Two scenarios for interpreting the three polarizations. A) Under the canonical rooting proteasome evolution would require several selective sweeps and large-scale loss. The monomer PyrD B would have evolved from one of the more complex quaternary structures, and the derived insert in EF-2 would occur after LACA. B) Under the gram-negative rooting, Anbu could be ancestral to both HslV and the 20S proteasome. PyrD could evolve via stepwise increases in structural complexity, and there is no need to invoke extinct stem archaea to explain the EF_G insert. We believe these transitions argue for a gram-negative rooting.