Somewhere between several million (Schloss et al., 2016) and one trillion (Locey and Lennon, 2016) bacterial species inhabit the earth and most of these are not accessible through culture-based techniques (Handelsman, 2004). As the first aim of evolutionary biology (Table 1) is to document life on earth, microbiome research advances this goal by identifying organisms that are otherwise invisible to science (Wright et al., 2009). This is especially true since novel habitats uncover novel biodiversity (e.g., Goffredi et al., 2008; Goffredi, 2010; Petersen et al., 2010; Hug et al., 2016). For example, hoatzins are the only folivorous, foregut-fermenting bird and their crop and lower esophagus function as an extended fermentation chamber for the leaves in their diet. Hoatzin crop microbiomes contain many novel bacterial strains: one study including six wild individuals found 94% of the bacterial OTUs belonged to unnamed species (Godoy-Vitorino et al., 2008). Another study, focusing on Archaea, found 17 of 24 methanogen OTUs likely represented new species and perhaps three new genera (Wright et al., 2009).

The second goal of evolutionary biology is to understand the lifestyle and history of an organism. Here we want to know what an organism does, where it does it and when it existed (Table 1). Describing traits such as home range, species range, daily and seasonal habits, physiological adaptations, phylogeography, etc., are essential for understanding how an organism functions in the world.

Vultures provide an excellent example of how microbiomes add biological value to our understanding of host traits. Vultures are carrion feeders and this dietary specialization exposes them to many microbes that are known pathogens. Roggenbuck et al. (2014) investigated whether the microbiome enables this lifestyle by providing resistance to pathogens. Black vulture and turkey vulture contain high bacterial diversity on their head skin and extremely low bacterial diversity in their gut. Quite unusually, the bacterial classes Fusobacteria and Clostridia dominate and their relative abundances are similar across species and in a captive vulture as well. Fusobacteria and Clostridia, which are frequently pathogenic to other bird species, are possibly contributing to carrion digestion in the vulture digestive tract and outcompeting other bacteria that may be suitable for the niche. The avian hindgut is a previously undescribed niche for both bacterial taxa. Thus, Fusobacteria and Clostridia may facilitate the vulture lifestyle.

There are many available methods to assess the microbiome as a trait. Phylogenies are one powerful tool that provide evolutionary information about the microbiome. Phylogenies can establish which bacteria are important to host or biogeographic micro-environment and can be used to calculate descriptive (alpha diversity) and comparative (beta diversity) statistics. Phylogenetic distance metrics like UniFrac (Lozupone and Knight, 2005) can determine how similar microbial communities are and permutation tests assess significance. Bacterial phylogenies can indicate how long a microbial taxon has inhabited a particular environment (Moodley et al., 2012; Moeller et al., 2016). Categorical statistical tests are another tool that correlate host traits with microbiomes (Godoy-Vitorino et al., 2012; Hird et al., 2014, 2015). Functional analysis of microbiomes using metagenomics can illuminate what genes and pathways are found in a sample. Microbiome functions may be more conserved than taxonomic composition (Lozupone et al., 2012), since convergence, horizontal gene transfer, and functional redundancy may select for function rather than taxonomy. When coupled with host genomic information, a powerful method to detect concurrent shifts in host and microbial function emerges (Gao et al., 2016).

The third major goal of evolutionary biology moves from describing natural history and phylogeny into elucidating the processes that shaped them. All of the major evolutionary forces—selection, drift, migration, mutation, diversification, extinction, and adaptation—are as fundamental to the microbiome are they are to hosts yet at this point we have more questions than answers (Table 1). The microbiome is under multiple levels of selection (Ley et al., 2006): how strong are the different levels and how do they structure the microbiome? Drift (including founder effects and bottlenecks) may be important during times of illness or early in life, when microbial populations are small. How genetic drift in the host affects the microbiome is unknown. The (relatively) rapid generation time of microbes and the ability of the community to change in response to stimuli may contribute to acclimation in hosts, which in turn could facilitate adaptation of a host to new or changing environments (Alberdi et al., 2016). The microbiome facilitated the mammalian expansion from carnivory to herbivory (Ley et al., 2008) and it contributes to more minor ecological shifts as well (like the ability to consume toxic plants; Kohl et al., 2014, 2016). Host migration and population structure may drive the extinction of distinct microbial taxa (Domínguez-Bello et al., 2008).

Comparing microbiomes in the wild contributes to all three of the major goals of evolutionary biology. Hird et al. (2015) generated a biodiversity catalog for the microbiomes of 59 Neotropical bird species and improved our understanding of both birds and microbiomes in an evolutionary context. First, since Hird et al. (2015) contained the first microbiome information from any of the host species, we described new environments for every microbe identified, thus contributing to Goal 1 of evolutionary biology (although whether the OTUs are residents or transients, living or dead, is unknown). The Hird et al. (2015) data also contribute to Goal 2: the microbiome—as a trait—has been described for the host species and revealed some interesting patterns. For example, Figure 2 shows the distribution of Fusobacteriales across a subset of the birds in Hird et al. (2015) and the distribution of the bacteria gives some information about the hosts. Some birds appear to be better suited for hosting Fusobacteriales—the two samples in dark green belong to two Galbula ruficauda (Galruf71828X072 and Galruf71831X222). They have many more Fusobacteriales than all other birds, including those from the same family (shown in lighter green) and those from the same sampling locality (site G). This supports the hypothesis that G. ruficauda may be selecting for Fusobacteriales in their gut; on a phylogenetic tree, many short branches from a single environment may indicate adaptive radiations (Ley et al., 2006), an intriguing possibility for G. ruficauda. Furthermore, the OTUs found in GALRUF71828X072 and GALRUF71831X222 are abundant and closely related, possibly indicating diversification of Fusobacteriales within the species, G. ruficauda or the individuals themselves.

Study design is of utmost importance for field microbiome studies because increasing sample size at a later date may be difficult to impossible. Additionally, microbiomes can change over time or with season (Bailey et al., 2010; Liang et al., 2015), so securing appropriate sample size during a field trip or field season is best. The research question will largely drive the sampling but the desired molecular data can greatly influence the budget. There is a direct tradeoff between number of samples one can analyze and the sequencing coverage per sample and this equation needs to be carefully considered. Collecting replicate samples is essential (if possible) because the variation within species (or groups) is likely unknown and the results from a single sample may be misleading. Microbiome samples are relatively cheap to collect in both time and money (Figure 3). This is especially true in comparison to the cost of getting to some field locations or effort to get the host organism in hand.

Sample preparation and storage of microbiomes is very important: decomposition begins within minutes of death (Vass, 2001) and triggers a successional change in the microbiome (Metcalf et al., 2016). Bacterial taxa in feces can change after <30 min at room temperature (Gorzelak et al., 2015). Freezing, with or without buffer, is thought of as a “gold standard” for sample preservation (Song et al., 2016). This is plausible for some field trips, if access to a freezer, dry ice or liquid nitrogen is available. However, field conditions often prohibit immediate freezing of the sample. High purity ethanol (95% or higher) can fix the bacteria in a sample and this method has been shown to work especially well with fecal samples; Hale et al. (2015) found that frozen and ethanol stored samples had the most similar microbial communities to fresh fecal samples after 8 weeks of storage. Liquid storage buffers are another field possibility and can preserve both RNA (e.g., RNAlater) and DNA [DNAgard or non-proprietary solutions like DMSO/EDTA/saturated sodium chloride (DESS)]. Consistency may be the most important factor for field storage of wild microbiome samples—several storage methods perform well but comparing across methods may be problematic.

As the costs involved in sequencing will affect sampling design, it is best to have a plan for the molecular work from the beginning of a study. The two main categories of microbial community genetic data are amplicon-based and metagenomics. Amplicon-based studies are those that amplify and sequence a single homologous locus (usually a variable region of 16S rRNA). These studies are an informative and economical first step in describing and analyzing microbiomes. The data provide taxonomic information (using large, publicly available databases) and diversity statistics that can be used to characterize and compare microbial communities. Amplicon-based studies are limited, though, in that they (1) rely on PCR, (2) produce short sequencing fragments, (3) contain only one marker. In many cases, information about the functional capabilities of a community are of interest. Functional information can be estimated from 16S rRNA data with computational approaches (Langille et al., 2013) but how well they perform in novel environments is unknown.

Alternative to amplicon-based methods are shotgun metagenomics methods, which analyze a random subset of genetic material from a sample (Zarraonaindia et al., 2013). One shotgun metagenomics approach is sequence-based metagenomics. Here, random DNA fragments from microbial communities are sequenced and used to infer the genes or metabolic pathways found in a sample based on available databases, e.g., KEGG Orthology (Kanehisa et al., 2004). With sufficient coverage and analytical tools, full or nearly complete genomes can be reconstructed from a metagenome (Tyson et al., 2004; reviewed in Sangwan et al., 2016). Another shotgun approach, functional metagenomics, can be used to screen genomic fragments for specific traits (e.g., antibiotic resistance). Here, fragmented DNA is cloned into a fosmid vector then transfected into a bacterial host (frequently E. coli), where functional experiments can be performed (e.g., Sommer et al., 2009). Metagenomics are generally more expensive than amplicon based studies, as they require greater sequence coverage to describe a sample and construct high quality contigs, and require more powerful computational tools to analyze. Metatranscriptomics is another option that uses the RNA in a sample to determine what genes are actively being transcribed at the time of sampling (e.g., Franzosa et al., 2014). Like metagenomics, this method can be expensive but is informative about the metabolic processes occurring at a given time.

The variable regions of the 16S rRNA molecule are by far the most popular choice for characterization of a microbiome. There are many software packages that are suitable for microbiome sequence analysis, the most popular of which are free and include extensive tutorials [e.g., QIIME (Caporaso et al., 2010), mothur (Schloss et al., 2009)]. Metagenomic and metatranscriptomic analyses are more involved than single locus analyses, requiring contig assembly before annotation and alignment to databases. Oulas et al. (2015) provides a detailed description about the computational steps involved in analyzing metagenomic data. All bioinformatics analyses require computational resources; these can range from relatively cheap, e.g., if one has access to an institutional cluster to extremely expensive e.g., if one needs to purchase hardware or cluster time. Additionally, analysis is frequently the most time-expensive step, which directly translates to dollars in many instances. Many sequencing facilities can add standard bioinformatics analysis to a project for a fee.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.