This exciting new preprint touches on a topic close to my heart: gene flow in gut communities. What’s even better, this paper shows how rich and complex the evolutionary dynamics are, and how much more remains to be discovered and understood.
Garud et al. (2017) examine evolutionary dynamics within different people’s gut microbiomes (the gut microbiome being the bacteria living in your intestines that help digest food, keep out pathogens, and properly regulate your immune system). The most abundant bacteria in the human gut belong to two phyla: Bacteroides and Firmicutes. This study focuses on Bacteroides (especially Bacteroides vulgatus) for the simple reason that abundant bacteria = good data. The data are good enough that Garud et al. can infer how haplotypes (chunks of linked genetic variants) change in frequency over time for several different species of bacteria in the intestines of 365 humans. Garud et al. do so by arguing that alleles that are found in > 80% of sequences from a particular species probably occur together. These are what Garud et al. call ‘confidently phaseable’ haplotypes, in other words, groups of mutations that move together in lock-step over time because they are found in the same genome.
One reason why I’m excited about this paper is that it shows that many selective sweeps — haplotypes that increase in frequency in the population, eventually establishing in the population— are driven by recombination with other strains and species in the gut, rather than de novo mutations. Several studies have pointed to the importance of recombination as a driving factor in bacterial evolution (Shapiro et al. (2012) and Denef et al. (2012), for example,). So-called ‘soft sweeps;’ that is, adaptation stemming from existing genetic variation in the population have also been shown to be more important than ‘hard sweeps’ stemming from de novo mutations in natural populations as well as in evolution experiments (Kosheleva and Desai (2017) and Garud et al. (2015) and (2016)). There are a couple good reasons why: first, existing genetic variation has been ‘field-tested’ by natural selection, and second, even highly beneficial de novo mutations have a high probability of going extinct through bad luck.
Garud et al. also report something quite puzzling: anomalously closely related strains found in different hosts found as far as different continents, compared to average divergence between strains from different hosts. This pattern isn’t universal across genera, which makes it more striking. What’s going on here? The authors don’t know yet and neither do I; perhaps some interesting new biology!
I was also gratified to see that the work done by Garud et al. fits a ‘recombination wind’ model of gene flow in the microbiome that I proposed in previous work. In that paper, I described a model of E. coli genome evolution in which genes are differentially affected by a ‘wind’ of mutations blowing in from other strains and species, due to the differing strength of purifying selection over the genome. I used this model to give one explanation for why the molecular clock seems to tick at different rates in different E. coli genes. My basic idea was that E. coli genomes sampled from different microbiomes would have the imprint of evolution related to horizontal gene transfer within their microbiomes, which would be a ‘hidden variable’ without more samples from those microbiomes. Garud et al. reveal the action of this previously ‘hidden’ variable; that said, Garud et al. do not provide evidence one way or another on whether gene flow is related (or not) to the strength of purifying selection. I’m quite interested to see what future research on the evolutionary dynamics in microbiomes will find.
In a previous post, I described two lines of evidence showing that my model is not necessary to explain apparent divergence times (measured by synonymous mutations) at different loci. First, precise experimental measurements of the mutation rate in the E. coli genome shows that the rates actually vary across genes. Second, the density of synonymous mutations varies in between the chromosome and plasmid of Buchnera endosymbionts that do not seem to be significantly affected by horizontal gene transfer (although note that there is evidence of pervasive recombination in Wolbachia endosymbionts) . So, I concluded that my model must be wrong. Now, I think I jumped the gun in rejecting my model; I think there’s a reasonable chance that it might be verified by future work. As I’ve argued before, even with genome-wide variation in the mutation rate, the combined effects of gene flow and transient hypermutability should dominate subtle genome-wide mutation rate differences—at least for microbes (such as E. coli) where both are important.