Disruption to the composition and function of the early life gut microbiome is now recognized for its role in irregular immune development (1, 2), metabolic disorders (3) and inflammation (4, 5). Several of these phenotypes have been reconstructed using animal models or epidemiological approaches providing a compelling link between aberrant microbiome development in early life and these negative health outcomes (6–9). Thus, if pandemic non-communicable diseases such as type 1 diabetes, obesity, allergy, and asthma are associated with impaired microbiomes during infancy, as suggested by emerging evidence (3, 10–15), then relevant modulation of the microbiome in early life provides a compelling solution for addressing the increasing public health burden associated with these diseases. However, evaluative parameters to identify desirable microbiome compositions and their potential interrelationship with health, are currently lacking.

The application of methods derived from ecological theory and evolutionary biology have been fundamental to elucidating the factors that shape the microbiome throughout the lifespan. In this work, we apply concepts from the “ecosystem services” framework (16) to guide the ecological assessment of the breastfed infant gut microbiome from a host-centered perspective. We first describe the ecological processes that shape and define the composition of the microbiome in early life. This description is centered on the hypothesis that human hosts select, via diet (human milk), for the enrichment of specialized symbionts that fulfill beneficial functions underlying the provision of ecosystem services that contribute to their fitness and well-being. We then propose a model in which the absence of these beneficial functions and the consequential reduction in one or more ecosystem services can be defined as dysbiosis. To demonstrate the applicability of the model, the discussion is centered on the coevolution of specialized bifidobacteria, namely B. infantis, for which clinical evidence is available (17). Finally, we summarize, as evidence for this model, large cohort studies indicating the absence of bifidobacteria in early life is associated with negative health outcomes.

Immediately following birth, the neonatal intestine becomes rapidly colonized by microbes from the mother and the surrounding environment. Infants delivered by cesarean section are more likely to become colonized by environmental microorganisms from the maternal skin, healthcare staff and hospital surfaces. Vaginally delivered infants come in contact with bacteria from mother's vaginal canal and the fecal microbiota (18, 19). From this initial load of microbes, the allochthonous, vaginally-derived and environmental species are then rapidly replaced by organisms adapted to the gut (20–23); however, the microbiome differences based on delivery mode persist over time (19). Nutritional resources that reach the gut are another major factor influencing the neonatal gut microbiome, in terms of both composition and function. In exclusively breastfed infants, human milk oligosaccharides (HMOs) represent the main nutritional resources for bacteria in the gut. As a result, the gut microbiome of exclusively breastfed infants exhibits lower alpha diversity and higher abundance of specialized taxa able to metabolize HMOs, namely bifidobacteria (24–27). In the absence of specialized infant-associated species of bifidobacteria, HMOs are under-utilized, resulting in excess resources with profound impacts on ecosystem function (Figure 1). Cessation of breastfeeding and the introduction of solid foods represent a major shift in the nutritional resource landscape and a more functionally complex community of microbes is then required to deplete the greater variety of dietary substrates reaching the large intestine (Figure 1).

Additional ecological events, including random processes, ultimately influence the overall composition of the infant gut microbiome; however, initial microbiome inoculation based on birth mode, and the subsequent environmental selection through the provision of selective substrates from human milk, are the two major ecological processes shaping the gut microbiome of breastfed infants (30–33).

Identifying a healthy gut microbiome in both infants and adults has proven to be a major challenge to the scientific and medical fields (34). Historically, diversity has been speculated to maximize functionality, in a generalization of the “insurance hypothesis” (35–38). However, diversity indices are of limited value alone and have proven insufficient to determine ecosystem functionality, or to categorize microbial ecosystems as healthy or unhealthy (37, 39) (see Box 1 for an in-depth discussion on the limitations of diversity). Moreover, taxonomic composition can be highly variable among individuals, while functions encoded by the gut microbiome are remarkably coherent (45) and breastfed infants across different geographies develop a common microbial functional core (15, 32, 33). This implies hosts are under a strong pressure to select high-fidelity microbial partners to maintain key ecosystem functions (38), and that breast milk establishes key niches that can only be occupied by specialized taxa (46) (Figure 1). Furthermore, given the host and its microbiome operate as a highly interconnected and co-evolved ecosystem in which interactions among members and community characteristics are governed by the principles of community ecology, we argue the evaluation of gut microbiomes can only be successful if based on ecological and evolutionary criteria. To this end, the ecosystem services framework has been implemented to link ecological processes of macro-ecosystems with elements of human well-being (47) and has recently been adapted to value the services of gut microbial ecosystems from a host-centered perspective (16, 48, Box 2). Therefore, we propose to use this framework to guide the assessment of the infant gut microbiome and to determine the ecological conditions within the gut that may increase host health and, ultimately, fitness.

In the following sections we outline three key ecosystem services: (1) supporting services; (2) provisioning services; and (3) regulating services underlying the relationship between the breastfed infant and the gut microbiome based on previously defined criteria (16). Specifically, we discuss the ability of B. infantis to efficiently utilize resources (i.e., HMOs) and produce organic acids as key functional traits that sustain the provision of these services.

Organic acids, including short chain fatty acids (SCFA), are the major metabolic products of anaerobic microbial fermentation in the gut and have demonstrated roles in human health (63, 64). In the breastfed infant gut, fermentation of HMOs into lactate and acetate depends critically on specialized primary degrader organisms that have the metabolic machinery to capture and metabolize these complex compounds (46). This process generates pioneer products (supporting service; Table 1) and releases energy that is otherwise inaccessible to the infant (provisioning service; Table 1). Selected strains of bifidobacteria and Bacteroides metabolize HMOs (24), but only B. infantis contains complete pathways enabling intracellular HMO-transport and degradation. Consequently, it is the only organism with the demonstrated capacity to significantly increase the production of lactate and acetate in the breastfed infant gut while simultaneously decreasing residual HMOs in the stool of breastfed infants (17, 65). In fact, in the absence of B. infantis, high concentrations of these HMOs are expelled into the stools of infants (1, 17, 66–68) which is a clear indication of low utilization of these resources (i.e., HMOs) in the gut, even when compared to infants colonized by other bifidobacteria. This observation highlights the importance of B. infantis in providing and provisioning services that underlie the overall function of the infant gut microbiome ecosystem (Table 1).

In addition to their role as pioneer substrates in the gut, organic acids and SCFA can enter circulation and directly affect the adipose tissue, brain, and liver (69–71). Acetate has been proposed to have an important role in inducing anti-inflammatory effects via the modulation of regulatory T-cells and anti-inflammatory cytokines (70), as well as improve mucosal epithelial integrity in the gut leading to protection from infectious disease in animal models (64). Lactate crosses the blood-brain barrier and functions as a modulator of neural activity, and is actively transported by gut epithelial cells (72–74). Acetate and lactate are also precursors of butyrate, which has anti-tumorigenic and anti-inflammatory properties and provides energy to gut epithelial cells (64). Overall, these microbially-produced organic acids have a major influence on host physiology. Thus, the presence of taxa able to efficiently metabolize HMOs into these key metabolites is critical to the delivery of fundamental ecosystem services that can affect the short- and long-term health of the growing infant.

One of the critical functions of the gut microbiome is to protect the immunologically naïve infant from acquiring exogenous pathogens and to prevent the overgrowth of opportunistic commensals (10, 75), a process known as colonization resistance (76, 77). Direct competition for resources, metabolic exclusion by production of organic acids, and indirect stimulation of the mucosal barrier system are well-characterized mechanisms by which the microbiome provides the host with this regulatory service (78). More competition for resources increases ecological stability at the expense of diversity by favoring the growth of specialized taxa, and limits the ability of invading microbes to establish and replicate (79). Thus, increased stability is central to the delivery of this regulatory service (Table 1) as stable ecosystems are inherently more resistant to external disturbances (42). In a clinical study, it was shown that colonization with B. infantis EVC001 significantly increases the stability of the infant microbiome (17). Moreover, consumption of HMOs by B. infantis produces acidic end-products mainly lactate and acetate, thereby altering the intestinal environment to prohibit the growth of pH-sensitive populations (e.g., Enterobacteriaceae and Clostridia) (69, 80, 81) including known enteric pathogens (17, 82), many of which carry antibiotic resistant genes (83–85). Further, the resulting high abundance of bifidobacteria contributes to maintaining intestinal barrier function through the production of acetate and tryptophan metabolites, and the reduction of mucus-eroding bacteria (86–89). Thus, the regulating services infants obtain from a microbiome abundantly colonized by B. infantis represents an archetypal model of protection, in which the host selects (via HMOs) microbial taxa most adept at strengthening epithelial defenses as well as creating biotic (i.e., competition for resources) and abiotic (i.e., pH) resistance barriers against invasion (28). A conceptual depiction of these concepts is shown in Figures 1, 3.

The application of traditional concepts from macroecology has proven successful in providing relevant insight into the ecological dynamics that govern the human microbiome (38, 49). According to ecological theory, ecosystem productivity can be measured by total biomass and by changes in the concentration of a limiting substrate (35, 62, 91). In the gut, dietary and host-derived carbohydrates are the primary resources for microbial metabolism (31, 32). Productivity of the ecosystem can thus be determined based on the efficiency of their utilization, and in concert, determining bacterial biomass (Table 1; Figure 3). Together, these two functions offer complementary and independent approaches to monitor productivity and to identify states in which the delivery of the ecosystem services is maximized. Thus, by combining evaluations of ecosystem productivity and the generation of ecosystem services we propose a model for the definition of dysbiosis of the breastfed infant gut as a low-functioning ecosystem, in which the gut microbiome community is characterized by (1) low stability even without perturbations (e.g., diet change or antibiotics); (2) high susceptibility to invasion by external taxa; and (3) low utilization of the available resources (i.e., HMOs). The alternative to dysbiosis or a “healthy” state is characterized as being a high-functioning ecosystem when the gut microbiome community is: (1) stable over time, (2) resistant to invasion by allochthonous bacteria; and (3) shown to exhibit a high conversion of HMOs to pioneer products and biomass of benefit to the host (Figure 3). Overall, by focusing on function, this model is agnostic to method and index of choice and provides a quantifiable and objective approach to evaluate the microbiome.

Humans live in symbiosis with the composite of microbial inhabitants residing in their intestinal tracts but the contribution of specific species to the overall ecosystem function and terms of the individual symbiotic relationships, which can range from commensal to parasitic, are less understood. Considering the ecosystem services the infant host obtains from selectively favoring the growth of B. infantis, it is evident that the symbiotic relationship is mutualistic. Free selectively consumed resources like HMOs are extremely rare in nature and the composition of HMOs is unique among mammals (92, 93). Of the thousands of species able to colonize the human gut, only a very limited number of species have the molecular machinery to utilize them (24). Within the genus bifidobacteria, only B. infantis encodes the complete set of genes required to transport, and intracellularly deconstruct and metabolize all the chemical structures found among HMOs (65), thus indicating maintenance of these genes is under strong selection. Indeed, since its discovery, B. infantis has so far been exclusively found in association with human beings (94) and phylogenetic analysis indicates humans and bifidobacteria have co-speciated (95). Taken together, the association of B. infantis and the breastfed infant host presents strong characteristics of an exclusive symbiotic alliance that has persisted over evolutionary timescales, whereby the human host requires the symbiont to access a significant portion of its diet (i.e., HMOs), while concurrently the symbiont benefits from the nutritional niche provided by the host. This concept is congruent with well-established models of coevolved symbioses (57, 96, 97).

Interdependent biological alliances are best understood in binary symbiotic models (57). One invariable lesson from decades of research in these model systems has been that aposymbiosis (i.e., the absence of the symbiont) can represent a major stressor to the host and often results in physiological and developmental deficiencies. For example in the well-characterized Squid-Vibrio model, external perturbations are markedly different between apo- and symbiotic squids (98, 99). Indeed, the presence of V. fischeri may help modulate the host stress responses (100). Similarly, the removal of nutritional symbionts (i.e., symbiotic bacteria that help their animal partners digest, absorb, and metabolize complex nutrients) is known to pose appreciable fitness costs to the host (96, 101, 102). Notably, the removal of vertically transmitted (from parent to offspring) nutritional symbionts has been shown to have the greatest negative impact on host fitness (102), which bears surprising parallels to the conspicuous depletion of B. infantis among infants with severe acute malnutrition (103) and with the inverse correlation between fecal pH and stunting (104). Further examples include aposymbiotic pea aphids which have reduced growth rates, attain a lower adult size, and are reproductively sterile (101) and fruit flies, for which the presence of the facultative symbiont Lactobacillus plantarum is critical to the growth and maturation of larvae ingesting nutritionally suboptimal diets (9). Together, these examples demonstrate broadly that the disruption of ancient symbiotic associations can have negative implications on the host.

All data indicate human infants have evolved to partner with key symbiotic gut bacteria specialized in metabolizing host-provided resources in the form of HMOs; however, it appears over time the role of B. infantis and the impact of its absence from the infant gut have become obscured, likely because the generational loss of B. infantis predates the advent of high resolution tools to investigate the gut microbiome. For instance, substantial fecal excretion of HMOs and high fecal pH are not considered abnormal, and considerable instability of the gut microbial ecosystem is considered normal in early life (33). However, historical records suggest bifidobacteria was once more prevalent among infant populations in developed nations than what contemporary reports indicate (105), and correlative evidence from large cohort studies suggest absence of this key symbiont comes with important negative acute and chronic health consequences during a critical developmental stage (2, 4, 103, 104, 106).

The importance of individual species to ecosystem function, and ultimately to the services, can become apparent through their loss. There is growing appreciation that interventions known to disrupt microbiome development may lead to the extinction of certain taxa across entire populations (107). Widespread antibiotic use, cesarean section delivery, and formula feeding are associated with altered gut microbiome compositions and subsequent negative health outcomes, including obesity and autoimmune diseases (3, 108, 109). In particular, the increased prevalence of these dietary and medical interventions has been associated with the decline of Bifidobacterium over the past century (20, 21, 60, 105, 110, 111). We pose the loss of critical functions in the gut resulting from the decline in the prevalence of B. infantis may have selected for microbiota that lack the resilience and stability during critical stages of immune and metabolic development. In fact, lower abundance of bifidobacteria has been associated with greater risk for developing colic, atopic dermatitis, asthma, food allergies, type I diabetes and chronic inflammation (2, 10, 11, 15, 112). Additionally, infants lacking B. infantis show signs of chronic enteric inflammation during the first 60 days of life (5), which has been directly linked to an increased risk of certain chronic disorders such as atopy and asthma later in life (113).

Interestingly, in geographic locations where breastfeeding rates are high and vaginal birth is widespread, Bifidobacterium is normally abundant in infant microbiomes (66, 114, 115). In contrast, the gut communities of infants in developed countries are largely unstable and highly variable (25, 111) and the distribution of Bifidobacterium is notably bimodal (26). This variation is clearly evident in a recent comparison of the gut microbiome of infants in geographically similar but developmentally diverse locations in which the level of Bifidobacterium was found to be higher in infants in more resource-limited locations, which correlated with decreased incidence of autoimmune and allergic diseases (2). Together these findings raise the question of whether the modern infant gut microbiome has been fundamentally altered from that of our ancestors and how the loss of key symbiotic species and the resulting disruption in immune development could be connected to the increased incidence of metabolic, autoimmune, and allergic diseases observed in developed countries today.

Fecal pH is another factor that has changed significantly over the past century and is consistent with the loss of Bifidobacterium (105). Fecal pH values directly correlates with the bacterial species colonizing the infant gut, particularly pertaining is the direct association between lower fecal pH and significantly decreased abundance of potentially harmful bacterial populations (i.e., Clostridiaceae, Enterobacteriaceae, Peptosteptoccocaceae, and Veillonellaceae) (105). These findings are intriguing, as an abundance of specific Enterobacteriaceae species induce gut inflammation (21), which has been positively associated with colic and crying in infants (116, 117). These adverse conditions may be due to the fact that Enterobacteriaceae-derived lipopolysaccharides induce stronger inflammatory activity compared with other lipopolysaccharide-producing bacteria (2, 118). In addition, lower fecal pH has been shown to be associated with better anthropometric growth scores (104) and improved thymic growth, a sign of immune system development (1). This may partially explain why B. infantis-colonized infants exhibit more robust vaccine responses (66, 119) and why there is a reduced incidence of autoimmune diseases in populations colonized with high levels of Bifidobacterium (2). Taken together, these data indicate functions provided by key symbiotic partners (i.e., B. infantis) during infancy have a strong impact on development and conversely, the absence of these taxa can have negative health consequences, underscoring the need to restore specific beneficial taxa to the infant gut.

With the growing recognition of the role of the microbiome in human health, the incorporation of microbiome-based diagnostics will inevitably become routine. In fact, a number of commercial tests are currently available to the general public and physicians are increasingly requested to interpret test reports. However, we currently lack a “gold standard” for what constitutes a healthy microbiome. Here, we proposed an anthropocentric model whereby gut microbiome function is determined in terms of ecosystem services that ultimately benefit the infant. Thus, microbiome composition can be evaluated objectively with regard to its contribution to host health, facilitating interpretation by health professionals. Furthermore, linking functional traits to specific ecosystem services may assist both the development of prognostic tools of infant microbiome function and probiotic interventions aimed at restoring the ecosystem services of the infant gut microbiome.

However, it is important to recognize that this model is limited to conditions in which the nutrient landscape in the gut is shaped by a single nutritional resource (i.e., HMOs) and will have to be re-validated for conditions known to shift the type and amount of resources as well as the distribution of biomass. Such conditions include, the introduction of complementary foods, formula feeding, antibiotic use and other microbiome-modifying practices. Moreover, the principles on which this model is based may be affected by stochastic events including niche pre-emption (i.e., “first come, first served”) driven by priority effects (120). We also recognize the overall dynamics of the infant microbiome involve complex intra and inter-species interactions which are not considered in our model. For example, the ecological relevance of Bifidobacterium species other than B. infantis, which are known to have limited capacity to metabolize HMOs but are found in the stools of infants, is currently unknown (121). Additionally, future models should aim to integrate the non-bacterial microbial inhabitants of the microbiome (e.g., virus, archaea, fungi, and other eukaryotes) which are increasingly recognized as important functional components. Nevertheless, the ecological principles presented here, can be broadly applicable to other host species, and evaluation of additional body sites, and can be adapted to inform the selection of taxa that may be relevant for health in other stages of life.

Lastly, we hope this work encourages the field to propose analogous models that incorporate ecological theory and testable frameworks to identify microbiome characteristics that are conducive to health or disease.

RD, BH, GC, and SF drafted and wrote this manuscript. All authors are responsible for idea conception, critical evaluation, and manuscript review.