Microbial Speciation

Abstract

What are species? How do they arise? These questions are not easy to answer and have been particularly controversial in microbiology. Yet, for those microbiologists studying environmental questions or dealing with clinical issues, the ability to name and recognize species, widely considered the fundamental units of ecology, can be practically useful. On a more fundamental level, the speciation problem, the focus here, is more mechanistic and conceptual. What is the origin of microbial species, and what evolutionary and ecological mechanisms keep them separate once they begin to diverge? To what extent are these mechanisms universal across diverse types of microbes, and more broadly across the entire the tree of life? Here, we propose that microbial speciation must be viewed in light of gene flow, which defines units of genetic similarity, and of natural selection, which defines units of phenotype and ecological function. We discuss to what extent ecological and genetic units overlap to form cohesive populations in the wild, based on recent evolutionary modeling and population genomics studies. These studies suggest a continuous "speciation spectrum," which microbial populations traverse in different ways depending on their balance of gene flow and natural selection.

Figure 1.

A model for bacterial speciation under different recombination/selection balances. Five stages in the process of speciation are illustrated under the r/s ≫1 regime (top). The illustration shows an ancestral population of bacteria (purple), which diverges into two incipient species (red or green) adapted to different habitats following the acquisition of niche-specifying genes (red or green arrows). Thin gray or black arrows show recombination events within and between incipient species, respectively. In the r/s ≪1 regime (bottom, not illustrated), stages 2–4 differ, but the start and end points are the same. The r/s ≪1 regime corresponds closely to the stable ecotype model (s.e.m.) (Box 3). Selection (s) is defined here as the average fitness difference experienced by a niche-specifying (adaptive) allele in different niches and recombination (r) is the recombination rate per locus per generation. The stages represent rough, potentially overlapping, and potentially terminal steps (e.g., stage 2 need not lead to stage 3). HGT, Horizontal gene transfer. (From Shapiro et al. 2012; adapted, with permission, from the author.)

Figure 2.

Islands and continents of speciation. Based on data from aligned core genomes of Vibrio cyclitrophicus (Shapiro et al. 2012) (upper panel) and Sulfolobus islandicus (Cadillo-Quiroz et al. 2012) (lower panel), the distribution of divergent single nucleotide polymorphisms (SNPs) (divSNPs) fixed between populations is plotted along the genome as black bars. Vibrio and Sulfolobus contain, respectively, 725 divSNPs distributed more than ∼1% of the genome, and 4232 divSNPs distributed more than ∼36% of the genome. Scale is approximate for divSNPs (y-axis) and genome position (x-axis). The y-axis is not to scale for SNPs rejecting differentiation between populations (regions of the genome shown in gray).

Figure 3.

Ecological differentiation via a gene-specific sweep. This illustration follows the basic steps of the symsim model (Friedman et al. 2013). At the first time point (t₁), a niche-specifying gene (red triangle) arrives into a homogeneously recombining population occupying a single niche. Between t₁ and t₂ (as between all time points), recombination (r) occurs at random from a sympatric pool (one to two events per genome, illustrated as arrows), then genomes reproduce clonally, and are culled to a carrying capacity of four genomes per niche. Because the niche-specifying gene confers a selective advantage(s) in the new niche, genomes that contain it grow exponentially until the carrying capacity is reached at t₃. Other genomes are culled at random, because the rest of the gene pool is neutral to fitness. By t₄, the gene-specific sweep is complete. The niche-specifying gene is in perfect association with the new niche, but all other genes are randomly distributed across niches. At this point, barriers to recombination between niches (dashed line) may or may not emerge. (Note that recombination events at t₄ are not shown for purposes of clarity, but this does not mean they do not occur.)

Figure 4.

Barriers to recombination emerge via a competition-dispersal tradeoff. (A) The particle-associated Vibrio population (L) attaches to nutrient-rich particles and forms biofilms, whereas the free-living Vibrio population (S) hovers near the surface and scavenges loose nutrients. (B) When a new particle becomes available, only the S population is able to rapidly disperse to the new nutrient source. (From Yawata et al. 2014; adapted, with permission, from the author.)

Figure 5.

Islands of speciation are distinct from genomic islands. (A) Different colors denote different allelic variants at a chromosomal locus. (B) Divergence between species (left y-axis) is often measured as the number of fixed nucleotide substitutions between species, or a measure, such as the fixation index (F_ST); metagenomic coverage (right y-axis) is simply the number of metagenomic sequencing reads that align at a given position of the reference genome.

Figure 6.

Evolutionary convergence as the basis for genome-wide association studies (GWAS). In this simplified example, the genome-wide core phylogeny has been inferred for a sample of mostly clonal bacterial isolates (A–L). A convergent (homoplasic) single nucleotide polymorphism (SNP) is identified in four genomes. Importantly, this corresponds to only two independent mutation events (T → A, indicated in the phylogenetic tree by the red “X” with an “A” above it), which associate perfectly with two independent transitions from the antibiotic sensitive to resistant (R) state. The significance of the association can be assessed by calculating a P-value by resampling from the genome-wide distribution of mutations and phenotypic states (resistant or sensitive) on the phylogeny. By failing to account for population structure (e.g., the phylogenetic information), four events would be counted, thereby overestimating the significance of the association. GWAS can also be performed considering entire genes, instead of individual nucleotide sites, as targets of convergent mutations (for examples, see Sheppard et al. 2013 and Farhat et al. 2013, 2014).