Microbiota from environmental sources overlap and interact with human microbiota, contribute to human microbial diversity, and provide beneficial immunomodulatory stimuli. Meanwhile, reduced diversity in human microbiota and immune dysregulation have been associated with a range of diseases. Emerging evidence suggests landscape-scale drivers of microbial diversity may influence our health, but the area remains understudied because of its multidisciplinary nature. Here, we attempt to widen the view on this subject by offering an environmental researcher's viewpoint, proposing a unifying conceptual framework to stimulate multidisciplinary interest. To focus research in this challenging area, we propose greater emphasis on multiscale ecological links and that landscape-scale proxies for potential underlying microbial mechanisms be investigated to identify key environmental attributes and health relationships worthy of subsequent detailed examination. Wherever possible, ecological epidemiological studies should account for the temporal nature of environmental microbiota exposures, especially with respect to the early development of the human commensal microbiota.
People often express an intuitive sense that being in nature is good for their health. In addition to well-established risk-exposure scenarios in environmental health, modern scientific approaches are increasingly discovering that there are a range of nontrivial mechanisms and co-benefits linking natural surroundings, biodiversity, and human health influence (box 1). Better understanding these relationships may have important implications for developing cost-effective and mutually beneficial outcomes to help address simultaneous challenges in public health and biodiversity conservation (von Hertzen et al. 2011, WHO and SCBD 2015).
Increased exposure to anthropogenic hazards and environmental pollution
Increased exposure to natural hazards, including
harmful biotic agents: emerging infectious disease arising from land-use change, encroachment, biodiversity loss, and altered human–animal–environment dynamics
physical hazards: due to reduced buffering from extreme weather and other natural disasters
Declining food security and nutritional deficiency
Lifestyle: health benefits from exercise and sunlight influenced by surrounding natural environments
Mental health and social and cultural well-being: linked to natural surroundings and sense of place
Global change: including climate change, globalization, and conflicts over depleting natural resources
Biomedicines: loss of biodiversity-related potential new pharmaceuticals and traditional biomedicines
Reduced contact with protective environmental microbial diversity
The last mentioned mechanism in box 1 is among the least understood while also having wide potential to influence human health because of the largely hidden but ubiquitous nature of microbes (or microorganisms). Microbes have dominated the evolution of life and constitute a dominant portion of the Earth's living biomass and its genetic diversity (Whitman et al. 1998). Microbes feature in every habitat where life is possible. The various human microbiotas, or communities of microbes (e.g., gut, skin, airway, oral cavity, genito-urinary), exist in interdependent symbioses performing much of our metabolism (Wikoff et al. 2009). Their importance to human physiology is reflected in current knowledge that the combined human microbial genome (or microbiome) expresses over 100 times more genes than the human genome (Belizario and Napolitano 2015). Beneficial connections between microbiota and host health—influencing bodily development, mood, and stress responses—have been observed in both humans and animal models (Round and Mazmanian 2009, Rook et al. 2013, Belizario and Napolitano 2015). The human microbiota is believed to play an important role in normal human development (of organs, gut, immune system, bone, and brain) and actively participate in the homeostasis of the human body (McFall-Ngai et al. 2013). With important metabolic, immune, and nutritional roles, the human intestinal microbiota has been described as a “super-organism” (Purchiaroni et al. 2013).
Dysbiosis of the human microbiota (i.e., reduced diversity or changes in composition, often with an increase in the ratio of pathogenic to commensal organisms) has been associated with a range of immunological, gastrointestinal, metabolic, psychiatric, and behavioral disorders observed in humans and animal models, as has been reviewed elsewhere (Round and Mazmanian 2009, Clemente et al. 2012, Parker and Ollerton 2013, Rook et al. 2013, Belizario and Napolitano 2015). Ongoing research into particular microbiota–disease associations is supporting the increasing recognition of host microbiota–mediated mechanisms across diverse disease outcomes, such as in obesity (Ridaura et al. 2013), type 2 diabetes (Forslund et al. 2015), rheumatoid arthritis (Zhang et al. 2015), stroke (Yin et al. 2015), depression (Zheng et al. 2016), and some cancers (Sivan et al. 2015, Vétizou et al. 2015).
Multifactorial influences are known to drive the composition and diversity of the human microbiota, including diet, genetics, antibiotic use, age, birth mode of delivery (natural or caesarean), and geographic location (Clemente et al. 2012, Voreades et al. 2014, Belizario and Napolitano 2015). However, at least a portion of the human microbiota is in dynamic exchange with microbes from the surrounding environment; therefore, natural microbial diversity is now appreciated as an important contributor to normal (healthy) human immunological (and potentially other aspects of homeostatic) functioning (von Hertzen et al. 2011, WHO and SCBD 2015). Emerging experimental evidence also supports this line of thinking. For example, mice exposed to soil, house dust, and decaying plants had enhanced gut microbial diversity and innate immunity when all other variables (diet, age, genetic background, physiological status, and original gut microbiota) were controlled for (Zhou et al. 2016). In a separate study, mice exposed to dog-associated house dust experienced changes in gut microbiome that were associated with protective immune responses against airway allergens and virus infection (Fujimura et al. 2014). Rook (2013, figure 3) suggested several potential pathways through which environmental microbiota might affect the human microbiota and/or provide immunomodulatory stimuli. These pathways may include transient contact or colonization, with either direct recognition by immune receptors or indirect responses following interactions that alter the host microbiota.
Having emerged, in an evolutionary sense, from largely natural and biologically diverse surroundings, a growing proportion of the global population are now surrounded by relatively depauperate (low biodiversity) urban, industrialized, or highly managed and largely monocultural agroecological landscapes. As we discuss later, these macroscale changes can translate to microscale changes in biodiversity and ecosystem composition (Adams and Wall 2000, Bulgarelli et al. 2013, Turner et al. 2013). Meanwhile, the science of aerobiology (e.g., Womack et al. 2010, Polymenakou 2012, Bowers et al. 2013) shows that human populations have a real biological connection to their ambient surroundings (in addition to any direct physical environmental contact). These lines of evidence suggest that different sources and compositions of environmental microbiota—through interactions with the human microbiota and other immunomodulatory pathways (von Hertzen et al. 2011, Rook 2013)—may inadvertently provide protective or adverse background influences on human health.
Indeed, many medical researchers, including the World Allergy Organization, now suggest that microbiota-mediated mechanisms—and disruption to these, arising from environmental change—at least partly explain the pandemic of allergic, auto-immune, and chronic inflammatory diseases (AACIDs, discussed later) occurring across developed nations in recent decades (Haahtela et al. 2013). Described variously as the microbial old friends (MOF) mechanism (Rook et al. 2013), the high microbial turnover hypothesis (Matricardi and Bonini 2000), the biodiversity hypothesis (von Hertzen et al. 2011), or the evolutionary mismatch of “biome depletion” (Parker and Ollerton 2013), it is suggested that a lack of microbial diversity—or the reduced contact with the right type of microbes (or MOF, as we discuss later)—in our modern surroundings is an important contributor to the rising incidence of immune dysregulation underlying AACIDs and possibly a range of other diseases, including some cancers (Rook and Dalgleish 2011). Highlighting concerns (shared by von Hertzen et al. 2011) for the impacts of biodiversity loss leading to reduced immunoregulation from natural environments, Rook (2013) proposed that environmental microbiota (as supplements to MOF) may provide an unappreciated ecosystem service that is essential to our well-being, and that “this insight will allow green spaces to be designed to optimize health benefits and will provide impetus from health systems for the preservation of ecosystem biodiversity.”
However, large gaps remain in our knowledge: “Hardly anything is known about the interactions between environmental and indigenous [host commensal] microbiotas” (Haahtela et al. 2013). There are still many unknowns concerning the protective roles and membership of MOF, their possible modes of action, and broader relationships with biodiversity and the surrounding environment (Stanwell-Smith et al. 2012, WHO and SCBD 2015). Important research questions in the context of potential environmental microbiota–mediated influences on human health include the following: (a) Can we characterize environments through their microbiota? (b) What are the effects of macro- to landscape-scale environmental change and biodiversity loss on environmental microbiota? (c) Is landscape-scale biodiversity associated with human health outcomes? (d) Are different types, or conditions (qualities), of environment potentially more beneficial than others? (e) How might protective environmental influences compare with recognized drivers of human health such as socioeconomic status, diet, and lifestyle risk factors? (f) Can we identify and prioritize particular environment (or environmental change) and health associations to target subsequent detailed research? (g) Under what circumstances might environmental microbiota (or other microscale bioactive agents) be associated with health benefits? Answers to these questions may help to build insight and hypotheses and prioritize research opportunities before tackling more detailed investigations of possible environmental microbiota–mediated mechanisms.
To date, the MOF mechanism has principally been investigated from a medical research focus, with limited emphasis placed on the potential role and analysis of broadscale ecology or environmental change. This is despite a call to “bridge the chasm between ecology and medicine/immunology” (Rook 2013). Here, we further the argument for greater integration of ecological insight and environmental analyses into studying potential protective environmental microbiota–mediated mechanisms. In particular, we suggest that a comprehensive examination of broadscale, spatially variable environmental attributes in the context of spatially distributed public-health data may advance knowledge in this area. If we adopt the view that microbial diversity in the environment should be viewed as an inherent ecosystem service that is essential to our well-being (Rook 2013) and that this can be related to environmental biodiversity (von Hertzen et al. 2011), then we suggest that protective health effects should be observable and associable with recognizable environmental attributes (e.g., land use, vegetation, soil types, and their diversity), acting as proxies for as-yet-unknown microbial agents and mechanisms.
Immunomodulation, “old friends,” and the biodiversity hypothesis
Microbes play a key role in educating and regulating the immune system (Purchiaroni et al. 2013, Belizario and Napolitano 2015, WHO and SCBD 2015). Having co-evolved with a diverse range of microbes (and their metabolic and decay products) in the surrounding environment, the human immune system has needed to develop defense mechanisms against harmful pathogens, as well as tolerance mechanisms to other commonly encountered, and mostly harmless, microbial agents. As developed societies around the world have improved standards of sanitation, we have witnessed a decline in infectious diseases. However, in recent decades, this has been paralleled by a corresponding increase in AACIDs (Haahtela et al. 2013).
Initial attempts to explain this trend gained most attention via the hygiene hypothesis (Strachan 1989). However, this has since been revised and expanded and is perhaps most notably described in terms of the MOF mechanism (Rook et al. 2013). Alternatively, WHO and SCBD (2015) use the terminology “supplements to the human symbiotic microbiota from the natural environment.” The MOF mechanism suggests that following prolonged microbial exposure over evolutionary timescales, a dependence evolved between the immune system of mammals and some microorganisms. Possibly, this involved ancestral humans losing the need for gene expression associated with essential functions that could be readily performed by these partner microorganisms. In particular, this concerns a key function of the immune system in recognizing when not to activate in order to avoid unwarranted and potentially self-harming inflammatory responses to the body's own cells and normally harmless microbes from the surrounding environment.
Exposure to a broad diversity of microorganisms following birth (e.g., from vaginal delivery, diet, human contact, and the environment) provides important training inputs to the human immune system (O'Hara and Shanahan 2006, Wopereis et al. 2014). Microbes are sampled by immune cells associated with mucosal barrier tissues, prompting the establishment of complex immunoregulatory circuits that balance inflammatory responses (to suppress dangerous pathogens) with tolerance mechanisms that induce, for example, anti-inflammatory cytokines (signaling proteins) and regulatory T cells (Treg) in order to avoid undue responses to common antigens (O'Hara and Shanahan 2006, von Hertzen et al. 2011, Purchiaroni et al. 2013).
In contrast, AACIDs have been associated with immune dysfunction, dysbiosis, and inappropriate inflammatory responses to (a) our own tissues, manifesting as autoimmune diseases such as type 1 diabetes, multiple sclerosis, and rheumatoid arthritis; (b) normally harmless allergens and foods, manifesting as allergic disorders, eczema, asthma, and hay fever; and (c) gut contents including commensals, manifesting as inflammatory bowel diseases such as ulcerative colitis and Crohn's disease. These associations are reviewed in detail elsewhere (e.g., Round and Mazmanian 2009, Clemente et al. 2012, Parker and Ollerton 2013, Purchiaroni et al. 2013, Belizario and Napolitano 2015). The risk of AACIDs may be further enhanced by lack of physical activity and sunlight, poor diet, pollution and other factors, which may act in synergy with dysbiosis of the gut flora (Stanwell-Smith et al. 2012, Haahtela et al. 2013). Haahtela and colleagues (2013) also reviewed and speculated on possible connections between dysbiosis and AACIDs. They suggested that it is possible that some common members of the normal (healthy) commensal microbiota may play an active role in the development of Treg cells, responsible for mediating suppression of T-cell mediated inflammatory responses. They speculated that altered environmental microbiota may play a role in the development of dysbiosis, such as through the reduced signaling of pattern recognition receptors (used by the innate immune system to identify particular microbes and thereby amplify or suppress responses). Reduced immune signaling may then lead to immune dysfunction, which enhances the colonization and growth of a biased microbiota, thereby reinforcing the host–microbe interaction toward an unhealthy state (Haahtela et al. 2013).
Failing immunoregulatory mechanisms can also lead to continuous background inflammation, even without a specific chronic inflammatory disorder. Persistent raised levels of inflammatory mediators have been associated with increased susceptibility to a range of diseases including insulin resistance, metabolic syndrome, type 2 diabetes, obesity, cardiovascular disease, reduced stress resilience, and psychiatric disorders such as depression (Parker and Ollerton 2013, Rook et al. 2013, Belizario and Napolitano 2015). Several forms of cancer are also associated with increases in AACIDs, which may be explained because chronic inflammation provides growth factors and mediators that stimulate the vascularization and metastasis of tumors (Rook and Dalgleish 2011).
Temporal dimensions of human and environmental microbiota interactions also require consideration. Early stimulation is viewed as particularly crucial for supporting the maturation of immunoregulatory mechanisms (Wopereis et al. 2014) and dysbiosis during early developmental periods may have lasting adverse health impacts (Cox et al. 2014). However, immunoregulatory effects associated with dysbiosis (Parker and Ollerton 2013, Belizario and Napolitano 2015) and helminth infections (Versini et al. 2015) are also observed in later childhood and in adults, while the immune-boosting effects of mycobacteria (a suggested old friend) are known to be transient (Matthews and Jenks 2013), suggesting that ongoing diverse exposures are also important (Matricardi and Bonini 2000). Temporal effects are also discussed later in relation to the variability of environmental microbiota exposures and addressing confounders in more detailed work.
Drawing on multiple lines of evidence, von Hertzen and colleagues (2011) extended the notion of a MOF mechanism to suggest that “declining biodiversity might actually increase the risk to humanity from chronic diseases.” This idea arises because transient beneficial members of the human microbiota overlap with environmental microbiota, suggesting a dynamic interaction with the environment. As has been reviewed elsewhere (von Hertzen et al. 2011, Stanwell-Smith et al. 2012, Haahtela et al. 2013, Rook 2013, WHO and SCBD 2015, and references therein), the grounds for the notion of a wider association among dysbiosis, AACIDs, and a lack of biodiverse microbial stimuli from the surrounding environment come from (a) metagenomic studies of the microbiota in the gut and other sites; (b) epidemiological studies on immigrants moving to more affluent but more depauperate countries; (c) urban–rural AACID comparative studies; and (d) studies of immunomodulatory effects due to epigenetic mechanisms, farm and livestock exposures, proximity to agricultural land, and exposure to pets. Reduced exposure to biodiverse environments and urban green space is also suggested to partly explain the higher incidence of AACIDs associated with lower socioeconomic status (Rook et al. 2014).
Emerging evidence lends support to von Hertzen and colleagues’ (2011) biodiversity hypothesis. Hanski and colleagues (2012) found associations between atopic sensitization (allergic disposition), skin microbiota and surrounding land-use types in a random sample of 118 adolescents living in a heterogeneous 100-kilometer (km) by 150-km region of Finland. Atopic individuals had reduced generic diversity of gammaproteobacteria on the skin compared with healthy individuals. In contrast, healthy individuals showed a significant correlation between the relative abundance of the gammaproteobacterial genus Acinetobacter and expression of interleukin (IL)-10, a key anti-inflammatory cytokine in immune tolerance. Atopic sensitization was significantly explained by land use, decreasing with the amount of forested and agricultural land within 3 km of the study subjects’ homes. In cohort studies from Finland and Estonia, Ruokolainen and colleagues (2015) found that land-use patterns explained 20% of the variation in the relative abundance of proteobacteria on the skin of healthy individuals, and the amount of green environment (forest and agricultural land had similar effects) was inversely associated with the risk of atopic sensitization in children. They concluded that “the environmental effect may be mediated via the effect of environmental microbiota on the commensal microbiota influencing immunotolerance.” There are, however, limited studies of this type, and more research to test the biodiversity hypothesis in different environments is needed.
Why focus on environmental proxies?
Sources of microbial diversity in the natural environment include soil, vegetation, animals, and aquatic and marine environments. The environment is highly multifaceted, and here, we discuss landscape-scale attributes as potential drivers of environmental microbiota diversity and therefore health. We might expect macroscale environments to be linked to microscale environments through the provision of characteristic feedstocks and microhabitats. Changes to aboveground macroscale features can affect microscale ecosystem dynamics of terrestrial, freshwater, and marine systems through (a) changes in resource supply, (b) physical and structural habitat heterogeneity, (c) biotic (ecological) interactions, and (d) cross-surface migration of above- and belowground organisms (Adams and Wall 2000). Broadscale geographic variation may also contribute to variation in human microbiota; for example, Suzuki and Worobey (2014) suggested that higher latitude colder climates are associated with changing proportions of dominant bacterial phyla linked to increased body mass. The key environmental themes linked to sources of microbiota are highlighted below.
Vegetation and land use
Different plant species are associated with different microbiota of the phyllosphere and rhizosphere (i.e., microbial habitats of aerial vegetation and belowground roots respectively; Bulgarelli et al. 2013, Turner et al. 2013). The composition of and similarities between plant microbiota are driven by factors including (a) biochemically induced mutualism between particular plant and microbial species, (b) genetic relatedness between plants, (c) climate, (d) anthropogenic influences (e.g., pesticide use), and (e) spatial proximity (Bulgarelli et al. 2013, Bringel and Couée 2015). The connection between aboveground (macroscale) and belowground (microscale) components within terrestrial ecosystems typically results from powerful mutual feedback mechanisms. For example, plant characteristics will dictate organic matter inputs to soil microbiota while soil microbiota will dictate the breakdown and re-supply of nutrients to plants. These feedbacks will vary depending on the natural fertility and productivity of an ecosystem (Wardle et al. 2004) and also with anthropogenic changes in land use and management (Coleman et al. 2004).
It is also possible that a range of (nonmicrobe) microscale bioactive agents may provide immunomodulatory influences. For example, Li and colleagues (2006) found immune-boosting effects from phytoncides (wood essential oils), and Stanwell-Smith and colleagues (2012) suggested that protective agents may extend beyond the living MOF themselves, to include their cellular components (e.g., endotoxin), decay products, and metabolites. In view of this potential wider context for health influences from the environment, plants are also known to emit pollens, aerosols, and a wide variety of volatile organic compounds (VOCs; Bulgarelli et al. 2013). VOCs can promote or inhibit (and thus shape) adjacent microbial communities, whereas phyllosphere microbiota are also active in the production, interception, and alteration of various plant-related VOC emissions (Bringel and Couée 2015).
Through interactions with their surrounding environment, animals may inadvertently sample and collect a wide variety of environmental microbiota. The characteristic microbial sources most relevant to human interactions will likely include fur or hides and fecal matter. At the landscape scale, different vegetation and land-use types are often associated with different animals (e.g., livestock grazing or feedlots on agricultural land, native species in conservation areas, and pest species in poorly managed areas). Human exposure to animal microbiota will be influenced by proximity, the amount and volatility of source material, as well as prevailing winds for airborne microbiota. Bowers and colleagues (2013) measured airborne bacterial signatures of cow fecal microbiota in a rural city surrounded by agricultural land containing cattle feedlots. Exposure to pet dogs in early infancy has been shown to reduce the risk of childhood allergic disease development, and dog-associated house dust has been found to be associated with beneficial immunomodulatory effects (Fujimura et al. 2014). The microbiota associated with animals and farm exposures are further reviewed elsewhere (Stanwell-Smith et al. 2012, Rook 2013).
Soils are the most complicated biomaterial on the planet (Young and Crawford 2004). They support an immense diversity of microbes that remain largely unexplored, with drivers of variability in soil microbiota including variation in soil types and microhabitats (arising from environmental conditions, anthropogenic and organic inputs, and soil texture or clay content; Torsvik and Øvreås 2002). Microbes from soils have produced many of the most important medicinal drugs, including the majority of antibiotics and many anti-cancer compounds (Charlop-Powers et al. 2015). Soil eating (geophagy) is widespread in vertebrates and many human cultures, typically targeting clay-rich soils and suggested to provide protective health benefits (Young et al. 2011); this is consistent with the mechanisms discussed here. Particular soil constituents may have biological effects and seasonal mobilization patterns, such as has been shown in studies of coccidioidomycosis (valley fever) caused by a soil-dwelling fungus (Kolivras et al. 2001). Loss of contact with soil (and associated microbiota) has been suggested as a possible contributor to the rise in AACIDs arising from broadscale sealing of soils in urban developments (von Hertzen and Haahtela 2006).
Aside from soil itself, biological soil crusts can constitute up to 70% of the living groundcover across many diverse natural environments (Belnap and Lange 2001). These crusts form an aggregation of soil particles and cyanobacteria, algae, microfungi, lichens, and bryophytes that live in or on the top few millimeters of soil and may also be important contributors to beneficial human–environmental microbiota contact.
Coastal and marine environments
The marine microbiome is also diverse and largely unexplored, has biomass (cell densities) concentrated in near surface layers, and shares over 70% of microbial gene functionality with the human gut microbiome (Sunagawa et al. 2015). Mobilization, via aerosols, of bioactive substances associated with marine microorganisms—thus influencing the health of near-coastal human populations—is demonstrated through the adverse example of harmful algal blooms or red tides (Weinstein 2013).
Aerobiology demonstrates there is a real biological connection between humans and ambient environmental microbiota. The air is alive with microbial diversity—including bacteria, viruses, fungi, pollen, and algae—and acts as a source of both pathogenic and beneficial microbes to humans (Womack et al. 2010, Polymenakou 2012). Spatial and temporal variability may be expected in the composition of airborne microbiota. From sampling the near-surface atmosphere across three distinct land-use types (agricultural fields, suburban areas, and forests), Bowers and colleagues (2011) found that the composition of airborne microbiota was significantly related to land-use type and that differences were likely driven by shifts in the sources of bacteria rather than by local meteorological conditions. Also, Bowers and colleagues (2013) observed seasonal fluctuations in the composition and sources of near-surface airborne microbes, with soils and leaves representing important microbial sources across both urban and rural sites and cow fecal bacteria (associated with neighboring feedlots) also featuring in the rural location on a seasonal basis. They observed that microbial sources varied in prominence under seasonal conditions, potentially explained by climatic conditions, deciduous plant growth and senescence, and seasonal soil disturbance from surrounding agricultural land-use practices. Continental-scale patterns in the distribution of dust-associated bacteria and fungi have also been observed (Barberán et al. 2015) where geographic patterns were associated with climatic and soil variables. That work also found that urban areas were exposed to more homogenized airborne microbiota compared with the geographic variability found across rural areas.
Given the preceding evidence of relationships connecting various ecosystems (or ecosystem components), environmental microbiota, and potential human health effects, it may be possible to improve human health outcomes through environmental management specifically targeting microbiota-mediated linking mechanisms. We envisage theoretical links among landscape-scale environmental change, environmental microbiota, and human health, as we show in figure 1. However, in order to prioritize where more detailed study of underlying mechanisms and/or public-health interventions might be most cost effectively targeted, it is first necessary to quantify the strength of associations between environmental exposures and health outcomes. In the absence of temporal change data, we might examine these relationships using spatial analogues (environmental mapping) for differences in the landscape. We suggest that the links depicted in figure 1, while not comprehensive, may provide a useful conceptual framework for multidisciplinary research. In the next section, we briefly discuss methodological approaches for establishing priorities and addressing confounders, and we outline subsequent more detailed approaches required to advance knowledge in this emerging field of study.
There are a number of factors supporting the use of environmental proxies to investigate the MOF mechanism and related biodiversity hypothesis. Knowledge of the membership of MOF is incomplete and inconsistent (Stanwell-Smith et al. 2012); it is not known whether biodiversity, total biomass, or the particular source or species of environmental microbe(s) is important (Rook 2013), and there are still considerable computational challenges and base-knowledge limitations in trying to characterize and understand the genetic makeup and biological function of complex natural environments such as soil (Howe et al. 2014). Knowledge is still building on the microbiota of different environments, such as through the Earth Microbiome Project (Gilbert et al. 2014). A range of previously unappreciated (nonmicrobial) microscale bioactive agents from the environment may also be contributing to human health. There is a growing consensus that living in close proximity to the natural environment can provide a broad range of health benefits (WHO and SCBD 2015), but what type of natural environment? And are some environments better than others? Using a spatial analogue approach may help answer this question.
A focus on identifying particular environmental microbiota–mediated mechanisms affecting human health may also be hampered by redundancy that is likely to be found both in the microbial agents providing immunomodulatory stimuli and human immune system pathways (Stanwell-Smith et al. 2012). Microbiota can drive epigenetic responses (Shenderov 2012), bringing further potential complexity and requisite expertise to the examination of underlying mechanisms. Required doses are unknown, and it may be that subclinical (asymptomatic) exposures are all that is required to deliver protective health benefits (Stanwell-Smith et al. 2012). If so, this poses a challenge as subclinical exposure is much harder to detect in epidemiological studies. This means for subsequent detailed study into underlying mechanisms, immunological markers will be important, not just disease outcomes.
When it comes to analyzing environmental exposures, the question of required dose is worth exploring further. This is because emerging knowledge of the predominantly beneficial role of microbes (as we discuss here) is at odds with the traditional focus of microbiology, concerned with the negative role of microbes in driving infectious disease. Here, we raise the possibility that hormetic, U-, or J-shaped dose–response relationships (i.e., characterized by low-dose stimulation and high-dose inhibition; Calabrese et al. 2007) may provide a bridging paradigm between protective MOF and the traditional toxicological view of common pathogenic microbes by spanning divergent health outcomes inferrable from varying microbial dosage rates. For example, known pathogens including Escherichia coli, Helicobacter pylori, species of Salmonella and Staphylococcus, enteroviruses, and parasitic helminth worms are among those microorganisms suggested to have protective roles (Stanwell-Smith et al. 2012). Calabrese and colleagues (2007) suggested that hormetic responses are generalizable and commonly encountered across a range of biological systems.
In figure 2, we conceptualize an idealized dose–response curve for a generic MOF. When otherwise expected, missing or very low doses of a MOF are associated with greater risk of AACIDs and related adverse health outcomes (zone 1). Nominally low to moderate doses are associated with protective benefits (zone 2; due to appropriate stimulation of immunoregulatory circuits). Increasingly elevated microbial doses are expected to be associated with disease (zone 3). Figure 2 mirrors Calabrese and colleagues’ (2007) biphasic response curve except that instead of setting the reference response level at 100% of control (zero dose), by setting the reference response at some low–moderate dose range (perhaps corresponding to an evolutionary norm), three response zones instead of two are depicted. In this context, an evolutionary norm would correspond to long-term exposures to diverse environmental microbiota and natural allergens consistent with expected normal immunomodulation via the MOF mechanism. Such a curve also parallels the triphasic deficiency–adequacy–toxicity concept familiar in plant nutrition (Smith and Loneragan 1997), further supporting Calabrese and colleagues’ (2007) claim that such hormetic curves are generalizable across many biological systems. Selecting spatial environmental attributes (proxies) that might mimic varying amplitudes of microbial exposure (e.g., soil erodibility and soil microbial activity indices) and using natural experiments may provide a means to test (or at least build support for) this hypothesis of a hormetic relationship.
Ecological interactions (e.g., competition, predation, mutualism, and commensalism) operate at the microbial scale (Coleman et al. 2004), so applying general principles, we might speculate on the ecological context corresponding to the three zones in figure 2. In zone 2, where some low–moderate concentration of the MOF is present, this would correspond to an evolutionary norm or long-term steady-state microbiota—consistent with the establishment of immunoregulatory norms. Such a long-term, well-established microbiota is also suggestive of a balanced composition with maximal biodiversity (and therefore buffering to change) compared with that of the other zones. In zone 1, we might envisage that environmental conditions or microbial ecosystem dynamics have reduced the populations of the particular MOF. Such a shift in environmental conditions (e.g., feedstocks, temperature, air, and moisture) will likely favor the proliferation of another microbial species to fill the vacant, or newly emergent, ecological niche. The loss of the MOF with a rise in some other remaining species would correspond to an overall reduction in biodiversity of the microbiota. Alternatively, the original environmental microbiota could have been largely substituted, for example, where people have moved from rural to urban areas. In zone 3, we might speculate that it is actually the particular MOF that has been favored by a shift in environmental conditions. This would be at the expense of reduced numbers or loss of other species, also corresponding to a loss of biodiversity. In this hypothetical scenario, it is interesting to note that appropriate, protective doses (and exposures) to a particular MOF might be entirely consistent with exposure to a high diversity environmental microbiota. However, a chain of evidence would be required to test this hypothesis in detail, such as we have outlined in figure 1.
Other known microbial ecology mechanisms also underlie the importance of microbiota composition and diversity. Greater diversity suggests greater redundancy in gene functionality, as well as genetic adaptability (including horizontal gene transfer). Quorum sensing will also play a role, referring to intra- and interspecies signaling used to synchronize gene expression among bacterial groups to control production of, for example, antibacterial substances, disease-causing virulence factors, and immune-system suppressors (Belizario and Napolitano 2015). Alcock and colleagues (2014) suggested that through mechanisms such as quorum sensing, more abundant microbial species can coordinate their secretions to influence host mood and behavior and even manipulate host eating habits to increase their survival.
At the landscape scale, the highly faceted nature of the environment is reflected in the growing availability of diverse large-area environmental mapping data sets. Geographic information systems (GIS) are being increasingly used in epidemiology studies, including the use of spatial association (e.g., proximity analysis) to design surrogate exposure metrics to better understand environmental influences on disease (Nuckols et al. 2004). Environmental proxies to investigate possible relationships between environmental microbiota (and other microscale bioactive agents) and human health could include spatial measures (a) of relative exposure to particular environmental features or attributes (e.g., via GIS focal statistics calculations of proportions of different classes of vegetation, land cover, land use, or other themes within a predetermined neighborhood) that might subsequently be related to changes in airborne microbiota; (b) of biodiversity where we might expect to find positive correlations with human health outcomes; and (c) that might mimic ambient exposure to particular MOF (e.g., soil erodibility and soil microbial activity indices) in which we may find nonlinear (e.g., hormetic or U-shaped) relationships with health outcomes.
Spatial environmental mapping data vary from expert-assessed polygon-based thematic mapping to raster-based remote sensing data (with varying levels of processing and interpretability) and statistically based spatial predictive modeling or mapping for all manner of environmental attributes. Following the approach of McBratney and colleagues (2003), the predictive mapping of soil microbiota, for example, may be developed using a wide array of environmental variables as potential predictors. Predictors can be chosen to span various soil-influencing themes such as previously measured soil attributes, climate, organisms (including vegetation and land use), topography and terrain attributes, lithology, age, and spatial or geographic position. By extension, we might also consider a wide array of environmental variables as potential predictors, or proxies, for as-yet-undefined potential protective environmental microbiota and nonmicrobial influences, in a broadscale environmental correlation analysis with spatially defined public-health outcomes (or ecological epidemiological study). Such correlative studies could potentially involve tens to hundreds of environmental variables where many of these variables are often correlated. Traditional multivariate approaches such as principal-components analysis can deal with correlation in predictor variables; however, this may be at the expense of ease of interpretation, such as when attempting to compare the relative importance of environmental variables (which may have lower effect size) among other known public-health predictors (e.g., socioeconomic status and lifestyle risk factors).
Contemporary machine learning methods such as the least absolute shrinkage and selection operator (LASSO) penalized regression (Tibshirani 1996) are designed to tackle high dimensional problems with large numbers of (including often correlated) potential explanatory variables, and yield interpretable results. Using LASSO penalized regression modeling in the environmental correlation analysis of Liddicoat and colleagues (2015) enabled direct interpretation of the relative effect and direction of important environmental predictors from the size and sign of standardized regression coefficients. Using alternative methods such as the LASSO in ecological epidemiological studies may complement traditional multivariate approaches to highlight key environmental attributes to assist in hypothesis building and establishing priorities for subsequent work. Therefore, the availability of diverse environmental spatial mapping data sets, coupled with natural experiments that influence human health, can provide a wealth of data from which to draw key associations and thus point the way for subsequent studies to investigate causal links.
Limitations, confounders, and more detailed work
A complex interplay of factors can influence human health, including known confounders (e.g., socioeconomic status, diet, lifestyle risk factors, exercise, health support services, genetics, age, and sex) and environmental influences (box 1). A broad human health–environmental correlation analysis will obviously not restrict findings to environmental microbiota mechanisms. Follow-up work will be needed in those environments of interest to test connections (see figure 1) through characterizing environmental features and their related environmental microbiota, human exposures, interactions with human microbiota, immunomodulatory responses, and consequent human health responses. Natural experiments, whereby particular population groups can be found that provide inherent controls for other important confounding factors (e.g., diet, lifestyle, and antibiotic use) will assist this subsequent detailed work. Focusing analysis on lower socioeconomic groups—reflecting their stronger association with AACIDs (Rook et al. 2014)—or children—because of the important role of early immune stimulation (Wopereis et al. 2014)—may also assist in the identification of environmental microbiota–mediated health mechanisms.
Individual responses to environmental microbiota are expected to vary because of differences in host commensal microbiota. Studies in mice (Seedorf et al. 2014) investigating the colonization of host microbiota have shown that established indigenous host microbiotas are resilient to perturbation and resist colonization by foreign microbiota. However, they also found that in the case of germfree or gnotobiotic (with no or limited known microbiota) mice, with a limited suite of environmental microbiota sources, there are reproducible selective processes that can drive initially disparate host microbiota compositions of separate co-housed animals to converge to similar phylogenetic structures. This included colonization of host gut microbiota by foreign microbes from highly divergent environmental habitats (e.g., soil microbiota). From this, we might speculate that the immunomodulatory influence of environmental microbiota could be greatest on individuals with immature or compromised (dysbiotic) commensal microbiota. Voreades and colleagues (2014) found that short-term diet interventions may transiently alter the gut microbiota composition but that long-term diet changes are required to shift to a new steady state. If we were to extrapolate these results more broadly, this could suggest that the lasting protective influences of environmental microbiota may depend on long-term exposures. Important temporal factors would need to be accounted for in any subsequent detailed work, such as (a) the timing and duration of exposure to potential beneficial environmental microbiota, (b) seasonal variations in environmental microbiota sources, and (c) short-term fluctuations, succession, and maturation in host commensal microbiota (Clemente et al. 2012, Wopereis et al. 2014).
Recognized health drivers may also be correlated with underlying environmental variables. For example, biodiversity and landscape productivity can be drivers of local economic activity, which in turn can drive higher socioeconomic status of communities. Investigating a large number of (including often correlated) environmental variables presents obvious challenges in attempting to identify links between health outcomes and microbiota-associated environmental proxies. Despite this challenge, medical researchers are calling for new approaches that invest ecological knowledge (Rook 2013) and investigate multiple interacting environmental influences that may potentially act across multiple health outcomes (Myers et al. 2013).
In spatial epidemiology, there are typically trade-offs between the availability and spatial resolution of health and key contextual data. Often, some level of spatial aggregation may occur for privacy or data-reliability purposes; therefore, environmental parameters will also need to be summarized to match the available area-based health data. In these situations, there can be difficulty in separating influences because of the scale and availability of data as well as spatial variability versus differences in ecological processes. Also, Ruokolainen and colleauges (2015) found that the spatial scale of land-use description affects the ability to detect a significant relationship between land-use gradients and allergic disorder; statistically significant relationships were observed at intermediate scales from 2 km to 5 km. The potential for ecological bias and ecological fallacy (Elliot et al. 2000) also needs to be recognized. When possible environmental proxies for MOF via spatial mapping are examined, these will at best represent potential exposure (not dose). Mapping data for environmental variables is often extrapolated from limited field-truthed sites to provide exhaustive spatial coverages. Such data often carry uncertainty that is unquantified but may represent the best available knowledge. These limitations need to be borne in mind but should not be seen as roadblocks for the purpose of hypothesis building and pointing to areas where more detailed research is required.
A sequence of progressively detailed studies is envisaged (in the context of potential links in figure 1). As we outline here, we recommend that broadscale environmental correlation analyses be examined to firstly identify particular environments and health outcome scenarios of interest. Where possible, this may take advantage of existing environmental and public-health data sets. In areas of interest, prospective epidemiological cohort studies are then recommended when possible to further establish possible associations between environmental exposures and possible protective health outcomes. Studies will need to account for recognized confounders (e.g., demographics, diet, social indicators, lifestyle risk factors such as smoking status, and environmental pollution); temporal factors including the timing, duration, and seasonality of environmental exposures; and incorporate immune biomarkers to track asymptomatic (or subclinical) exposures. As we discussed earlier, a focus on children (reflecting the importance of early immune system development) and/or lower socioeconomic groups (reflecting a stronger association with AACIDs) may also assist in the identification of environmental microbiota–health mechanisms.
More detailed ecological epidemiological studies based on environmental proxies may lead to hypotheses that can subsequently be tested with experimental studies on animal models. This has been demonstrated elsewhere; for example, Hanski and colleagues (2012) reported a special role for the gammaprotebacterial genus Acinetobacter in enhancing immunotolerance and in increasing the expression of anti-inflammatory cytokine IL-10. Subsequently, Fyhrquist and colleagues (2014) reported strong support for this hypothesis with a mouse model. Further research to understand relationships between potentially beneficial environmental microbiota and corresponding recognizable environmental features (e.g., plant species and soil types) will also benefit subsequent implementation of public-health policy, such as in translating new knowledge of protective environmental microbiota–mediated mechanisms into new urban green-space design.
Microbes and other microscale bioactive agents provide a real biological connection to our surrounding environment and represent an understudied influence on human health. Emerging evidence suggests that microbial old friends and/or diverse environmental microbiota supplement human microbiota and may provide protective background immunomodulatory stimuli, whereas their absence may play a role in dysbiosis, immune dysregulation, and disease. To advance understanding, we advocate the use of environmental proxies as a pragmatic investigation tool. In this, we suggest that soils have been underrepresented in studies to date as a source of environmental microbial diversity with the potential for a protective role in microbiota-mediated human health. Similarly, the influence of different types of vegetation, land cover, and land use (among other themes) also remain largely untested. We suggest that comprehensive environmental correlation analyses examining recognizable environmental attributes and allergic, autoimmune, and chronic inflammatory diseases (as well as other dysbiosis-associated diseases) could help build understanding and provide greater focus for subsequent detailed studies of potential underlying microbiota-mediated mechanisms. In this way, we can advance the “eating of the elephant”—that is, we can provide a first step. The timing and duration of environmental microbiota exposures also require consideration, with respect to the establishment, maturation, and long-term stability of the human commensal microbiota. Knowledge gaps regarding potential sources of microbial old friends and their relationship with recognizable features in the environment need to be addressed, for example, to prescribe new urban design (green-space) health treatments. In short, it remains to be demonstrated convincingly that landscape-scale environmental influences can affect our human microbiota and health. However, the public-health implications of such a connection warrant further research into this area. Such work will ultimately inform concurrent improvements in environmental stewardship, biodiversity conservation, and human health.
We thank the anonymous reviewers and editors from BioScience and Professor Peng Bi from the University of Adelaide School of Public Health for their thoughtful comments and suggestions on earlier versions of this manuscript.