A mapping framework of competition–cooperation QTLs that drive community dynamics

Genes have been thought to affect community ecology and evolution, but their identification at the whole-genome level is challenging. Here, we develop a conceptual framework for the genome-wide mapping of quantitative trait loci (QTLs) that govern interspecific competition and cooperation. This framework integrates the community ecology theory into systems mapping, a statistical model for mapping complex traits as a dynamic system. It can characterize not only how QTLs of one species affect its own phenotype directly, but also how QTLs from this species affect the phenotype of its interacting species indirectly and how QTLs from different species interact epistatically to shape community behavior. We validated the utility of the new mapping framework experimentally by culturing and comparing two bacterial species, Escherichia coli and Staphylococcus aureus, in socialized and socially isolated environments, identifying several QTLs from each species that may act as key drivers of microbial community structure and function.


Supplementary Tables
Supplementary Table 1 The estimates of growth parameters (and their standard errors) for microbial abundance by Gompertz (G), logistic (L), and Richards (R) equations in monoculture and by these three equations and Lotka-Volterra (LV) equations in co-culture. According to F-test, the optimal equation is the R for both bacterial species in monoculture, since F-values for both G and L are significant (see the text). The AIC values calculated identify the LV as optimal equations for both species in co-culture. Power was empirically calculated as the proportion of the number of simulation replicates in which significant QTLs were detected over the total number of simulation replicates (1000) for the data simulated under the assumption of QTL occurrence. False positive rates (FPR) were calculated as the proportion of the number of simulation replicates in which significant QTLs were detected over the total number of simulation replicates (1000) for the data simulated under the assumption of no QTL.

Supplementary Figures
Supplementary Figure 1 Goodness-of-fit of the LV equation (2) to observational abundance data for E. coli and S. aureus in co-culture. a, the fitted mean curve (thick line) of all strains for each species, with raw data shown by grey lines. b, random scatters of residuals over predicted values across each strain (circle) by the ODEs, warranting the statistical behavior of data fitting.
Supplementary Figure 3 States of two interacting species at four genotypic combinations C/C, C/T, T/C and T/T for E4614704 and S188004. By setting dNe/dt = 0 and dNs/dt = 0 to obtain zero isoclines for each species, ODE (2) charts a state-space of the combination of abundance between two species. The space has four types: (i) a predation/parasitism of E. coli over S. aureus, (ii) a predation/parasitism of S. aureus over E. coli, (iii) the cooperation between two species by benefiting from each other, and (iv) the competition by excluding from each other to exist in unstable equilibrium. A, plot of the abundance of S. aureus (solid line) against E. coli (slash line) at each genotypic combination. Four interspecific genotypes at QTLs E4614704 and 188004 present consistently a similar pattern of competition, but the degree of such competition varies strikingly among the genotype combinations, suggesting that these two QTLs, through their across-genome combination, are involved in the regulation of competition between E. coli and S. aureus. B, limit cycles of four genotypic combinations between two species.

Supplementary Figure 4 Biological validation of competition-cooperation mapping (CoCoM).
The overall growth of E. coli (A) and S. aureus (B) in co-culture (thick blue line) fitted to raw abundance data (thin blue lines) was decomposed into the independent growth (red solid line) and dependent growth (green line) at two alternative genotypes C and T for E4614704 and C and T for S188004. The broad agreement of the independent growth in co-culture estimated by CCM with the growth in isolation for each genotype (red triangles fitted by slash curves) suggests that the model can accurately capture the biologically grounded rules of ecological interactions.
Supplementary Figure 6 Simulation results by mimicking one significant QTL pairs detected from the real example in terms of sample size, heritability and ODE parameters. The overall curve, independent curve and dependent curve of each species are denoted by solid, slash and dot lines, respectively. The pink background represents the 5% confidence interval of the estimated overall curve under the sample size of 45 and heritability of 0.1.