From individual behaviors to collective outcomes: fruiting body formation in Dictyostelium as a group-level phenotype

Collective phenotypes, which arise from the interactions among individuals, can be important for the evolution of higher levels of biological organization. However, how a group’s composition determines its collective phenotype remains poorly understood. When starved, cells of the social amoeba Dictyostelium discoideum cooperate to build a multicellular fruiting body, and the morphology of the fruiting body is likely advantageous to the surviving spores. We assessed how the number of strains, as well as their genetic and geographic relationships to one another, impact the group’s morphology and productivity. We find that some strains consistently enhance or detract from the productivity of their groups, regardless of the identity of the other group members. We also detect extensive pairwise and higher-order genotype interactions, which collectively have a large influence on the group phenotype. Whereas previous work in Dictyostelium has focused almost exclusively on whether spore production is equitable when strains cooperate to form multicellular fruiting bodies, our results suggest a previously unrecognized impact of chimeric co-development on the group phenotype. Our results demonstrate how interactions among members of a group influence collective phenotypes and how group phenotypes might in turn impact selection on the individual.


29
Individuals frequently form groups and engage in cooperative behaviors within them that benefit the 30 members. In some cases, group benefits arise from a higher-level phenotype that emerges from the 31 interactions among its members. For example, herding behaviors can produce an optical illusion that

40
Prior studies have shown diverse effects of the number and identity of individuals in a group on 41 collective phenotypes. On one hand, genetic diversity sometimes confers benefits, referred to as social 4 cell fate (i.e., survival versus death), means that selection can favor selfish behaviors. In 83 Dictyostelium, a genotype that can avoid the stalk fate or induce its partner to form it should have a 84 selective advantage, referred to as cheating. Apparent cheating behaviors are indeed observed among 85 different natural isolates when they are co-developed to form chimeric fruiting bodies (Strassmann et

100
To date, studies of chimerism in Dictyostelium have largely focused on whether spore production in 101 chimeras is equitable-that is, on quantifying the extent of cheating. By comparison, less effort has 102 been made to understand the consequences that chimerism entails for group-level phenotypes 103 (although see Foster et al. (2002) and Votaw and Ostrowski (2017)). Yet, individual-level behaviors 104 5 have the potential to impact the phenotype of the collective. For example, groups comprised of stalk-105 avoiding cells might produce fruiting bodies with shorter stalks, leading to poor spore survival. In 106 short, group-level phenotypes have the potential to impact the cost-to-benefit ratio of the collective 107 behavior and feedback to influence the decisions of individuals about whether to participate.

109
Indeed, perhaps owing to a history of conflict, strains of D. discoideum have evolved mechanisms that 110 allow them to recognize and avoid unrelated strains following aggregation (Ostrowski et al. 2008).

111
This kin recognition system involves a pair of heterophilic cell membrane proteins, tgrB1 and tgrC1, 112 that are highly polymorphic in natural populations (Benabentos et al. 2009). It is thought that these

122
Here we take advantage of the Dictyostelium discoideum model system to quantify how group 123 composition-specifically, the number of strains and their geographic and genetic relationships to one 124 another-impacts the size, number, and productivity of collectives. We employ a statistical approach 125 to detect and quantify genotype interactions involving the estimation of Walsh coefficients 126 (Beerenwinkel et al. 2007;Weinreich et al. 2013;Poelwijk et al. 2016). This approach has been used 127 previously to quantify genetic interactions among alleles that influence the phenotype of a cell, termed

133
In addition to quantifying the G ´ G interactions, we assess other metrics of group composition and 134 their influence on collective phenotypes. We ask how the cooperative sporulation and morphology of   We assessed the evidence for pairwise and higher-order genotype interactions by estimating Walsh 148 coefficients (see Methods and SI.) (B) We assessed whether the co-occurrence of strains in the same 7 (=sympatric) vs different (=allopatric) soil samples led to higher or lower co-sporulation when co-150 developed together. Whereas higher sporulation of co-occurring strains could indicate mutually 151 beneficial co-adaptation, worse sporulation of co-occurring strains could indicate enhanced levels of 152 conflict owing to local competition. Finally, genetic divergence in allopatry could lead to 153 incompatibilities that reduce the ability to co-sporulate. We tested this hypothesis by asking whether 154 allopatric groups showed a significant negative correlation between among-strain genetic diversity 155 and spore production.

158
Strain isolation - Table S1 provides a list of the strains and the GPS coordinates of the sites from 159 which they were isolated. Briefly, three D. discoideum isolates were obtained from soil samples at

165
To isolate D. discoideum, 6 g of each soil sample was mixed with 30 ml of sterile water. The soil-166 water slurry was agitated briefly, and approximately 200 µl was deposited on hay infusion agar plates, 167 with 400 µl of the bacterium Klebsiella pneumoniae as a food source. Spores were collected from a 168 single fruiting body per plate and frozen at -80º C in KK2 buffer (per L: 2.25 g KH2PO4, 0.67 g 169 K2HPO4) with 20% glycerol. Each isolate was subsequently cloned to ensure genetic homogeneity by 170 plating the spores at low density on a lawn of K. pneumoniae on a Sussman's medium plate (SM; 171 Formedium, Ltd; with 2% agar) and picking fruiting bodies from the center of a single, isolated 172 plaque. Spores of cloned isolates were grown to high density and then stored in KK2 buffer with 20% 173 glycerol in a -80º C freezer.

175
Overview of Experimental Design -We grouped the 12 strains into two sets of six, where each set 176 contains three strains from two sites, as shown in Figure 1B         indicate the group produced more spores than expected, based on the strains' clonal sporulation.

235
We assessed overall evidence for additivity by regressing the observed sporulation of a group on its 236 expected sporulation. Slopes and intercepts were estimated using standardized major axis regressions 237 to account for error in both the independent and dependent variables. Regressions were estimated in R 238 using the package 'smatr', with slope.test=1 and elev.test=0. Principal components analysis was 239 carried out using the 'prcomp' function in R, with centering and scaling. Statistical models of 240 observed sporulation as a function of expected sporulation, geography (allopatry/sympatry), number 241 of strains, genetic distance, and block were carried out in R using glm. Owing to the small number of 242 sites and strains, all factors were modeled as fixed effects.

244
Epistasis Coefficients -To estimate first-order (i.e., individual strain), pairwise, and higher-order 245 epistasis coefficients, we use the group sporulation data to calculate Walsh coefficients (Weinreich et

277
The mean effect of mixing on sporulation efficiency also differed between the two sets of strains. For 278 NC/MA strain combinations, the mean deviation from expected sporulation was significant and 279 positive (=7%), indicating that combining strains tend to enhance the spore production of the group 280 (one-sample t-test: t=8.99, df=56, P<0.0001). This finding was largely driven by the NC strains, 281 which produced many more spores as a group than expected based on their sporulation in 282 monocultures. In contrast, the MA strains performed somewhat worse than expected. For VA/NH 283 strain combinations, the average deviation between expected and observed spore production was 284 significant and negative (-3%; one sample t-test: t = -5.2, df=56, P<0.0001), indicating a cost of 12 mixed groups on spore production, at least with respect to spore number. We discuss these results in 286 more detail in later sections.

295
Pairwise and higher-order genotype interactions influence group sporulation.

296
To determine the nature of interactions leading to group effects on spore production, we estimated

307
(1-6; see Table S1) mixed and analyzed to estimate the Walsh coefficient directly above. Each column 14 expected based on how well members sporulate individually. We also looked for the converse-320 strains that consistently elevate the sporulation of their groups-which we refer to as 'good eggs'. To 321 do so, we used ANCOVA to determine how the presence or absence of each strain impacts the 322 sporulation of its groups after controlling for their expected sporulation (Fig. S3). These analyses 323 show that several strains are 'good eggs': strains 1, 2 (both NC), strain 6 (from MA), and strain 1 324 (from VA). Adding these strains to a group increases its sporulation by ~4% from its expected values.

325
Our analyses show a consistent negative effect for only one strain, strain 4 (from NH); groups with 326 this strain do 2.8% worse than expected based on their clonal sporulation. Together, these results 327 indicate that certain strains interact with the presence of other strains in a consistent way.

329
Strain diversity enhances sporulation and alters fruiting body morphology.

330
To address the sporulation patterns we observed, we also asked whether increasing the number of 331 strains in a mix enhances or reduces the sporulation of the group. We expected that sporulation might 332 decline as the number of strains in the group increases, in part because many of these strains are 333 genetically and geographically distant (>1000 km) from one another (Fig. 2). We hypothesized that

339
Given these expectations, we were surprised to see the opposite relationship (Fig. 5). For NC and MA 340 strain combinations, there was a significant positive correlation between the number of strains in the 341 mix and the number of spores they produced. For VA/NH strains, the relationship was not significant, 342 but the sign was also positive, and the 5-way and 6-way mixes were amongst the highest performing 343 mixes in terms of spore production. At least for this metric of productivity, increased strain diversity 344 was associated with improvements in group performance, suggesting that D. discoideum shows social

359
Spore production of the group is one metric by which we can measure the success of fruiting body 360 formation. However, additional group attributes might influence the collective's success, aside from 361 the number of spores they produce. For this reason, we also measured fruiting body size, number, and 362 shape (Fig. 6). For NC/MA strain combinations, as the number of strains in the group increased,

392
Geography had inconsistent effects on the ability of strains to co-sporulate ( Fig. 7; Table S2)

417
However, the differences among these groups were not significant (F2,336=1.64, P=0.196). Taken 418 together, these results reveal a wide range of interactions among sympatric strains: from synergism 419 and enhanced spore production as a group (NC strains) to performance that is significantly worse-420 than-expected (NH strains) and even worse than allopatric combinations (e.g., MA strains do worse 421 with each other than they do with NC strains.)

423
Limited evidence that evolution in allopatry leads to accumulation of social incompatibilities.

424
A major hypothesis in social evolution is that social interactions among conspecifics lead to 425 adaptations and counter-adaptations that drive region-specific co-evolutionary trajectories (West-

431
We did not find that increased genetic divergence was associated with reductions in sporulation for 432 allopatric mixes, defined as those where at least two strains come from different sites (Fig. 7B). The

435
Although each correlation was not significantly different from zero, the two correlations were 436 significantly different from one another (P=0.03, confidence intervals and difference estimated using 437 'twocorci' package in R).

439
Putting it all together: how does the composition of groups influence collective phenotypes?

440
We used principal components analysis (PCA) to develop a better understanding of the relationship 441 between the different group morphological attributes, sporulation, and how these traits are influenced 442 by the number of strains in the group and the genetic differences between them (Fig. 8).

457
To assess the statistical significance of each explanatory factor, we modeled the observed sporulation 458 of each group as a function of its expected sporulation (i.e., mean clonal sporulation, which accounts 459 for any differences among strains in clonal spore-stalk allocation), geography (allopatry versus 460 sympatry), genetic distance, and the number of strains. These analyses are summarized in Table S3.

481
In contrast to the predictable effects of genotypic diversity on group size, which makes sense given

482
what is known about how kin recognition works in Dictyostelium, the observation that more diverse 483 groups also produced more spores was more puzzling, and the most parsimonious explanation is that 484 cells alter their spore-stalk allocation in response to detection of non-self. We thus wondered whether  morphology-that are likely to be closely tied to fitness in nature. Our experimental design is novel in 518 that it applies experimental approaches and mathematical analyses that have been successfully used to 519 quantify the form and extent of g × g epistasis (that is, genetic interactions among alleles in an 520 individual) to the detection of epistasis at a higher level of biological organization.

522
Our findings indicate a consistent influence of particular strains on group phenotypes, but also the 523 existence of extensive pairwise and higher-order epistasis. One factor that strongly influenced the 524 group morphology and behavior was the number of strains in the group. Although we had predicted 525 that more strains (i.e., more genetic diversity) would lead to the production of fewer spores, we found 526 that groups with more strains tended to have greater, rather than lesser, spore production. This effect

555
To date, studies with Dictyostelium have focused primarily on one type of fitness cost that cells might 556 experience when they join forces to build a collective: they might be cheated on, whereby one 557 genotype contributes more than its fair share to the production of the dead stalk, thus gaining the  cheating-those who are choosy will not gain the full benefits of cooperation (Crozier 1986

589
We also found that genetic diversity typically leads to increases in spore production of the groups.

590
Thus, to some extent, Dictyostelium strains show social heterosis, at least for productivity measured as 591 the number of spores produced (Nonacs and Kapheim 2008). This result is hard to explain unless we 592 accept that the strains can facultatively shift their cell fate towards spore production in response to

629
Addressing how and why group behaviors evolve is challenging, but the Dictyostelium model system