and societies suggest that gut resistomes vary not only in the richness and evenness of ARGs but also in the genetic diversity within ARG families. Chinese resistomes have more ARG types and greater abundances of specific types than the resistomes in Danes and Spaniards and also have distinct patterns of ARG SNPs (41, 42). These population-specific resistome patterns raise the possibility of inheritance across generations; in a study of 147 Norwegian mother–infant pairs, the abundances of class 1 integrons in fecal samples of the infants correlated with abundances in their mothers but not with antibiotic use and persisted over the first two years of life (43).

Antibiotics also have clear and profound effects on the microbiomes of individuals. These effects are age dependent and are presumed to accumulate over time as the number of exposures to antibiotics increases, although the confounding effects of illness and other aspects of aging can be difficult to disentangle. In a study of three healthy adults with no antibiotic exposure in the preceding year, a five-day course of ciprofloxacin produced immediate changes in the structure of the fecal microbiota, including decreases in the abundance of roughly 25–50% of the bacterial taxa to undetected levels and dramatic shifts in the numbers of Faecalibacterium and other Ruminococcaceae, Lachnospiraceae, and Bacteroides (44, 45). There was also interindividual variation in some of these changes and in the timing of recovery, with complete recovery after several weeks in one subject and slow directional change over months toward an interim state in another. However, a second five-day course of ciprofloxacin given six months after the first led to a new shift in the microbiota of all subjects that persisted for at least two months after this second exposure. These results probably reflected both direct effects of the antibiotic and of ARGs and indirect effects due to ecological interactions among members of the microbiota community. These observations were reminiscent of a classic ecological phenomenon: Compounded perturbations of a type, magnitude, and/or frequency distinct from and greater than those to which a community has had an opportunity to adapt can lead to permanent and unexpected changes (46). Other studies of the experimental use of antibiotics in healthy volunteers have revealed similar basic findings: evidence for both shared responses (e.g., decreased overall taxonomic diversity) and individualized responses, with very different kinds and durations of effects, varying as a function of body site and antibiotic classes (47). Some, but not all, responses can be predicted based on the microbiome composition before exposure (48). A predictive understanding of microbiome stability and of the impact of antibiotic disturbance on microbial ecosystem function is an important but unmet need.

The roles of past exposures to antibiotics and of age at the time of these exposures in the development of antibiotic resistance are not sufficiently understood. It has been shown that intrapartum antibiotics for preventing group B streptococcal disease in the newborn have consequences for the microbiome of the baby (49). Not only taxonomic composition but also SCFA production and ARG content are altered during the first three months of life in babies born to women who receive intrapartum prophylactic antibiotics (49). In general, the durability of antibiotic-induced changes depends upon the antibiotic class. In an analysis of fecal bacterial taxonomic and functional composition in two- to seven-year-old children, the effects of macrolides lasted at least two years, unlike those of β-lactams, which lasted less than one year. Macrolide-associated changes were similar regardless of the age of the child at time of exposure and regardless of prior antibiotic exposure (50). In contrast to the persisting effects of macrolides on microbiota taxonomic composition, the early increase in levels of macrolide resistance reversed soon after cessation of antibiotic use and resolved by the end of one year postexposure. Children with more macrolide exposure, especially before the age of two years, had higher rates of asthma by seven years of age.

Prospective study designs with denser time sampling offer a more robust basis for causal inferences about the effects of antibiotics on microbiome development, function, and resistance. Gibson et al. (51) took this approach for a study of premature infants and found antibiotic exposure associated with reduced species richness and diversity in fecal microbiota and with enrichment of ARGs for noncognate antibiotics, probably because of selection for multidrug-resistant strains (usually members of Enterobacteriaceae). Yassour et al. (52) arrived at similar conclusions with a population of young children born at term. In particular, they found that reduced diversity was most apparent within species at the strain level and that ARG expansion was longer lasting for episomal than for chromosomal genes, highlighting the importance of horizontal gene transfer (HGT). Others have provided evidence for age effects in ARG abundance, number of types, and diversity across longer periods of the human lifespan (53).

Work elucidating the contributions of transmissible elements and HGT to the emergence of antibiotic resistance originated in the 1960s (54–56) and implicated commensals as donors, recipients, and reservoirs of ARGs with experiments involving isolates in a laboratory setting (57). Microbial genomics and metagenomics have afforded a more expansive view of the contributions of HGT to microbial functional capacity and have facilitated inferences about conditions and habitats in which HGT appears to be more common and potentially more important. As an example, a large survey of microbial genomes suggested that HGT is more common among human-associated bacterial isolates than among other isolates and especially between organisms that occupy the same body site and ecological niche, such as in the microaerobic gut (58). On the other hand, ARGs were more likely than other genes to show evidence of transfer across ecological niches and habitats. Interestingly, isolates from farm animals and human food were especially likely to share ARGs with human isolates (58). The connectedness of human, animal, and environmental microbiomes is a key issue that underlies the design of efficient and effective interventions against the selection for and spread of antibiotic resistance. Comparative metagenomic methods have revealed extensive ARG sharing, especially ARGs associated with putative mobile genetic elements, among microbial communities in humans, animals, and the environment in a rural village in El Salvador and a periurban shantytown in Lima, Peru (59). These data suggest structuring of resistome and microbiome phylogenetic composition along an ecological gradient organized by the degree of input from human feces.

By understanding better the conditions that lead to mobilization and horizontal transfer of ARGs in the gut microbiome, new approaches might be designed for reducing AMR. For example, specific drugs or inflammatory states that are unusually potent in stimulating HGT and species or strains that serve as the most efficient ARG donors and recipients might be used for risk profiling of patients and, respectively, could be avoided, treated, or targeted. In the meantime, efforts to replace, suppress, or exclude highly resistant strains with fecal microbiota transplants and other general approaches for microbiome restoration are under study (see below).

Role of the Microbiome in Mediating Vaccine Effects

Microbiome-Dependent Immunogenicity.

A limited body of evidence suggests that the host microbiome at the time of immunization can affect the immunogenicity of a vaccine and thereby change the degree of protection it offers. This finding may provide a basis for understanding some of the host-to-host variability in vaccine responses. The evidence focuses on TLR ligands, which have strong modulatory effects on host responses to infectious disease agents such as poliovirus (60) as well as on responses to vaccines. An important example concerns the contributions of flagellin to antibody responses to influenza vaccine: Elevated levels of TLR5 transcripts in peripheral blood early after influenza vaccination were found to correlate with subsequent levels of antiinfluenza antibodies (34). In a murine experimental model, early B cell responses to this vaccine were dependent on the presence of commensal gut bacteria and on flagellin. Flagellin was shown to be sufficient to stimulate plasma cell differentiation. These and other results have led to the view that the gut microbiota serves to some degree as a natural adjuvant for vaccination. The question is whether these findings can be exploited to enhance protective responses to vaccines that are unadjuvanted or in subjects with insufficient native microbiota-associated adjuvanticity. Efforts to discover differences in human microbiotas that are correlated with vaccine responders and nonresponders have had mixed success (61–63), probably because the key determinants of a vaccine-specific microbiota-based adjuvant effect are related to cell-surface features that are not easily inferred from taxonomic or even metagenomic data. In addition, there are many confounding factors (64), e.g., nutritional status, history of enteric infectious disease, and host genetics, whose effects are not exclusively mediated by the microbiota. The use of defined probiotic strains to enhance immune responses is a potentially promising approach (65–68) but is not yet a precise science with predictable, optimized effects based on a mechanistic understanding.

Insofar as the efficacy of vaccines in preventing infection and disease is modulated by the condition of the microbiota at the time of immunization, the specific effects of vaccines in preventing antibiotic use and preventing antibiotic-resistant infections will be modulated in the same way.

The Microbiome as an Innocent Bystander Protected by Vaccines.

Any vaccine that reduces the incidence of an infectious disease has the potential to reduce the incidence of antimicrobial use (69). To the extent that antimicrobials are used to treat the infectious disease, the entire microbiome may be exposed and, if so, will experience selection for resistance, along with the pathogen, as well as mobilization of ARGs. Hence, the burden of ARGs and antibiotic-resistance organisms in that host will increase, at least transiently, as described above. Every course of treatment with an antimicrobial agent also risks off-target killing of commensals in the treated host. Therefore, every course of antimicrobial treatment averted by a vaccine has potential benefits in reducing this “bystander selection.” Consider, for example, influenza vaccine (a nearly identical argument can be made for other respiratory virus vaccines). Influenza frequently leads to antimicrobial treatment (70, 71). Sometimes this treatment is inappropriate, e.g., when a purely viral acute respiratory infection that should not be treated with antimicrobials is nonetheless treated. Sometimes this treatment is appropriate, e.g., when influenza infection leads to secondary bacterial infection and pneumonia, which then receive antimicrobial therapy. In either case, the entire microbiome, insofar as it is exposed to concentrations of antibiotics capable of selecting for resistance, is subject to such selection. Antimicrobial treatment, whatever the indication, may lead to increases in the prevalence of resistance within potentially pathogenic microbiome members such as oral streptococci (72), Streptococcus pneumoniae (73, 74), Haemophilus influenzae (75), Enterobacteriaceae (76), Staphylococcus spp. (77, 78), and even concurrently infecting strains of Neisseria gonorrhoeae (79). As discussed above, increases in the frequency of resistance genes and other, nonpathogenic microbiome components within these populations may increase the risk of dissemination of these genes by horizontal transfer to potentially pathogenic species within the individual (80, 39). Efforts are just beginning to quantify the impact of vaccines in preventing bystander selection on the microbiome. Meanwhile, an important step toward incorporating this consideration into estimates of vaccine impact is to include antimicrobial use as a secondary outcome in randomized trials and postlicensure observational impact studies of vaccines.

Microbiome as a Site of Vaccine Action.

Several key pathogens targeted by vaccines are members of the normal microbiome and become pathogenic only when they translocate to and multiply in protected sites (81–83). Eradication of the entire targeted species in an individual through use of a vaccine may be difficult or impossible, because of the limitations of immunity at the mucosal sites where these species live and/or because of antigenic variation (84), which hampers the ability of a vaccine to target a whole species. For example, serotype-specific conjugate vaccination against S. pneumoniae has nearly eliminated targeted serotypes but has left overall carriage levels approximately unchanged (85, 86) because of strain replacement. Similarly, development of vaccines against Staphylococcus aureus and members of the hospital-acquired Enterobacteriaceae has been hampered by strain variation. The group B Streptococcus is an appealing target for vaccine development, in part because preventive treatment in the peripartum period with antibiotics carries an especially high cost by altering the microbiome of the newborn (discussed above) (49).

In such cases, the effect of a vaccine may be to shift the competitive balance within vaccinated hosts against strains of the bacterium targeted by the vaccine and thereby in favor of strains competing for the same niche that are not targeted (85, 86). This may affect the prevalence of resistance within the species. In the case of pneumococcal conjugate vaccines in the United States and elsewhere, the seven serotypes targeted by the original vaccine were disproportionately the serotypes associated with resistance (87–89). Thus, a welcome benefit of deploying that vaccine was a disproportionate reduction in drug-resistant pneumococcal infection (90). Focusing on the microbiome, total carriage of pneumococci did not change appreciably because of vaccine use (85, 86), but the serotype composition did change, away from vaccine types and toward nonvaccine types, which on the whole were less likely to be resistant. Unfortunately, this benefit was temporary, as resistant nonvaccine types became more frequent over time (85, 86). It was renewed with the switch in the United States to PCV13, which included the highly resistant serotype 19A (91, 92), but here, too, the disproportionate benefit against resistant strains may be temporary as nonvaccine-type resistant pneumococci increase in frequency.

Inspired in part by this serendipitous benefit of modulating the pneumococcal constituents of the microbiome toward drug sensitivity, a potential strategy to target directly the resistant variants of such pathogens with future vaccines has been proposed (93, 94). The idea here is that a vaccine that elicits immunity against resistance-conferring proteins or alleles could specifically reduce the prevalence of these resistance elements by shifting the competitive balance in the commensal microbiota in favor of drug-susceptible strains. Resistance-conferring proteins have been traditionally overlooked as targets for immunization because they may not be present in all bacteria of a targeted species and because they may be less immunogenic than other targets due to regulated expression or intracellular location. However, these cautionary notes may not apply in all cases (such as for cell-wall enzymes targeted by β-lactams). More importantly, mathematical modeling of the impact of such vaccines suggests that they could markedly reduce the prevalence of resistance even if their efficacy against resistant strains were very modest. The reason for this can be understood in two ways (93, 94). One is that such vaccines are not required to eliminate the pathogen altogether but only need to shift the competitive balance toward drug sensitivity; in this way, the competition from nontargeted (susceptible) strains augments the vaccine’s effect (95). Another is that the selective impact of antimicrobial treatment is intense but spatially limited to those (comparatively few) hosts receiving treatment at any given time. Such localized, intense selection could, in principle, be counteracted by much weaker selection against resistance that is distributed among a larger number of vaccinated hosts. To date this proposed benefit is only theoretical, but efforts are under way to test it in several species.

Effects of Vaccines on the Host–Microbiome Relationship.

Some emerging evidence suggests that an underappreciated effect of the pneumococcal conjugate vaccine is to decrease the rate at which commensal respiratory bacteria—including those not targeted by the vaccine—cause otitis media. It is believed that this effect occurs because the vaccine prevents severe episodes of otitis media from targeted bacteria in early infancy, and this in turn prevents inflammation-induced remodeling of the middle ear that would otherwise increase the risk that asymptomatic carriage of other bacteria could lead to otitis. To the extent that this effect is general, it represents a mechanism distinct from herd immunity by which a vaccine can prevent infection—by changing the host’s ability to “contain” its microbiome—and thus prevent resistant infections and antimicrobial use.

Recent studies from Israel of trends in otitis media show that the introduction of the 7-valent pneumococcal conjugate vaccine (later 13-valent) was associated with a decline in otitis media not only (as anticipated) from the pneumococcal serotypes included in the vaccine but also from those not included in the vaccine (96). Modeling has attributed this change to a decline in the “progression rate” of bacteria, i.e., the probability in a given unit of time that a strain or species present in the microbiome progresses to cause otitis media (96). Thus, the prevalence of nonpneumococcal constituents of the upper respiratory microbiome is approximately unchanged, and the prevalence of nonvaccine pneumococccal serotypes is increased since vaccine introduction, but these nonvaccine serotypes have become less likely to cause otitis media. This finding is thought to reflect vaccine-induced protection of the middle ear from early-life severe otitis, such that a given species or serotype in the nasopharyngeal microbiota is less likely to cause clinically apparent disease. If this phenomenon is generally true, modulating the host’s ability to prevent a potentially pathogenic member of the microbiome from causing illness could be another route by which vaccines prevent disease, including resistant disease, and consequent antimicrobial treatment (97).

By preventing disease and colonization by a targeted agent, vaccines produce differences in microbiota composition in vaccine recipients. Use of the 13-valent pneumococcal conjugate vaccine, but not the 7-valent vaccine, has been associated with higher diversity and stability of the nasal microbiota in immunized infants (98). An earlier study of the 7-valent pneumococcal vaccine revealed a shift in the composition of the nasopharyngeal microbiota at 12 months but not at 24 months postimmunization (99). The difference in the effects of these two pneumococcal vaccines on the microbiota was thought to be due to the 13-valent vaccine’s opening a more durable and larger niche that was filled by nonpneumococcal streptococci and anaerobes, but more effective disease prevention may have played a role as well. In studies of the rotavirus vaccine, short-term effects on the fecal microbiota have not been readily discerned (100, 101).

Future Directions

The interwoven relationships between the microbiome and host immunity offer important points of leverage that can be exploited for preventing and mitigating the emergence of AMR. Two basic strategies emerge: specific targeting of pathogens and antibiotic resistance and direct manipulation of the microbiota. Specific targeting includes the development of vaccines or narrow-spectrum antimicrobial agents against pathogenic and/or drug-resistant strains and species. Microbiota manipulation includes delivery of individual, small collectives or whole-scale communities of strains or of nutrients or other growth factors to strengthen or restore beneficial functions of the microbiota or exclude invasive and/or antibiotic-resistant strains and species. Both strategies may lead to a reduction in the use of broad-range antibiotics and hence a reduction in the selection for antibiotic-resistant organisms and the preservation of beneficial services by the normal microbiota. Both could also lead to the elimination or exclusion of antibiotic-resistant organisms.

Specific Targeting.

The development and use of vaccines against pathogens that warrant the use of disease-causing antimicrobials offer a number of attractive features and have been discussed above. Vaccine strategies directed against antibiotic-resistant organisms are less well explored and will require additional planning, including cost-effectiveness analyses. The challenges are numerous and include complexities of antibiotic-resistant strain and species diversity (102), the roles of some of these organisms as commensals, the consequences of opening niche space through their reduction in numbers, the limited information about antibiotic-resistance determinants per se as antigens, and the likely need for adjuvants or other approaches for manipulating host responses (103, 104). The known potent properties of certain commensals for inducing specific types of host immunity raise the possibility of using commensals as vaccine “platforms,” but more work is needed before the costs and benefits of this approach can be carefully assessed (105, 106).

The microbiome contains a broad diversity of small molecules and other products that deserve attention for their possible use in preserving native beneficial species and targeting invasive, resistant, and/or harmful species. Donia et al. characterized biosynthetic gene clusters (BGCs) in genomes of human-associated bacteria and discovered BGCs predicted to produce thiopeptide antibiotics. One was identified in the genome of Lactobacillus gasseri, a vaginal commensal, and was shown in purified form to have activity against vaginal Gram-positive pathogens but not against commensals (107). Others have discovered that some nasal isolates of Staphylococcus lugdunensis produce lugdunin, a thiazolidine-containing cyclic peptide antibiotic with activity against S. aureus. Colonization by these strains was associated with reduced colonization by the latter (108). CRISPR-Cas systems using suitably designed guide RNAs and effector Cas machinery that can be delivered with bacteriophages have been engineered to degrade chromosomal or plasmid-encoded ARGs. Challenges here include phage host range, evolved bacterial defense strategies, and human immune responses to the phage (109).

Microbiota Modification.

Manipulation of the microbiome, including the administration of probiotics, fecal microbiota transplantation (FMT) or transplantation of synthetic microbial communities, or the use of diet shifts or specific nutrient administration may reduce antibiotic resistance by displacing or excluding resistant organisms or indirectly by enhancing vaccine responses.

Traditional probiotics have been given concomitantly with oral vaccines in children with the goal of enhancing protective immune responses but have yielded mixed results (68). Alternatively, commensals with beneficial immune-stimulating properties might be engineered to present heterologous antigen as well. Some of the other commonly hypothesized beneficial properties of classic probiotic species might be expected to alleviate the use of antibiotics or to mitigate some of their adverse effects and by so doing reduce the ARG and multidrug-resistant organisms (MDRO) content of the gut microbiome (110). However, because of wide variation in the nature of probiotics, study populations, and immunization regimens, it remains unclear whether and how this general strategy might have value. The data so far do not provide consistent support for such benefits.

FMT appears to have some useful effects in reducing antibiotic resistance and is also a subject of vigorous study. It, too, as currently practiced, suffers from imprecision in content and procedure but suggests some potential paths forward. A number of case reports and case series have reported reductions in the types and frequencies of ARGs and/or colonizing MDROs and associated disease (111, 112) in patients following FMT. In some of these cases, FMT was deliberately undertaken for this purpose in the absence of the currently approved FMT clinical indication, recurrent Clostridium difficile infection (CDI).

Millan et al. (113) studied 20 patients with recurrent CDI who underwent colon cleanout after a course of an anti-C. difficile antibiotic and then FMT. CDI patients on average had greater numbers and diversity of ARGs than healthy controls in general or than the healthy stool donors used in this study. Those patients who had a favorable clinical response to FMT had a reduction in resistance-gene abundance and resistance-gene diversity; these reductions were sustained for as long as one year. The investigators found an inverse relationship between ARG diversity and overall bacterial diversity. However, there was no control group without FMT, so it is unclear how much of the effect might be due to the resolution of CDI and/or the avoidance of further antibiotic use (114). Patients were older than donors, and, as noted above, there is evidence that the relative abundance of ARGs increases with age (53). Furthermore, as Halpin and McDonald (114) point out, the investigators looked only at the relative abundance of ARGs, not at specific resistant strains or at the future occurrence of clinically significant infections caused by resistant organisms. There have been few formal assessments of fecal microbiota absolute counts and overall bacterial load during microbiome reconstitution with FMT after C. difficile colitis; thus, it is difficult to rule out a resistance gene dilution effect with the addition of resistance gene-impoverished donor microbiota. Of course, if the donor microbiota includes ARGs, they may become assimilated into the recipient microbial ecosystem, and donor microbiotas will need to be screened for these features (115).

The apparent success of this strategy for subjects with CDI as well as MDRO colonization may relate to the relatively unusual circumstances of the CDI gut ecosystem immediately following treatment, in which an inflammation-associated niche is closed and a normal health-associated habitat is partially restored, providing an open ecological niche for a health-adapted microbiota. Sets of related strains may engraft in an all-or-nothing manner (116). The success and stability of a restored or reconstituted health-associated community may predict the likelihood that invasive resistant organisms and resistance genes will be excluded and prevented from invading the restored ecosystem. In the many other circumstances in which resistant organisms and genes accumulate in the human microbiota and in which it is more difficult to turn off “drivers” of inflammation or in which inflammation is not so prominent, one would expect that microbiota restoration will be more difficult and that stable replacement with a low-resistance-gene microbiota will be more challenging. In a prospective study of eight immunocompetent patients colonized in the gut with either carbapenem-resistant Enterobacteriaceae (CRE) or vancomycin-resistant enterococci (VRE), FMT was tested for its ability to decolonize these patients of these resistant organisms (117). CRE and VRE were no longer detected in fecal specimens after three months in only three of these subjects; the overall microbiome composition before and after the intervention was not reported. Ongoing clinical trials in subjects with a history of recurrent MDRO infections are designed to determine if FMT will prevent additional MDRO infections (118, 119). Reductions in intestinal colonization and ARG content may be accompanied by similar effects at other body sites.

What about other approaches for enhancing or restoring the “colonization-resistant” properties of the microbiota, especially with respect to exclusion of pathogens and other organisms with relatively high ARG content? The mechanisms of colonization resistance include nutrient competition and direct bactericidal activity, immune stimulation, and various aspects of metabolic activity (120–122). Diet shifts may be associated with decreases in the types of organisms that often carry ARGs (123). When further elucidated, these mechanisms are likely to provide good leads for informed interventions to counter the burden of AMR.

In conclusion, the human microbiome can be viewed as providing myriad points of leverage in the effort to address and mitigate the growing problem of AMR. By manipulating or targeting microbiome membership, content, or activities in a deliberate and informed manner, one might enhance protective immune responses against pathogens and other “invasive species,” avoid the collateral damage of broad-range antimicrobials, and reduce the burden of AMR genes and resistant organisms in the human body. This is a struggle of “our wits versus their genes” (124).