Some years ago, it occurred to me and indeed to others that the monophyly of the three groups that constitute the three domains of life might not be as robust as people had said. My PhD thesis was on ribosomal RNA phylogeny of environmental sequences and depending on the dataset I was using, I was not getting a monophyletic archaebacterial group.
Martin Embley was the external examiner for my PhD and he queried this with me (it was 1995) and wondered why I was not getting good support for the archaebacteria and why on occasion in my thesis, I presented a phylogenetic tree where they were not monophyletic. I suggested that the sequences lacked signal, that the phylogenetic methods were immature and were improving all the time and that I was pretty sure they would prove to be monophyletic if this, that and the other.
Today I view these results in a different light. My default position is that the archaebacterial are not a monophyletic taxon. There are many reasons for my change of mind and they are outlined in our new paper in Nature Reviews Microbiology.
In 2007, when Davide Pisani was working with me as a Marie Curie fellow, we noticed that we recovered almost no well-supported phylogenetic trees that displayed the archaebacteria as a monophyletic group when eubacteria and eukaryotes were also in the dataset. There were almost no phylogenetic trees that displayed eubacteria as a monophyletic group, when archaebacteria and eukaryotes were also in the dataset.
It was beginning to dawn on me that the three domains hypothesis was about to fall. There are not three primary monophyletic groups on Earth.
Our methods were supertree methods, so I guess until you do some kind of standard analysis using ribosomal RNA or ribosomal proteins, then people won't believe you.
That paper came out in 2008, when Cox et al. showed that they could not recover a monophyletic archaebacteria using reasonably good method - they could only recover archaebacteria as a monophyletic taxon only when they used quite poor quality methods.
As time has gone on, more and more phylogenetic analyses are revealing that the eukaryote gene phylogenies place the eukaryotes within, not outside the archaebacteria.
The fossil record has prokaryotes at 4 billion years, the first eukaryotes at about 1.7 billion years.
Bioenergetic inferences say that eukaryotes cannot arise by standard autogeneous evolution - only by a "merger".
Protein Interaction data suggest three communities of interacting proteins - the archaebacteria preferentially interact with archaebacteria proteins, eubacteria with eubacteria and eukaryote-specific with eukaryote-specific proteins.
Recently, the "Dark Matter" paper suggested strong support for the three domains of life hypothesis. This paper messed up in so many ways. First of all, they use this Phylosift automatic pipeline and it seems that this pipeline doesn't adequately deal with non-tree-thinking. Therefore, the analysis contained genes of mixed histories - nuclear, mitochondrial and plastid. Therefore, the eukaryote genes of mixed history tended to pull away from the two prokaryotic groups and reinforce the three groups. However, Williams and Embley have shown that this analysis is certifiably wrong and indeed there is now a question about whether the paper should be retracted because the analysis is flawed.
Eukaryotes arose via some kind of merger event. The hybrid genomes of every eukaryote shows this.
I met Lynn Margulis once and she asked me why the rRNA sequences from mitochondrial genomes never featured on "Three-domains" trees. She really did have a point here. Those sequences were carefully ignored and in some places are ignored today. There is no reason for this and it has promoted a generation of poor-quality thinking.
In any case, we are using the names "Archaebacteria", "Eubacteria" and "Eukaryote/Eukaryota" because of the association between names and hypotheses. The three-domains of life hypothesis has been falsified and therefore the names "Archaea", "Bacteria" and "Eukarya" should now be put away.
Obviously, future evidence would change my mind, but for now, the proposal for the three domains has fallen, according to the data.
What should now be the "standard model" is shown in the image at the top of this post.
In fact, a better model would have more reticulations.
James O. McInerney, Mary J. O'Connell & Davide Pisani (2014) The hybrid nature of the Eukaryota and a consilient view of life on Earth. Nature Reviews Microbiology (2014) doi:10.1038/nrmicro3271