Models with unequal transition rates favor marine origins of Cyanobacteria and photosynthetic eukaryotes.

Using a broadly sampled, time-calibrated phylogeny of Cyanobacteria and photosynthetic eukaryotes, Sánchez-Baracaldo et al. (1) report that the mostrecent common ancestor (MRCA) of Archaeplastida likely inhabited a low-salinity (freshwater) habitat, providing an advance on a longstanding debate about the ecological context for the origin of plastids (2). The freshwater ancestry for Cyanobacteria and Archaeplastida was based on reconstructions that assumed equal transition rates among marine, brackish, and freshwater habitats (1), which likely oversimplified the 2.5-billion-year history of these groups. We relaxed the assumption of a single transition rate by fitting a model with additional parameters for independent transition rates among marine, brackish, and freshwater habitats. This model provided a better fit to the data (relative model probability, wi, was 99%) and revealed strongly unequal transition rates, with the probability of colonizing freshwaters from the marine habitat more than double the rate of the reverse (qm→f = 0.36 vs. qf→m = 0.16 transitions per billion years). Consequently, the marine environment was strongly supported as the likeliest ancestral habitat for Cyanobacteria and Archaeplastida (3). It is possible that direct transitions between marine and freshwaters are uncommon (4) and proceed through an intermediate brackish habitat (5). We fit models with ordered transitions and again found unequal transition rates, with marine taxa far more likely to colonize brackish waters than their freshwater counterparts (3). Models that assumed equal rates, either unordered or ordered, were highly improbable (wi < 0.01%), whereas an unequal-rates model with ordered transitions (freshwater↔brackish↔marine) provided the best overall fit to the data (wi = 85.6%). With this model, the freshwater habitat was no longer favored for the MRCAs of Cyanobacteria (marginal probability for freshwater, Pf < 0.01%) and Archaeplastida (Pf = 2.8%). When assuming unequal rates for multistate characters, low frequency states, like the brackish state in this dataset (just 2 of the 119 species are brackish), require extremely high transition rates away from them to account for their low observed frequency and can, as a result, affect ancestral reconstructions (6). The data deposited by Sánchez-Baracaldo et al. (7) contain a second coding scheme, with brackish taxa coded as marine and a different scoring for one Nostoc accession (scored as marine instead of freshwater). With this dataset, a model with unequal transition rates was again favored (wi = 61% vs. 39%). We accounted for uncertainty in model choice by averaging transition rates across the two models. The results again favored asymmetry in the rates of marine–freshwater transitions (qm→f = 0.56, qf→m = 0.34) and strongly supported marine ancestry for both Cyanobacteria (Pm = 94%) and Archaeplastida (Pm = 82%). The comparative framework employed by SánchezBaracaldo et al. (1) represents a promising way forward for reconstructing the history of endosymbiosis. However, as demonstrated here, even modest heterogeneity in the evolutionary process can clearly have profound effects not just on the rates of transition among states, but also—and perhaps most importantly—on inferences about events that have occurred in the very distant past (8).