Satellite Species

by Bill Hanage
These days I tend to notice when one of our papers has gone live through the twitter account of Cameron Thrash (@jcamthrash), who keeps a close eye on the literature from his perch at LSU and helpfully spreads the word.

This morning he tweeted this:

This is work with the Marvelous Pekka Marttinen, now a University Lecturer at the Helsinki Institute for Information Technology, but a former postdoc and long time collaborator. It builds on some work I did years ago with Christophe Fraser and asks how the clusters of more or less closely related things that we call ‘species’ persist in recombining bacteria. Here specifically we’re looking at two closely neighboring clusters. One of them is a famous and well-characterized pathogen, the pneumococcus. The other is not. It is a cluster of things usually called ‘atypical pneumococci’, but which we have previously shown is distinct from common or garden pneumococci both in their core and accessory genomes. This sort of thing was only detectable through the genomics, because if you look at only a few housekeeping genes, the two clusters are really not that different. Atypical pneumococci are different from the regular sort in several ways, notably they lack a polysaccharide capsule. These capsules, which are near ubiquitous in regular pneumococci, can be divided into more than 90 different serotypes and are a major virulence factor and target for current vaccines. Atypical pneumococci cannot be serotyped, and it’s not that the serotypes are different, or the genes are down regulated, the capsule genes are just not there.

The absence of capsule genes hasn’t stopped these bugs being successful; they’ve spread widely, and in some cases colonize a lot of the population. They’ve also been found to cause outbreaks of conjunctivitis, which is a distinct disease tropism. Altogether my opinion is that they’re not atypical pneumococci. They’re a new as-yet-unnamed species.

That said there are some things about them that are very odd. Notably, and like the pneumococcus, these things recombine. That means that genetic material from one lineage can be readily transferred to another. But if they recombine, how do the clusters remain distinct? Why don’t they merge together as their gene pools get mixed up, and form just one cluster? Christophe showed that this could happen pretty easily with the sorts of rates we are talking about.

The new paper tries to look at this using a model that incorporates ecology. Simplifying the complicated reality a LOT we assume that, for recombination to occur between two things, they have to be in the same part of ecological space. Very little gene flow is expected between things adapted to a rain forest, and a coral reef. But what about the more subtle distinctions that obtain in the place where these bacteria normally reside, which is the human nasopharynx?

Figure 2 from the paper Figure 2 from the paper. Here we are thinking of niche space in the sense of Hutchinson, ie as a notional n-dimensional hypervolume in resource space where things can survive. B shows a schematic of the model with ‘species’ A and B in their own niches a and b, and an area of overlap ab. A cannot survive outside of a+ab, and vice versa.

We found (or tbh, Pekka found), that given the amount of recombination we would expect all other things being equal, the ‘niches’ of the pneumococcus and ‘atypical’ pneumococci overlapped by about 40%. In other words, they encountered each other and were able to recombine about 40% of the amount of time that would have been expected were they ecologically identical. We did this using a model, in which individuals could migrate into and out of the area of niche overlap and recombine there. I should note that this model is quite different from some other ones in population genetics. The way to see this is to imagine what would happen if you set migration to zero, and ran the model forward. In the version we use here (see above), three populations would eventually emerge, one each from the two ‘private’ portions of niche space and one from the overlapping region. Another way to do it would be to instead combine recombination and migration, and have the two niches completely different with absolutely no overlap. Occasionally one ‘species’ would wander into the ‘wrong’ niche and survive long enough to recombine (ie the product of recombination and migration). In that case if you set migration to zero, you would get two distinct clusters, one from each niche and none from the overlap, because it doesn’t exist. Our version notably does not have explicit selection, but that is included with the niches. We like the model and think it is very flexible, but one reviewer disagreed, urging us to simplify it further. They were however extremely gracious, in noting that while they stood by their objection, this was not enough to prevent publication.

The other interesting thing was that when we looked at the ‘ecoSNPs’ which are those SNPs that are ‘private’ to one cluster or the other, their distribution was consistent with the two clusters not moving rapidly apart, and instead being roughly static in sequence space. This need not have been the case. If recombination dropped enough between two clusters, or niches became so distinct that they were unable to recombine, they are expected to rapidly diverge as they accumulate these ‘private’ mutations. In other words recombination is acting rather like a gravitational force here, preventing the clusters from separating. And when we think about these two ‘species’ they are like satellites of each other, rather than like the voyager spacecraft destined for a lonely future off beyond the known solar system. I immediately coined the catchy term ‘satellite species’ for these. Doubt it will catch on.

Satellite species are not, however, necessary outcomes in recombining bacteria. Sam Sheppard kindly directed us to a similar dataset of campylobacter, in which the story appears to be much more complicated (if you want to know how, I suggest you read the whole paper!)

Some important things that this paper does NOT do: as I have intimated, the model is a bit peculiar. I’m easy with that, but others might not be. One of the most important things about it is that it explains what we would see now that these ‘species’ are now things each with a distinct ecological ‘address’, determined by a multitude of things in their core and accessory genomes (and, likely, with a load of hitchhiking variation too). But how did this come to be? What are the steps by which these are accumulated, when presumably just one niche-defining gene would move rapidly between two clusters in the early stages of separation? While I am pretty sure that this shows that at least one example of ‘satellite species’ exists, the details of the origin of satellite species is a topic for further work.

Written on July 5, 2017