Agriculture currently uses 11% of the world’s land surface for crop production. Since 1961, while total cultivated land has shown a net increase of 12 percent to 2009, land under irrigation has more than doubled (FAOSTAT)¹. Farming activity has escalated since World War I to reach massive proportions. The world cattle inventory in 2018 is at one billion heads, with half of these animals in India and Brazil, and the third-most in China². Data from the Food and Agriculture Organization of the United Nations indicates that the world’s average stock of chickens is estimated at almost 23 billion, and pigs account for 770 million (Lawrence, 2019; Metcalfe, 2019). Interestingly, such an “animal invasion” has frequently occurred in combination with a decline in animal diversity due to anthropogenic selection of a limited range of animal varieties of economic interest. The predictable effect is the increased possibility of interactions among large numbers of a few animal types with large numbers of humans.

There has also been a spectacular increase in the population of particular crop plants due to the technologically driven “Green Revolution” starting in 1950s and 1960s (Matson et al., 1997), which has recently intensified by intercropping, growing two or more crops in proximity (Martin-Guay et al., 2018). We should not forget frequent animal co-culturing (interbreeding), or animal and plants co-production. The net result is the enlargement of fields promoting a mixing ecology of vegetables, animals and humans. For instance, intensively managed rice farming paddy soils might constitute unique agroecosystems, providing opportunities for mixing different microbiomes (Tanskul et al., 1998). As we will state in the next section, the dimensions of this mix are and will be critical in shaping the problem of antibiotic resistance.

However, the density of contacts also depends on anthropic local interventions, which are deeply influenced by sociology and economics. Urban ecosystems have established barriers to exclude contact with rural (farming) areas dominated by animals and plants. Such barriers are still weak in some developing countries; thus, the number of interactions is high. However, the challenge of feeding 11 billion people in 2050 implies an increase in contact between humans and animals in the agricultural use of water, antibiotics, and fertilizers, all of which are risk factors for the development of infectious diseases and antibiotic resistance (Rohr et al., 2019). In summary, if some general trends occur at a global scale, such as the increase in the population of humans, food animals, and plants, the investigation of local conditions influencing One Health dynamics will become essential to shaping the dimension of the local risk of interactions based on microbiome merging (Ley et al., 2008; Pehrsson et al., 2016; Flandroy et al., 2018; Trinh et al., 2018). Note that microbiome merging is expected to occur preferentially among hosts sharing some basic common features. For instance, the intestinal microbiota of most vertebrates is dominated (in various proportions) by the same taxa, typically Firmicutes, Bacteroidetes, and Proteobacteria, regardless of whether the host is herbivorous or omnivorous, including marine mammals; a quite different pattern is obtained in invertebrates (Ley et al., 2008; Nelson et al., 2015; Colston and Jackson, 2016). A more detailed work is waited for lower taxa, more dependent on the habitat conditions. However, at the level of species, the intestine-adapted microorganism Escherichia coli is ubiquitous in mammals (Gordon and FitzGibbon, 1999). In fact, the coincidence of taxa reflects the fact that evolution of animals has occurred in parallel to the evolution of their microbiomes (McFall-Ngai et al., 2013). The terms “phylosymbiosis” was in fact coined to refer to the concordance between a host phylogeny and microbial community (microbiome) dendrogram (Theis et al., 2016). Horizontal accessory gene transfer, and the potential spread of antibiotic resistance genes, occurs preferentially among phylogenetically related bacteria (as is illustrated in Figure 1), even though transfer might bridge different taxonomic levels (Boto, 2009).

It is also worth mentioning that most probably one of the first barriers to preventing colonization by bacteria originated in other hosts (eventually antibiotic resistant) is the presence of a dissimilar microbiome able to outcompete the alien novel microorganisms. Gut microecology is based on ensembles of bacteria that have evolved in protocooperation or synergy; if the “right partners” are absent, alien organisms are difficult to establish, or might produce dysbiosis and reduction in diversity. For instance, because of the influence of pig’s microbiota, the microbiota of farm workers is less diverse than in neighbor villagers (Sun et al., 2017). Such reduction in microbial diversity decreases colonization resistance (Van der Waaij et al., 1971; Buffie and Pamer, 2013) facilitating acquisition of antibiotic resistant strains. That has a correspondence in the host-occupied environments, as it was observed in built environments (as private homes, workplaces, hospitals) where a loss of microbiomes’ diversity following interventions to decrease microbial load correlates with an increase of antibiotic resistance (Mahnert et al., 2019).

The increasing density and connectivity of individuals of different species (humans, animals, plants) produces a major increase in the total size of a limited number of particular microbiomes, facilitating microbiome merging, a condition for the interbacterial spread of antibiotic resistance genes (Hernando-Amado et al., 2019; van Bruggen et al., 2019). This effect is possibly modified by the short lifespan of farm animals, reaching their slaughter age much earlier than their potential maximum life expectancy, approximately 10 times earlier for cattle, 50 for pigs, and 100 times for chickens³. On one hand, the slaughtering process eliminates the microbiota, including antibiotic-resistant populations (except in the case of slaughterhouse wastewater contamination). On the other hand, the replacement with newborn animals implies an intensive buildup of new microbiotas (high rate of microbiota reproduction). During the construction of the microbiota in each new individual, not only the access of microorganisms from other hosts and environments is facilitated, but the higher bacterial (and community) replication rate in a underexploited nutrient-rich habitat provides more opportunities for genetic exchange, and if antibiotics are present in the environment, selection of resulting resistant populations.

The emergence of novel opportunistic pathogens of non-human environmental origin, particularly among particular species or clones in Enterobacterales, Acinetobacter, Pseudomonas, and Enterococcus associated with the use of antimicrobial agents, might be linked to this process. When involved in clinical infections, these organisms necessarily colonize the mucosal surface of the host and interact with the local microbiota, most probably with phylogenetically close organisms sharing similar Hutchinsonian niches. An Hutchinsonian niche an imaginary space with many dimensions (hypervolume), in which each dimension or axis represents the range of some environmental condition or resource required for the optimal growth of a sublineage or genotypic group (Holt, 2009).

Environments under stress (natural or anthropogenic, including antimicrobial agents) tend to reflect reduced microbial alpha diversity, the number of species/clones harbored in particular microbiota (Rocca et al., 2018). To a certain extent, stress tends to produce species-deficient habitats, providing the opportunity for alien colonizers to invade. Stress-driven reduction of species and clones reduces the diversity of microniches that are dependent on their functions. Empty niches (in terms of resources that are not exploited) act as attractors for members of neighboring organisms. Many of these free-microbe-specific microniches are now occupied by more stress-resistant, less niche-specific organisms, which originated from other microbiotas. As in the case of the naturally bacterial-free intestinal habitat of newborns, the opportunity arises for microbiome collisions and hybridizations. Reduction in alpha diversity could be expected to reduce the possibility of asymmetric dispersion of a variety of rare organisms in neighboring habitats (reduction in beta diversity). However, the collision of microbiotas creates new configurations in which rare taxa could emerge (Rocca et al., 2018, 2019). The application of zeta diversity metrics (measuring the degree of overlap in the type of taxa present between a set microbiotas) will be most useful to illustrate the spatial structure of multispecies distributions in various environments, and therefore the dimensions of microbiome merging (Hui and McGeoch, 2014).

Certainly, the case of mother-child microbiome transmission illustrates the importance of shared microbiomes among closely related families of hosts (Arrieta et al., 2014). In fact, this is a case of “microbial community reproduction,” given that not only individual cells, but communities such as microbiotas also reproduce and evolve (Baquero, 2014; Gordo, 2019). As in other cases of biological reproduction, the reproducibility of the original microbiota pattern is imperfect, that is, not maintained in its integrity; thus, differences can be detected among individuals (Vallès et al., 2014). Most importantly, the reproduction of a recently acquired microbiota among closely related newborns maintains a highly conserved composition; with time, however, “microbial evolution within hosts” both in terms of species genetic evolution and migration of strains, takes place (Theis et al., 2016), increasing differences between them, which is beta diversity (Garud et al., 2019; Gordo, 2019). Thus, animals that are slaughtered early should have a more homogeneous microbiota, and the amount of this particular type of microbiota should increase in the environment.

In a shared environment, the dispersion of microbiota in highly related hosts, for instance in the same or closely related species and under similar dietary regimens (Moeller et al., 2013; David et al., 2014) ensures that the more advantageous variants emerging in a particular individual (Garud et al., 2019) can be spread to the other members of the host’s community. To a certain extent, there is a “collective microbiota optimization” favoring the health of the herd. This phenomenon is the other side of the coin of that giving rise to deleterious epidemics. Indeed, microbiota homogenization in ensembles of hosts is a major factor facilitating specific interactions and genetic transfer.

The process of microbiome merging and hybridization that might give rise to (at least partially) novel assemblies of bacteria originated in different environments is a phenomenon known in ecology as “community coalescence” (Rillig et al., 2015; Rillig, 2017). The combined increase in number and collapse in diversity of animals interacting with humans should facilitate reiterative coalescence events between the same microbiotic types, and thus interbacterial gene transfer.

The dynamics of microbiome merging are insufficiently understood (Roughgarden et al., 2018). New observations, suggesting the modular structure of microbiota (Earle et al., 2015; Tropini et al., 2017), indicate the possibility of a “recombinational merging” within and between microbiomes, eventually resulting in emerging taxa and emerging communities (Rocca et al., 2018, 2019). In the soil, specific microbial aggregate communities can be considered “microbial villages,” periodically connected through wetting events, where soil moisture is increased as a result of rainfall infiltration, allowing for the transfer of bacterial organisms and genetic material (Wilpiszeski et al., 2019).

Fecal transplantation provides an excellent case study for microbiota coalescence, at least in closely related types of hosts. In fact, bacterial species that might graft in the receptor host could probably be predicted by the abundance and phylogeny of bacteria in the donor and the pre-transplant receptor (Smillie et al., 2018). In the immediate period after transplantation, the invading microbiota from the donor tends to prevail, however, the before-transplant microbiota tends to progressively be restored, typically after 3 months (Seekatz et al., 2014). This is a type of “resilience effect,” where the recovery of the remaining minorities of substituted populations, or the reacquisition of the lost strains, or their functional equivalents from the environment, reconstruct the original pattern of the microbiota (Allison and Martiny, 2008). However, novel strains or species introduced by transplantation procedures can colonize without necessarily replacing the indigenous strains or species of the recipient; in fact, the presence of a kin-strain might facilitate colonization (Li et al., 2016). Thus, for a prolonged period of time there is a transient hybrid microbiota in which new consortia can be established, facilitating genetic interactions.

Intermicrobiome communication can be facilitated by shuttle bacterial species (or clones within species) belonging to generalist taxa, able to multiply in the microbiomes of various hosts. In fact, a part of these taxa can be considered “colonizing opportunistic pathogens” (Price et al., 2017). The specialist-generalist paradigm predicts that specialists will have local advantages (narrow resource utilization but high performance), and should predominate in specific microbiotas; whereas generalists, which are probably less abundant locally (broader resource utilization but lower performance), are able to colonize diverse microbiotas (Mariadassou et al., 2015). Generalist taxa are identified by a wide Levin’s niche width index, detecting a broad range of niche conditions that a species could inhabit and successfully survive and reproduce, which can be obtained considering the proportion of operational taxonomic units (reflecting the bacterial diversity) in various microbiomes (Pandit et al., 2009). These taxa act as “microbial hubs” in scale-free networks, linking diverse microbiotas. Eventually, these shuttle taxa could have a deeper effect on the architecture of the recipient microbiota than could be expected by their abundance (Muller et al., 2018), thus representing “keystone species” (whose presence facilitate the establishment of many others) such that these species disappear or have reduced fitness, and the health (optimal composition) of several apparently unrelated microbiomes might be impaired (Berry and Widder, 2014).

Incoming bacteria might compete for those organisms in the recipient microbiota sharing the same function (functional redundancy) (Koskella et al., 2017). The complexity of most natural habitats likely frequently precludes the extinction of competing organisms, following a kind of “imperfect exclusion principle.” Indeed, microbiome merging depends on the local metabolic landscape, which largely determines the “Hutchinsonian niche” of bacterial species and communities (Holt, 2009).

Certainly, the degree of microbiome merging could be positively influenced by bacterial species niche overlap, which occurs when incoming and recipient organisms share the same resources and other ecological factors (Cornell, 2012; Moeller et al., 2013). Local microecological similarities between some areas (habitats) in the colonized hosts or between hosts and external environments should facilitate merging. However, available data regarding the microecology of colonizable habitats, such as the intestine of humans and animals, remain scarce (Baquero, 2015). Progress in metabolomics and metabolic reconstruction will soon remediate this important gap (Abubucker et al., 2012). In fact, the structure of microbiomes is ecologically determined by their metabolic networks (Muller et al., 2018). The field termed “metabobiomics” has been suggested to study the correlations between the composition of the intestinal microbiome and the metabolome (Xu et al., 2015).

An important but insufficiently explored issue is the role of ecological gradients, containing a series of partially overlapping niches, in the bridging process of microbiome merging. For instance, members belonging to different host microbiomes (including animals, plants, soils, humans) might transiently coexist in inland water sewage, wastewater treatment plants, in contaminated agricultural puddled areas, or simply in the soil of farms and human habitats in regions with poor sanitation (Baquero et al., 2008; Berendonk et al., 2015; Pärnänen et al., 2019). Transient, but reiterated coexistence between various microbiotas (or microbiotic modules) might provide the opportunity for new associations of bacteria originated in different hosts, led by generalist taxa. In areas with poor sanitation, such associations could be introduced by continuous exposure to contaminated water or food in the microbiota of humans or animals, providing the opportunity for microbiome evolution, eventually reaching a generalist-like, “broader spectrum microbiota” (Fondi et al., 2016).

Ecological connectivity is certainly the basis of the formation of gene exchange communities, as well as of common mechanisms of niche construction or even of task distribution (Smillie et al., 2011; Fondi et al., 2016). However, gene exchange might favor the members of the community in an asymmetrical manner, particularly if the transmitted element is of great profit, as occurs in the case of antibiotic resistance. This asymmetry could have consequences for increasing the population sizes of the organism under selection, and thus for connectivity with other populations. In this regard, it would be important to document whether the recent spread of MGEs carrying antibiotic resistance genes across various microbiotas is contributing to remodeling (maybe expansion) of the borders of gene exchange communities, by recruiting novel partners able to communicate with new potential gene receptors. This possibility is in line with the concept of cumulative genetic evolution or “genetic capitalism,” in which the more adapted organisms increase in population size and consequently in connectivity and genetic interactions (Baquero, 2004).

One aspect that is not yet fully understood is the construction of gene exchange communities, particularly when the microbiome emerges de novo in virgin habitats, as in newborns (Skippington and Ragan, 2011; Boon et al., 2014; Mansfeldt et al., 2019). An important aspect to be considered is whether these communities “replicate” in newborns, maintaining an identical member composition, or whether new members (originated in other hosts or environments) are accepted in the genetic exchange club in these early stages of microbiota construction. A model based on puzzle construction (Baquero and Nombela, 2012) suggests that the building of the microbiota depends (as in the pieces of a puzzle) on the successive mutual recognition of the components of the community, which is independent from the order of accession. However, this is a “degenerated puzzle,” so that various pieces (different but functionally related species) can occupy the same space and establish the same, or very similar, interactions with the other pieces. This situation provides an opportunity to create variant genetic exchange communities.

Under natural circumstances, transenvironmental colonizers are probably a minority among those that are transferred. However, exposure to high inocula and/or population amplification by antibiotic selection can facilitate local adaptation, successful colonization, and integration into new genetic exchange communities (Baquero, 2018; Sheppard et al., 2018).

Communication between microbiomes is a condition for the propagation of genes between bacterial populations. Let us first clarify that the concept of antibiotic resistance genes is extremely anthropocentric. With few exceptions, possibly antibiotic producers (Davies, 1990), antibiotic resistance genes were not born to resist antibiotics (Linares et al., 2006; Martínez, 2012). They simply belong to a large pool of genes encoding paraphysiological adaptive functions. In this section, we will focus on trans-specific mobile antibiotic resistance genes: those that can be detected, with a high degree of nucleotide identity, in various bacterial species.

The pangenome is the gene repertoire of a given bacterial species, that is, the ensemble of all genes contained in all individuals within the species (Tettelin et al., 2008). In these studies, the notion of “species” should be defined in robust way, the ensemble of organisms with at least 95% of average nucleotide identity, as obtained in all-versus-all sequences comparisons. In most species, the pangenome is much larger than the “core genome,” accounting for genes that are contained in every individual of the species, involved in the basic machinery of cell functioning. The difference between the pangenome and the core genome is due to a collection of genes that can be present or not in a given population inside the species, unique genes responsible for functions that are adaptive, niche-specific, and eventually of a contingent nature. The genes are considered “dispensable” (at least in basic culture conditions) or “accessory” (complementing the core genome). The fact that they are not present in all individuals of a species means that they can be gained or lost; in addition, the same genes (with high sequence homology) can be found across various species, indicating interspecific mobility (Segerman, 2012). Interspecific gene mobility can be neutral if the genes are transported and acquired unspecifically only because they are hosted in MGEs, but do not provide a current benefit for the recipient bacteria. These genes can, however, be “markers” to trace genetic transfer. In many cases, interspecific accessory gene flow has an adaptive function, and the transmitted genes are critical for survival in particular environments and contribute to bacterial eco-specific diversification (Wiedenbeck and Cohan, 2011; Rouli et al., 2015). Of course, acquired antibiotic resistance genes are accessory genes, and are transferred among microorganisms by using the same MGEs than other pre-antibiotic accessory genes encoding inhibitors-resistance, including those determining resistance to heavy metals (van Hoek et al., 2011).

The “convergence of adaptive needs” among bacterial species should foster interspecific communication. Antimicrobial exposure is forcing many different organisms to survive, and there are a limited number of genes able to provide resistance. If these genes are carried by the mobilome of a particular microbiota where these species can coexist, an increased possibility of interspecific genetic transfer is expected to occur. This transfer suggests the interesting possibility that antibiotic exposure could trigger interspecific gene flow. Certainly, the accessory genes (constituting most of the pangenome) reflect the ecological needs of organisms and might be useful to redefine species and subspecies (Laing et al., 2010; Caputo et al., 2015).

Species located in very stable, reduced, highly specialized niches are less exposed to the gene-traffic circuit; thus, their pangenome is close to their core genome (Martínez et al., 2017; McInerney et al., 2017). Significant examples are Listeria monocytogenes, or Legionella pneumophila, able to exploit intracellular (stable, isolated) niches, which have larger core genomes (“closed genomes”) than other members of their phylogenetic relatives, indicating less exposure to horizontal gene transfer (Gomez-Valero et al., 2011; Collins and Higgs, 2012). On the contrary, many bacterial organisms of importance in public health, and particularly those able to colonize different environments, have an “open pangenome” that is open to the immigration (capture) of a wide variety of genes. In the case of E. coli, a recent study has estimated a pangenome of 15,950 genes in 60 strains, 13,076 for the accessory genome and 2874 for the core genome (Her and Wu, 2018). In one of the largest available studies of E. coli (more than 2000 genomes), the authors estimated 3188 core gene families (defined as being present in 95% of genomes) and approximately 90,000 unique gene families (Land et al., 2015). The discovery of a cumulative number of new genomes in species with “open pangenomes” suggests that the number of potential accessory genes has no real limit (Lapierre and Gogarten, 2009). Why does this massive amount of horizontal genetic flow not cause a significant phylogenetic disruption in bacterial species? Probably because the preservation of the species’ “core genome” in different circumstances and environments is assured by the acquired accessory genome (Ochman et al., 2005). On the other hand, in most cases, the recent origin of these accessory genes can be traced in organisms sharing a common or convergent eco-evolutionary history with the receptor (Smillie et al., 2011; Fondi et al., 2016). This asymmetrical pattern of gene transfer allows us to identify highways of gene sharing (Beiko et al., 2005).

The ensemble of antibiotic resistance genes is the resistome (D’Costa et al., 2006; Fajardo et al., 2008; Wright, 2010). The term can be applied to the resistance genes of a given bacterial population, a species or any higher taxa, or to the whole microbiota. However, the estimated size of the resistome is highly dependent on the definition of the resistance gene (Martínez et al., 2015). A mutation in a chromosomal gene might result in a resistance phenotype, but this mutated gene is rarely transferred to other bacterial species. In fact, the majority of the resistance genes detected in metagenomes are permanently associated with the same microorganisms (Fondi et al., 2016); i.e., they are intrinsic resistance genes (Olivares-Pacheco et al., 2013; Forsberg et al., 2014; Ruppé et al., 2019). However, most of antibiotic resistance genes of importance in public health are located into MGEs (Martínez et al., 2015). Any type of genetic interaction based on horizontal gene transfer favors the spread of antibiotic resistance (Huddleston, 2014). High-risk, transtaxa antibiotic resistance genes are prone to horizontal gene transfer by being included in structures such as plasmids, integrative and conjugative elements, conjugative islands, phages, and phage-like elements. To calculate the real proportion of antibiotic resistance genes among accessory genes transmitted by MGEs is presently a difficult task, given these elements are over-represented in the available databases, which are enriched with antibiotic-resistant organisms.

The spread of accessory genes, antibiotic resistance genes being a fraction of these, occurs asymmetrically between bacterial species (Hu et al., 2016). The antibiotic resistance gene flow between species can be envisioned as interbacterial roads and highways, which are used by the mobile elements serving as “vessels of the communal gene pool” (Beiko et al., 2005; Norman et al., 2009). Note that antibiotic resistance function (phenotype) might depend on the horizontal co-transfer of neighbor (clustered) non-resistance genes, when are part of an operon. The operon organization is beneficial as enables the transfer of functionally coupled genes (Lawrence, 1999). Knowing the roads and highways by which the accessory genome flows should help us predict the itineraries that will be used by antibiotic resistance genes. In other words, antibiotic resistance genes circulate in the same channels as the accessory genome, comprising most genes involved in cell-environment adaptive interactions (Paquola et al., 2018).

The accessory gene flow among Gammaproteobacteria has been represented as a bipartite network, where the edges (links) connect two independent sets of entities, in our case bacterial genomes and antibiotic resistance proteins (genes) (Lanza et al., 2018; Figure 1). The distance between two bacterial species is proportional to the number of connections, that is, the number of shared proteins. This representation was based on the study of 21 Gammaproteobacterial species, represented by 47,885 genomes, analyzed using the Porous material Analysis Toolbox⁴ platform, based on AcCNET software (Lanza et al., 2017). As mentioned above, the bipartite network includes nodes belonging to different categories, in our case genomes and proteins (gene sequences were translated into proteins), so that each genome has links with their corresponding proteins. A statistical module allows inferring both genome clustering and protein clustering. Genome clustering arranges the genomes into groups (units) that share specific antibiotic resistance proteins. Protein clustering illustrates the possibility of the co-occurrence of specific proteins that are found in the same group of genomes.

This representation suggests that the accessory genome gene flow circulates in gamma proteobacteria favored by a phylogenetic neighborhood. In Enterobacterales, the flow is preferential between a “flow core” constituted by Escherichia, Klebsiella, Salmonella, Citrobacter, and Enterobacter, linked with the closer genera Serratia and Yersinia. Frequently transited pipelines linking this flow core to other distant species occur between Vibrio and Salmonella; Escherichia-Salmonella and Pasteurella and Haemophilus; Klebsiella-Serratia with Acinetobacter and Pseudomonas. However, many links occur outside these high roads, including a few reaching the far-located Legionella. With the precaution of considering the biased composition of available genetic databases, these roads correspond well with the history of recent antibiotic resistance events.

These gene flow highways are highly consistent with the genome-based phylogeny of the bacterial organisms. Seven phylogenetic groups or clades have recently been proposed in Enterobacterales (Adeolu et al., 2016). The first, the Escherichia-Enterobacter clade, comprises Escherichia, Klebsiella, Enterobacter, Raoultella, Kluyvera, Citrobacter, Salmonella, Leclercia, and Cronobacter, and corresponds to the organisms more involved in gene flow between human and animal microbiomes (Hu et al., 2016, 2017). Other clades, such as Erwinia-Pantoea, Pectobacterium-Dickeya, Serratia-Yersinia, Hafnia-Edwardsiella, Proteus-Xenorhabdus, and Budvicia can certainly be considered as candidates in transenvironmental and transmicrobiome genetic transfer of antibiotic resistance genes to the species of the Escherichia-Enterobacter clade.