Host-bacterial coevolution and the search for new drug targets

Abstract

Understanding coevolution between humans and our microbial symbionts and pathogens requires complementary approaches, ranging from community analysis to in-depth analysis of individual genomes. Here we review the evidence for coevolution between symbionts and their hosts, the role of horizontal gene transfer in coevolution, and genomic and metagenomic approaches to identifying drug targets. Recent studies have shown that our symbiotic microbes confer many metabolic capabilities that our mammalian genomes lack, and that targeting mechanisms of horizontal gene transfer is a promising new direction for drug discovery. Gnotobiotic (“germ-free”) mice are an especially exciting new tool for unraveling the function of microbes, whether individually or in the context of complex communities.

Introduction

The ~10¹⁴ microbes that live in and on each of our bodies belong to all three domains of life on earth — bacteria, archaea and eukarya. They outnumber our own cells by a factor of 10, and contribute many physiological capabilities, including metabolism of glycans and amino acids, synthesis of vitamins and isoprenoids, and biotransformation of xenobiotics[1••]. A deeper understanding of our human biology thus requires understanding of our microbial communities and the genes that they harbor (‘our’ microbiome) (Fig. 1). The notion that we have a ‘meta-genome’ composed of microbial and human components, and a ‘meta-metabolome’ that reflects metabolic activities carried out by both our microbial and our H. sapiens cells, has implications for the definition of health, discernment of disease susceptibilities, and diagnosis of human pathologies. This view of ourselves also opens up another dimension to therapeutics, including treatment strategies that accommodate microbial metabolism of drugs that target our human cells, and a new generation of therapeutics that affect the structure and function of our indigenous microbial communities. The vast majority of our microbes live in the gut. Thus, the current challenge is to understand the extent to which each individual’s gut microbiota affects the bioavailability and host/microbial responses to orally or parenterally administered drugs, and the impact of interventions that alter our microbial ecology.

Fig. 1. Understanding the microbial part of ourselves.

A key part of understanding human metabolic capabilities is to understand our microbial symbionts, and the genomes of those symbionts.

The concepts of coevolution and co-differentiation

One key step in understanding our microbiota is to identify lineages that have coevolved with humans (or with mammals in general), and to identify the genomic consequences of this coevolution. Coevolution between a host and a beneficial symbiont, or a pathogen, is defined as reciprocal adaptation of each lineage in response to the other[2••]. For example, genetic changes that increase production of a metabolite by an intestinal bacterium may trigger selection of changes in the host genome that promote uptake or prevent synthesis of that metabolite. Coevolution can also result in co-differentiation. Co-differentiation is defined as the diversification of host and symbiont lineages in parallel through a history of constant association; however, coevolution and co-differentiation can occur independently of one another [2••]. Co-differentiation can be detected by showing that host and symbiont phylogenies match (see [3•,4•] for detailed reviews of methods). Some methods for detecting co-differentiation can also generate hypotheses about processes causing specific differences between host and symbiont phylogenies. These differences (Fig. 2) include (i) the absence of a symbiont in a host lineage, due to extinction of the microbial species in the microbiota that occupies a given body habitat in the host, or due to under-sampling; (ii) speciation of a symbiont within the same host, so that the host contains two closely related species of the symbiont; (iii) failure of the symbiont to speciate when the host speciates, so that two closely related species of the host contain the same symbiont; and (iv) host-switching, i.e. transfer of a symbiont to a different host[3]. For example, in the case of (iv), unrelated mammal species that independently acquired similar diets may support the same symbiont species, which evolved in one and then found conditions in the other to be suitable.

Fig 2. Processes affecting host-microbe coevolution.

Large gray tubes indicate relationships between host lineages, while thin colored lines indicate relationships between microbial lineages (top tree). Importantly, divergence between bacterial divisions (black lines) occurred over a much longer time-scale than divergence between metazoan host lineages such as, for example, humans, chimpanzees and bonobos (see scale). Processes depicted include co-divergence of host and microbial lineages (A), diversification of a microbial lineage while associated with a host (B), extinction (C), host-swapping (D) and association of a free-living microbial lineage with a host (E). Co-divergence, adaptation to a novel host, host-swapping or diversification within a lineage all produce splits on the microbial phylogenetic tree (bottom tree). Although great caution and deep sampling are required, studies of the distribution of microbial lineages among hosts (numbers at tips of bottom tree), and comparison of the microbial phylogeny to the host phylogeny, may help to resolve which evolutionary processes are responsible for observed divergence between host-associated microbial lineages.

An extension of the concept of co-differentiation between host and symbiont is co-differentiation of an entire microbial community with a host animal lineage. Here, an entire microbial community would be passed vertically from host to offspring. Over the course of speciation events in the host lineage, the microbial communities would differentiate in a way that would mirror host phylogeny. Such a scenario would be expected in host lineages where parents inoculate their offspring with a microbial consortium that is highly adapted to a specialized diet. An example of such hosts is the Koala bear: mothers inoculate their young with “pap”, a specialized dropping that allows the young to make the transition from milk to a folivorous diet of Eucalyptus leaves and branches[5].

Coevolution has been invoked to describe the relationship between mammals and their gut microbial communities, because communities differ between species (mouse, cow, pig and human[6]). However, these differences could instead stem from selection of microbial lineages by a host’s diet or immune system. Little is known about variation within each species, so differences between samples could also primarily reflect differences between individuals rather than between species. Similarly, gut microbial communities may be composed of environmental microbes pre-adapted to the chemical milieu of the gut, rather than microbes that have co-evolved with their hosts. Unambiguous demonstration of co-diversification of mammals and their gut microbes has not yet been achieved: for example, patterns of community similarity obtained by comparing the gut microbiotas of a range of mammals (e.g. using distance measurement algorithms such as UniFrac[7]), might mirror the phylogeny of the mammals. Such a test requires a systematic survey of the microbial communities associated with animal hosts representing a range of taxonomic orders and diets.

Identifying genes critical for symbiosis

Functional and comparative genomic analyses of human gut symbionts are revealing genes critical for adaptation to the gut environment, and mechanisms for horizontal transfer of these genes. These studies, along with in-depth analyses of symbionts of invertebrate hosts, provide a necessary framework for designing and interpreting metagenomic studies of the human microbiome.

In contrast to the human gut, which houses a diverse microbial community, many invertebrates (e.g. aphids, sharpshooters, and stinkbugs) have simple communities that are either maternally transmitted directly to the offspring[8,9••] or are eaten as maternally-deposited capsules shortly after hatching[10]. These symbiont genomes have dramatically reduced gene content, but retain genes for key metabolic capabilities that complement host physiology, including vitamin and amino biosynthesis[8,9••]. However, some human gut symbionts, including members of a large division of bacteria known as the Bacteroidetes, have maintained a larger genome size[11•], perhaps because they must survive outside the host to be transmitted.

Studies of mice raised to adulthood in sterile isolators without any exposure to microbes (‘germ-free’ animals) are especially useful complements to genomic approaches for understanding the function of gut microbes. For example, germ-free mice colonized with a prominent human gut symbiont, Bacteroides thetaiotaomicron, demonstrate that this bacterium can selectively induce a set of its genes that degrade otherwise indigestible dietary polysaccharides[12••]. Genomic analysis of Methanobrevibacter smithii showed that this species likely promotes energy harvest in hosts by consuming a range of fermentation products of other gut bacteria, and may be a good target for anti-obesity drugs[13]. Comparative genomic analyses of gut and non-gut Bacteroidetes[11•] revealed that gut Bacteroidetes possess large arsenals of genes that sense the nutrient environment. These nutrient sensors are linked to gene clusters encoding proteins involved in acquiring specific classes of glycans, and degrading these glycans by glycoside hydrolases and polysaccharide lyases. The products of these polysaccharide utilization gene clusters are used by other members of the microbiota that are ill-equipped to degrade complex glycans, but are well-endowed with genes involved in importing monosaccharides and converting them to fermentation products that can be utilized by other components of the microbiota, and the host[14•]. These types of studies can be expanded to model communities of sequenced gut symbionts that are introduced into normal or genetically engineered germ-free mice: the effects of diet and or drugs can be carefully monitored in these ‘gnotobiotic’ mouse models under conditions where potentially confounding variables, such as host genotype and diet, can be constrained. Gnotobiotic mouse studies will provide better understanding of the rules and forces that govern the assembly and operations of microbial communities, proof-of-principle experiments that ascertain the contributions of specified groups of microbes to community and host operations, and proof-of-concept tests of the efficacy of new types of anti-microbial drugs that target horizontal gene transfer between members of a microbiota or the activities of virulence factors embedded in a microbiome (see below).

A preliminary study of microbial gene content in the fecal microbiota, which mirrors the microbiota of the distal gut, of several healthy humans showed that compared to our H. sapiens genome and the genomes of all sequenced microbes, there is an enrichment of the representation of genes involved in vitamin biosynthesis, degradation of diet- and host-derived polysaccharides as well as xenobiotic metabolism[1••]. Further analysis revealed that the gut microbiome is also enriched in a family of conjugative transposons, consistent with a pronounced role of horizontal gene transfer (HGT) in shaping gut microbial genomes[15].

Horizontal gene transfer and coevolution

HGT is an important factor in the evolution of microbial communities that promotes adaptation to novel or changing environments, including mammalian host environments. HGT is of intense medical interest, not only because of its contribution to the spread of antibiotic resistance genes, but also because it can cause closely related strains to differ drastically in clinical parameters. For example, type III secreted effectors may cause differences in host specificity between strains of Salmonella enterica[16]. On a longer time scale, the acquisition of the type III secretory systems encoded by the SPI-1 and SPI-2 pathogenicity islands is a defining feature of host adaptation for S. enterica as a whole[17].

Several novel strategies for drug development are being pursued in response to the challenge posed by the horizontal transfer of genes involved in both antibiotic resistance and virulence. These strategies include the development of compounds that directly inhibit gene transfer[18••] or virulence[19].

Targeting virulence factors with small-molecule inhibitors directly presents several potential advantages. Such targeting may cause less collateral damage to the indigenous microbiome than traditional antibiotics, may exert less selective pressure for the evolution and transfer of resistance, and may be effective against divergent organisms that have acquired a particular virulence factor by HGT.

Genomic islands contain a rich source of genes of unknown function[20•] that may yield novel virulence factors appropriate for small-molecule inhibitors. For example, a recent screen for Salmonella enterica serovar Typhimurium genes involved in survival and replication within macrophages (a key feature of persistent infection by Salmonella) found that such genes were dramatically overrepresented within putatively transferred regions, such as prophages and pathogenicity islands[21].

Gene transfer systems themselves are also being targeted with small-molecule inhibitors. Such inhibitors could be co-administered with antibiotics to prevent the in-vivo acquisition of resistance factors by susceptible pathogens during the course of antibiotic therapy. For example, the bisphosphate-containing compounds clodronate and etidronate inhibit the F plasmid TraI relaxase in vitro and conjugative transfer of F plasmid in vivo. These findings are particularly significant because relaxases are essential components of conjugative transfer systems, and the F plasmid TraI relaxase is closely related (~99% sequence identity) to the relaxases of many plasmids known to transfer antibiotic resistance genes[18••].

The promise of metagenomic approaches

The vast majority of phylogenetic diversity in microbial communities associated with the human body (and other ecosystems) is represented by organisms that are difficult or impossible to culture in the laboratory using currently available methods. Standard culture methods are especially problematic for understanding symbiosis, because microbes may express completely different sets of genes or may not grow at all outside the host. Metagenomics allows us to observe the genes contained in this vast uncultured majority through the isolation and sequencing of DNA directly from the community. Typically, 16S rRNA gene sequences are used as a phylogenetic marker to probe community structure and diversity (‘who’s there, and in what abundance?’). The rest of the genes in the microbiome are characterized through shotgun sequencing of whole microbial community DNA. Although the short 200–250 nucleotide reads currently obtained by the latest generation of massively parallel DNA sequencers (i.e. pyrosequencers) are sufficient for characterizing communities based on their 16S rRNA gene content[22], characterizing other genes typically requires a combination of short pyrosequencing reads, longer Sanger-sequencing reads, and, ideally, complete reference genomes (i.e. genomes of cultured representatives of major phylogenetic lineages present in the community) (Fig. 3). In simple communities, such as those present in certain invertebrates or in the environment, metagenomic data can be sufficient for assembling the genomes of their constituent microbes. Nonetheless, assembly of sequenced genomic DNA fragments from a microbiome into relatively large contiguous sequences of physically linked genes derived from a given organismal genome remains challenging, especially in complex communities such as the human gut microbiota [1••,15,23•]. Despite these challenges, metagenomic studies have revealed specific genes that are enriched in the gut microbiome of humans (see above), as well as microbial genes whose representation is enriched in mouse models of human diseases, including obesity[23•].

Fig. 3. A comprehensive understanding of our microbial ecology requires integration of many data sources.

Community profiles with 16S rRNA indicate which types of organisms are present, metagenomic profiling allows us to identify specific functional categories of genes that are critical for differences in function, and complete genomes act as scaffolds for understanding changes in gene content through loss, amplification, and HGT that allow microbes to adapt to functional roles in different environments.

Because assembly is difficult, metagenomic analysis of complex communities is currently ‘gene-centric’: DNA sequences are mapped to known genes and genomes to infer the relative abundance of different genes and metabolic pathways[23•,24••]. One major challenge in metagenomics is to link these ‘gene centric’ functional predictions to the organism that contained each gene. This goal is complicated by a lack of complete reference genomes and complete/consensus reference taxonomies, and by the short DNA fragment generated by pyrosequencing. Functional assignments typically rely on homology searches. Taxonomic assignment is more complex, and usually involves aligning homologs, and building phylogenetic trees. Annotations can then be assigned based on the best matches or closest homologs in a reference set of genes. Thus, taxonomic assignment is challenging and computationally expensive even with full-length sequences, and results are strongly affected by the reference set.

Another major challenge is that individual labs usually do not have the capability to completely characterize a complex community through metagenomics: many key analysis tools do not scale to large datasets, and the cost of sequencing and the computational resources can be formidable. The most successful studies have come from collaborations spanning a range of disciplinary expertise, including physiology, microbiology, molecular biology, statistics, molecular evolution, ecology and high-performance computing. Collaboration and coordination at different project scales are clearly needed[14•].

Conclusions

Coevolution between mammals and their microbial symbionts is probably more complex than co-divergence, and HGT appears to play a key role. 16S rRNA and metagenomic profiling of mammal-associated microbial communities associated with health and disease, in combination with carefully chosen complete genomes, will likely reveal new drug targets for the manipulation of microbial community structure and function. This is one of the hoped-for outcomes of the international human microbiome project, which is now being launched in several countries [14•].