To examine the relationship between the genetic similarity of strains and the amount of segregation they exhibit during the formation of fruiting bodies, we performed pairwise mixes of a reference strain and a panel of natural isolates (Table S1
). To estimate the genetic distances between strains, we genotyped them at 12 polymorphic microsatellite loci, which were dispersed throughout the genome. We calculated the standardized Euclidean distance between strains based on their microsatellite allele sizes and used it as an estimate of genetic divergence, and thus as a proxy for the probability that strains share alleles (Table S2
). Genetic distance is thus similar to relatedness in that both measures are estimates of identity by descent, although they differ formally, since the latter is expressed relative to allele frequencies in a reference population [29
]. More important, because genetic distance takes into account not just allelic identity but distance between alleles, it provides greater resolution than relatedness measures based on shared alleles for divergent strains sampled from different geographic locations.
We first mixed the laboratory strain AX4-GFP (labeled by transformation with the gene for green fluorescent protein) with each of 14 natural isolates, the strain from which it was derived (natural isolate NC4), and unlabeled AX4 (control). For each mix, we combined labeled and unlabeled amoebae in equal proportions, deposited the mixture on damp nitrocellulose filters, and allowed them to aggregate and form fruiting bodies. We sampled ten fruiting bodies from each mix and determined the ratio of fluorescent to nonfluorescent spores in each one. We used the average variance in this proportion across fruiting bodies, based on a minimum of three temporally independent replicates, as an estimate of the degree of segregation for a given pair of strains.
The mixing experiment could have several outcomes (). In the absence of any discrimination, all fruiting bodies should show identical proportions of the two clones, resulting in low variance in that measure and no differences between mixes of isolates at different genetic distances (A). Under exclusive self–nonself discrimination, individuals would be expected to cooperate and form fruiting bodies with genetically identical cells but segregate from all other strains, resulting in a strongly binary response (B). Alternatively, if the degree of discrimination depends on the genetic similarity between the strains, we expect to see a graded relationship between genetic distance and the degree of sorting (C).
Hypothetical Patterns of Discrimination
The results of the mixing experiments support the third model (). When AX4-GFP was mixed with either unlabelled AX4 (control) or with the parental wild isolate NC4 (rank genetic distance 1 and 2, respectively), the proportion of GFP-positive spores was similar between different fruiting bodies (A) and the variance was low (B), indicating low sorting. However, mixes of AX4-GFP with isolates of increasing genetic distance resulted in greater segregation, reflected in the higher variance, and mixes of the most genetically distant strains resulted in fruiting bodies of two classes, indicating that the strains segregated from one another (). We observed a highly significant correlation between the genetic distance and the variance (Pearson correlation coefficient: r = 0.773, n = 16, two-tailed p <0.0001), indicating that segregation increased in proportion to the genetic distance between strains. Because the genetic distances were non-normally distributed, we also performed a nonparametric correlation, which was also highly significant (Spearman rank correlation: ρ = 0.631, n = 16, two-tailed p = 0.009). Finally, despite limited resolution to discriminate between the more distantly related strains, analyses in which genetic distance was estimated based on the number of shared alleles rather than allele size differences produced similarly significant results (Spearman rank correlation: ρ = 0.798, n = 16, p = 0.0002).
Segregation Increases with Genetic Distance in Mixed Fruiting Bodies
To test the generality of the result, we repeated our experiments with a different combination of strains. We chose two natural isolates (QS32 and QS33), which mixed poorly with AX4-GFP but were closely related to one another (identical at all microsatellite loci we examined), and a third strain (QS38), which was equally dissimilar to both (Table S1
). If the degree of discrimination can be predicted on the basis of genetic similarity, then the genetically similar strains QS32 and QS33 should mix well with one another and segregate from the genetically distant strain QS38. To test this prediction, we labeled the strain QS32 with a vital fluorescent dye and developed it in pairwise mixtures with the other two strains and with unlabeled QS32 cells as a control. We observed little segregation in the control mix and in the mix of the genetically similar strains QS32 and QS33 (). By contrast, mixing QS32 with the genetically distant strain QS38 resulted in fruiting bodies with more variable proportions of labeled spores, indicating stronger segregation. These results are consistent with the studies performed with the labeled laboratory strain, suggesting that the property of genetically related segregation is transitive [31
] and robust to strain choice.
The Property of Segregation Is Transitive and Robust to Strain Choice
Segregation could result from differential aggregation or from post-aggregative segregation. To distinguish between these possibilities, we transformed one of the natural isolates with a GFP-expression vector (QS44-GFP) and mixed it with a genetically dissimilar strain, labeled with a DsRed expression vector (AX4-DsRed, A). As a control, we also mixed AX4-GFP cells with AX4-DsRed cells (B). In both the experiment and control mixes, the labeled and unlabeled cells were well mixed when initially plated on agar to induce development (A and B, 0 h). As development progressed, both mixes initiated aggregation and all the cells moved toward the same aggregation centers regardless of their genetic similarity (A and B, 9 h). However, clusters of differentially labeled cells became increasingly evident in mixes of the genetically dissimilar strains (A, 9 h) whereas the genetically identical cells remained intermixed (B, 9 h). Segregation of the genetically dissimilar strains continued throughout the aggregation stage, at which point there was partial separation of labeled and unlabeled cells into different aggregates, as well as segregation within aggregates (A, 13 h). Thus, genetically different strains segregate, but they do so imperfectly. The control mixes showed no segregation at that stage or at any later time (B, 13 h, and unpublished data). We observed similar segregation in mixes of AX4 with the genetically different isolates QS32 and QS38 (unpublished data). The post-aggregative nature of segregation suggests that the sorting does not result from differences in developmental timing or the use of different chemoattractants, which are known to reduce interspecific chimerism [27
Sorting of Strains during Multicellular Development