Population genomics of Rhizobium leguminosarum – 3
Here are some more thoughts on issues raised by our new paper:
Kumar N, Lad G, Giuntini E, Kaye ME, Udomwong P, Shamsani NJ, Young JPW, Bailly X. 2015 Bacterial genospecies that are not ecologically coherent: population genomics of Rhizobium leguminosarum. Open Biol. 5: 140133.
First, I should note that our paper has been covered in an article in the online newspaper, The Speaker.
Phenotype: the problem with polyphasic taxonomy
Ganesh Lad grew our 72 strains on Biolog plates to test their ability to grow on each of 95 substrates. He found that almost every strain had a unique pattern, just as we had found with gene content. As a geneticist, I naturally believe that such phenotypes are determined by genes. We were able to prove that connection for one substrate, gamma-hydroxybutyrate, but it was not easy to map phenotypes to genes more generally. There were no patterns of substrate utilisation that were characteristic of particular genospecies – by contrast, there were some accessory genes that were found only in one of the genospecies, so there might be some genospecies-specific phenotypes, but they were not among the “easy” ones that we looked at.
Many of these “easy” phenotypes are the ones that taxonomists conventionally report when describing a new species. When we looked up the species description for Rhizobium leguminosarum, we found a long list of substrates that this species can, or cannot grow on. However, when we cross-checked this list against our data for the same substrates, we found that the majority of these assertions were actually wrong – they were disobeyed by some of our strains, even though we have good evidence that they do belong to this species, as normally defined. It seems clear that the tables of differences between species in substrate utilisation, beloved of taxonomists, are largely a fiction based on an inadequate sampling of the variation within species. Fifty years ago, phenotypes were all that bacteriologists had to go on when trying to classify bacteria, but nobody should rely on these metabolic phenotypes to identify bacteria these days, when DNA sequence is so much easier and more conclusive. Ernesto Ormeño-Orrillo and Esperanza Martínez-Romero (2013) have already made a similar point, and I agree with them.
Now that we know so much more about bacterial genomes, thanks to thousands of genome sequences, we can propose a bacterial classification that reflects the reality of core and accessory genomes. The sequences of core genes provide a robust phylogeny that allows stable species to be defined and recognised, even if strains differ in phenotype. On the other hand, the possession of important phenotypes, conferred by clusters of accessory genes, should play no part in defining species, but can be recognised as “biovars”. These phenotypes include pathogenesis, metabolic traits, or – in the special case of rhizobia – symbiotic capabilities. They are the “apps” that bacteria download from the microbial internet, and they determine most of the “interesting” things that bacteria do.