Fig. 3. Obligate symbiosis enables evolutionary shifts to diets deficient in B vitamins.

a,b, The ancestors of lineages that evolved obligate symbioses had similar levels of B5 (a) and B9 vitamins (b) in their diets compared with lineages that did not evolve obligate symbioses. c,d, After acquiring obligate symbionts, lineages evolved diets with significantly lower levels of B5 (c) and B9 vitamins (d) (‘non-obligate maintenance versus obligate maintenance’ comparison). The evolutionary loss of obligate symbiosis was also associated with increases in dietary levels of vitamin B5 and B9 compared with where obligate symbiosis was maintained (c,d: ‘obligate maintenance versus obligate loss’ comparison). Vitamin B concentrations are standardized amounts per gram for each ancestral node, estimated from the diets of extant insect families (‘Nutrient data’ in Methods; for reconstructed levels of B vitamins plotted on the tree and robustness of estimates to rate shifts in B vitamins, see Extended Data Fig. 4). Numbers along the x axis in a and b indicate the numbers of transitions. Violin density curves represent the posterior distribution of estimated ancestral levels of B5 and B9 vitamins (1000 samples) estimated using a BPMM. The violin width corresponds approximately to the most likely estimate of B vitamins. A BPMM was used to test for significant differences between transitions (*pMCMC <0.05, **pMCMC <0.01, ***pMCMC <0.0001; exact pMCMC values are given in Supplementary Table 10): In a and b we tested if the posterior distribution of B-vitamin estimates for a given transition was above or below the comparison transition, and in c and d we tested if the posterior distribution of the difference in B-vitamin estimates for transition comparisons was above or below 0. NS, not significant.