Opinion by Florent Lassalle
In a recent article titled “Improved genome of Agrobacterium radiobacter type strain provides new taxonomic insight into Agrobacterium genomospecies 4” (PeerJ 7:e6366), Gan et al.(2019) propose to create two subspecies within A. radiobacter ( = Agrobacterium genomospecies 4) to resolve the apparent conundrum of the type strain for A. tumefaciens (B6T) and A. radiobacter (NCPPB 3001T) occurring within the same bona fide species (Agrobacterium genomospecies 4) by proposing two new subspecies, of which each strain would be th type. Due to taxonomic priority of A. radiobacter, they rightly recommend that the whole species be called A. radiobacter (in accordance with the statement from the Judicial Commission of the ICSP, [Tindall 2014]), and that its subspecies be called A. radiobacter subsp. tumefaciens and A. radiobacter subsp. radiobacter.
I recognize the great effort of Gan et al. to describe the fine diversity within A. radiobacter, and the advanced knowledge provided in terms of prediction of the differentiated metabolic capabilities of the various strain clusters. However, I would not claim these to be subspecies.
Proposal of subspecies should be supported by sufficient population genomics evidence.
In principle, there is nothing prohibiting the proposal of subspecies for bacterial groups, in order to reflect significant separation of ecological or sexual populations within a named species. The process of bacterial diversification is mostly clonal and generates a fractal pattern of genetic diversity over long time scales, which naturally lends itself to hierarchical taxonomic classification. However, bacterial diversification also involves significant gene flow through horizontal gene transfer, with no strict barriers between lineages (species or any other biological unit), which blurs – or even erases – this fractal structure, especially at short evolutionary time scales. To investigate within-species population structure, one must thus conduct a thorough assessment of the differential gene flow between populations and of the occurrence of fixed genomic traits in each population. Nowadays, such population genomics analyses are feasible thanks to affordable genome sequencing techniques and increasingly standard bioinformatics methods. A good example is the study by Cavassim et al. (pre-print) delineating 5 genospecies within R. leguminosarum – in that particular case the fate of these genospecies might be to be renamed as bona fide species under the umbrella of the Rhizobium leguminosarum species complex, but in principle, observing similar structure at a lower divergence level could warrant the delineation of a new set of subspecies.
In the case of the study by Gan et al.(2019), it seems that an attempt to clarify the taxonomy by providing names than referred to previously equivalent taxa took precedence over the considerations discussed above. In any case, I do not think enough data is provided to justify the delineation of subspecies. Specifically, not enough strains are compared: only 11, which is unlikely to cover the extant diversity within A. radiobacter ( = Agrobacterium genomospecies 4) and thus to provide a robust evidence for genomic differentiation within this species. In fact, within each of the two proposed subspecies Agrobacterium radiobacter subsp. tumefaciens and Agrobacterium radiobacter subsp. radiobacter, each covering three strains, two strains are duplicates (derivatives of the same original culture with 100% whole-genome ANI): B6 and B6*, and LMG140 and NCPPB 3001, respectively. This leaves only one pair of non-redundant strains within each proposed subspecies, i.e. B6 and B140/95, and CFBP 5621 and NCPPB 3001. This seems insufficient to prove that these lineages have their own genomic coherence and are well separated from other clusters of diversity within A. radiobacter. Notably, hybrids of these strain lineages with other lineages within A. radiobacter could occur, and only dense sampling of the natural diversity of A. radiobacter could provide an idea of their frequency. Only with the demonstration of the absence – or exceptional rarity – of hybrids within a large sample could in my opinion make these lineages fit to be given a taxonomic label.
That said, this study carries a valuable in-depth comparative genomics investigation of the diversity of A. radiobacter, and presents very interesting facts. It highlights differences of gene content with predicted impact on the metabolism of nucleotide sugars and expression of key phenotypic traits such as the surface lipopolysaccharide (with possible impact on host range). These findings suggest a potential ecological differentiation of these lineages, which could find its counterpart in terms of sexual isolation – or only reflect the specific characters of two strain lineages among a spectrum of many occurring within this species. Further exploration of the genomic diversity and adaptive properties of members of this species are therefore encouraged at a greater scale, as they could actually reveal the occurrence of differentiated populations within A. radiobacter.
Agrobacterium tumefaciens: a confusing name referring to both a species and a species complex
Reading this article leads me to raise another issue about the choice of taxon names. This problem is not the fruit of the work of Gan et al., but it seems to me that this publication might reinforce a pre-existing confusion in the use of taxon names in the genus Agrobacterium.
Gan et al. specifically recommend that “Agrobacterium tumefaciens to be reclassified as Agrobacterium radiobacter subsp. tumefaciens”. While this could be taxonomically correct – setting aside the concern of lack of sufficient evidence for delineation of these subpsecies – this statement may come as a source of confusion. Indeed, A. tumefaciens (valid name of a species and later heterotypic synonym of A. radiobacter) could be confused with the “Agrobacterium tumefaciens complex”, a species complex which includes several genomospecies (numbered 1 to 20, [Mafakheri et al. 2019], some of which having recently been named [de Lajudie and Young,2018]). The taxon “Agrobacterium tumefaciens complex”, also known as Agrobacterium biovar 1, is rightly placed in the NCBI Taxonomy tree (TaxonomyID: 1183400) above the species level, and includes A. radiobacter (TaxonomyID: 358). The “Agrobacterium tumefaciens complex” is an evolutionary clade that is well distinct from other groups within the Agrobacterium genus, such as the clade grouping A. larrymoorei and A. bohemicum (Lassalle et al., 2011 & 2019), and should thus be preserved as a practical taxonomic unit.
It is thus important that readers of the article of Gan et al. have in mind that “Agrobacterium tumefaciens complex” and A. tumefaciens are two different entities, and avoid repeating. Indeed, there has already been regular confusion in the past between these them. As a result, many new strains appearing to belong to the “Agrobacterium tumefaciens complex” but not to a specific species under it (for instance due to lack of resolution of the sequenced phylogenetic marker) that should have been named Agrobacterium sp. were instead named A. tumefaciens or one of its synonyms: A . radiobacter, Rhizobium radiobacter or R. tumefaciens (the latter two emerging from the long controversy on the polyphyletic status of the genus Agrobacterium, now resolved [Mousavi et al., 2014]), even though these names all points to a specific species within this species complex (i.e. Agrobacterium genomospecies 4).
Such confusions created an apparent anomaly in the diversity landscape of Agrobacterium, with A. radiobacter (or its synonyms) grouping strains from the whole species complex; this became clear in recent phylogeneomic studies, with A. radiobacter appearing as a polyphyletic group in trees. This problem is not due to any complicated evolutionary phenomenon, but rather to a propagation of mistaken identity: an original mis-naming of Agrobacterium sp. strains as A. tumefaciens, and its occasional (often automated) translation into A . radiobacter.
One solution to avoid such confusion would be to disuse altogether the name A. tumefaciens in future bacterial taxonomy and diversity surveys. Instead, I would recommend the use of the less confusing names Agrobacterium biovar 1 for the species complex, and A. radiobacter for its component genomospecies 4.
Florent Lassalle
Florent Lassalle
