(A) Direct fitness effects. To define direct effects, we use a framework similar to Chapter 10 of Peters et al. (2017). For simplicity, consider a commensal community where population A benefits population B. (i) Populations A and B grow over time t1, t2, etc. Basal growth rates of monoculture populations are marked by dashed arrows. A also releases a metabolite which promotes the growth of B (solid diagonal arrow). Thus, B(t2) will depend on both B(t1) and A(t1). (ii) Suppose that at time t1, A acquires a mutation (green) which does not alter basal growth rate, but increases benefit supply to partner (thick green arrow). This will increase B(t2) even if we had held the dynamics of mutant A to that of the ancestral A. We define such a mutation as “strictly partner-serving”. (iii) At time t1, A acquires a mutation (blue) that increases A’s basal growth rate (thicker blue arrow), but not benefit supply rate. This mutation will promote B(t3) via increasing A(t2), but will not promote B(t3) if we had held the dynamics of mutant A to that of the ancestral A. Because increased B(t3) is indirect (mediated by increased A(t2)), we define this mutation as “strictly self-serving”. (iv) A win-win mutation (brown). (B) Mutation types in a mutualistic community and their evolutionary fates. Mutations that exert a positive direct effect on self (selfish, strictly self-serving, and win-win) are favored in a well-mixed environment. In a spatially-structured environment, effects on self and on partner are both important. For example, a spatially-structured environment may favor an altruistic mutation that confers a large benefit on partner at a small cost to self. Parentheses indicate that selection outcome (favored or disfavored) depends on quantitative details of the fitness effects on self and partner (see Momeni et al., 2013b for an example). Note that a mutation that is strictly partner-serving or altruistic could still rise in frequency in a well-mixed environment by “hitchhiking” with other self-serving mutations (Morgan et al., 2012; Waite and Shou, 2012). (C) CoSMO. CoSMO is an engineered mutualistic community consisting of two non-mating S. cerevisiae strains (Hart et al., 2019a; Shou et al., 2007). Thus, the two strains may be regarded as two species. The mCherry-expressing L-H+ strain is unable to synthesize lysine (L) and overproduces the adenine precursor hypoxanthine (H). The complementary GFP-expressing H-L+ strain requires hypoxanthine and overproduces lysine. Both overproduction mutations render the first enzyme of the corresponding biosynthesis pathway insensitive to end-product feedback inhibition control (Armitt and Woods, 1970; Feller et al., 1999). In minimal medium lacking exogenously supplied L and H, the two strains form a mutualistic community where live cells from both strains release overproduced metabolites (Hart et al., 2019a) and support each other’s growth.