Rapid and widespread de novo evolution ofkin discriminationOlaya Renduelesa,1, Peter C. Zeeb,c, Iris Dinkelackerd, Michaela Amherda, Sébastien Wielgossa, and Gregory J. Velicera,b,d

aInstitute for Integrative Biology, ETH Zürich, 8092 Zürich, Switzerland; bDepartment of Biology, Indiana University, Bloomington, IN 47405; cDepartment ofBiology, California State University-Northridge, Northridge, CA 91330; and dDepartment of Evolutionary Biology, Max-Planck Institute for DevelopmentalBiology, 72076 Tübingen, Germany

Edited by Richard E. Lenski, Michigan State University, East Lansing, MI, and approved June 1, 2015 (received for review February 3, 2015)

Diverse forms of kin discrimination, broadly defined as alterationof social behavior as a function of genetic relatedness amonginteractants, are common among social organisms from microbesto humans. However, the evolutionary origins and causes of kin-discriminatory behavior remain largely obscure. One form of kindiscrimination observed in microbes is the failure of geneticallydistinct colonies to merge freely upon encounter. Here, we first usenatural isolates of the highly social bacteriumMyxococcus xanthus toshow that colony-merger incompatibilities can be strong barriers tosocial interaction, particularly by reducing chimerism in multicellularfruiting bodies that develop near colony-territory borders. We thenuse experimental laboratory populations to test hypotheses regard-ing the evolutionary origins of kin discrimination. We show that thegeneric process of adaptation, irrespective of selective environment,is sufficient to repeatedly generate kin-discriminatory behaviors be-tween evolved populations and their common ancestor. Further, wefind that kin discrimination pervasively evolves indirectly betweenallopatric replicate populations that adapt to the same ecologicalhabitat and that this occurs generically in many distinct habitats.Patterns of interpopulation discrimination imply that kin discrimina-tion phenotypes evolved via many diverse genetic mechanisms andmutation-accumulation patterns support this inference. Strongincompatibility phenotypes emerged abruptly in some popula-tions but strengthened gradually in others. The indirect evolutionof kin discrimination in an asexual microbe is analogous to theindirect evolution of reproductive incompatibility in sexual eu-karyotes and linguistic incompatibility among human cultures,the commonality being indirect, noncoordinated divergence ofcomplex systems evolving in isolation.

social evolution | nonself recognition | kin recognition | cooperation |territoriality

The behavior of many vertebrates (1, 2), invertebrates (3, 4)and microbes (5–7), as well as some plants (8), is altered by

social encounters with conspecifics in a relatedness-dependentmanner. Such kin discrimination is commonly interpreted in thecontext of inclusive fitness theory to have resulted from positiveselection for kin discrimination per se (7, 9–11) specifically becausesuch discrimination promotes preferential cooperation among kinthat share cooperation alleles (12–14). However, despite any the-oretical plausibility of kin selectionist explanations, the actualevolutionary origins of kin discrimination in natural populationsare often difficult to infer due to their temporal remove (11, 15).Kin discrimination, broadly defined (SI Appendix, Methods), mightalso arise indirectly as a byproduct of alternative evolutionaryforces (15–20). Despite decades of rigorous empirical and theo-retical investigation of kin discrimination and its causes, a biologicalsystem that allows both the documentation of kin discriminationorigins and causally proximate analysis of the evolutionary forcesresponsible for those origins has been lacking.Among motile microbes, one form of biological kin discrimina-

tion is reduced merger of genetically distinct swarming coloniesupon encounter relative to “self–self” controls. This phenomenonwas first documented in 1946 with the observation of “Dienes lines”

that formed between distinct nonmerging colonies of the bacteriumProteus mirabilis (21) and has since been found in several otherspecies (20, 22). For example, a large number of such colony-merger incompatibilities evolved among closely related genotypes ofthe cooperative bacterium Myxococcus xanthus in a natural centi-meter-scale population (20). Irrespective of their original evolu-tionary cause(s), the emergence of colony-merger incompatibilitiesin nature is likely to have profound implications for the distributionof social interactions among genotypes during cooperative pro-cesses. For example, barriers to colony merger may promote themaintenance of genotypes that are inferior social competitors inchimeric groups and promote cooperation by hindering the terri-torial spread of socially defective cheaters (20), as might other formsof kin discrimination (23).In this study, we first use natural isolates of M. xanthus to test

whether kin discrimination affects patterns of cooperation duringfruiting body development when migrating social groups encoun-ter one another. To do so, we quantify the frequency of chimerismamong fruiting bodies that form near the border of colony terri-tories, both for pairings of distinct colony-merger allotypes andself–self controls. Subsequently, we use experimentally evolvedlaboratory populations to test whether (i) generic adaptation bymigrating populations (irrespective of variable selective con-ditions) tends to generate colony-merger incompatibilities to-ward an ancestral genotype and (ii) whether mere independenceof the adaptive process causes allopatric populations to evolvetrait differences that generate kin discrimination phenotypesupon secondary contact. We also characterize temporal patternsof kin discrimination evolution and test whether kin discrimi-nation evolved by one or rather multiple molecular mechanisms.

Results and DiscussionNaturally Evolved Kin Discrimination Reduces Chimerism at Colony-Territory Borders. In response to nutrient deprivation, M. xanthus

Significance

Relatedness-dependent behavior modification is commonamong social organisms and has been a major feature of socialevolution theory for decades. However, the evolutionary cau-ses of kin discrimination are often unclear. Here, we documentmany spontaneous origins of kin discrimination in a social mi-crobe that appear to arise as indirect byproducts of adaptationat other traits and we show that kin discrimination evolves bydiverse genetic mechanisms.

Author contributions: O.R., P.C.Z., and G.J.V. designed research; O.R., P.C.Z., I.D., and M.A.performed research; O.R. and G.J.V. analyzed data; O.R. and G.J.V. wrote the paper; I.D.performed selection experiments; and S.W. performed genome analysis.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. Email: olaya.rendueles@env.ethz.ch.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1502251112/-/DCSupplemental.

9076–9081 | PNAS | July 21, 2015 | vol. 112 | no. 29 www.pnas.org/cgi/doi/10.1073/pnas.1502251112

cells cooperatively construct multicellular, spore-bearing fruitingbodies (24). We tested whether colony-merger incompatibilitiesbetween natural isolates of M. xanthus affect the spatio-geneticstructure of social interactions by reducing chimerism withinfruiting bodies that form near the interface of oncoming colo-nies. We examined three distinct isolates (A23, A47, and A96)sampled from a centimeter-scale natural soil population in whichmany colony-merger incompatibilities were previously docu-mented (20). Colonies were initiated from separate locationsand allowed to swarm toward one another on a low-nutrientagar surface on which colonies grow and swarm but eventuallydeplete growth substrates and initiate fruiting body development(SI Appendix, Fig. S1). Colonies of the same genotype markedwith distinct antibiotic-resistance markers were used as self–selfcontrols in which no lines of demarcation between colonieswere visible (e.g., Fig. 1A and SI Appendix, Fig. S1B). In con-trast, nonself encounters between distinct natural isolatesresulted in a visually evident line of demarcation between col-onies (Fig. 1B).By reducing developmental coaggregation of distinct geno-

types along intercolony borders, colony-merger incompatibilitiesshould benefit genotypes that compete poorly within chimericsocial groups in which mean relatedness among interactants islow relative to their performance within high-relatednessgroups (e.g., pure cultures) (20). One such genotype is A23,which sporulates at a high level (similar to that of A47) in puregroups, but which makes very few spores when forcibly mixedwith A47 at a 1:1 ratio before development (SI Appendix, TableS1) (20). In contrast, A47 sporulates at high levels in forced 1:1mixes with A23 as well as in pure culture, giving A47 an extremefitness advantage in those mixed groups.

To test the above hypothesis, we staged motility-driven encountersbetween colonies of A47 and A23 and assessed the frequency ofchimerism among fruiting bodies that formed on both sides of theinterface of oncoming colonies growing on low-nutrient medium. If acolony of A47 were capable of freely merging with an A23 colonyand coaggregating into fruiting bodies that form near the colony-interface border on the A23 side of the initial encounter zone atnearly a 1:1 ratio, A23 should suffer a great fitness cost in this regiondue to its poor competitive performance within mixed groups con-taining A47 (SI Appendix, Table S1). We first quantified chimerismin self–self encounter controls between differently-marked variantsof A23 and A47 and found that more than half of all sampledfruiting bodies across all experimental replicates were chimeric forboth controls [14 of 25 (56%) and 31 of 59 (53%) for A23 andA47, respectively]. In contrast, in nonself encounter assays be-tween A23 and A47, chimerism was rare among fruiting bodiesthat formed near colony-interface borders with only 2 of 58fruiting bodies sampled from both sides of the demarcation borderbeing chimeric across all replicates (Fig. 1C). On the A23 side ofthe demarcation line, the strong mixed-group competitor A47 waspresent in far fewer fruiting bodies [2 of 23 fruiting bodies sampled(9%)] than was predicted from control experiments (P < 0.0001for deviation from expectation of 13 chimeric fruiting bodies basedon A47 controls, two-tailed binomial test).Similar results were obtained in nonself colony encounters between

isolates A47 and A96. Like A23, A96 loses to A47 during de-velopmental competitions on low nutrient medium when populationsare forcibly mixed at a 1:1 ratio before development, although to alesser degree (SI Appendix, Table S1). In the colony-encounter assays,only one fruiting body among 24 (4%) isolated near intercolonyborders from the A96 side (across four replicate experiments) wasfound to contain A47 as well as A96 (Fig. 1D, P < 0.0001 for de-viation from expectation of 12 chimeric fruiting bodies based on A47controls, two-tailed binomial test). Like for A23, the colony-mergerincompatibility between A96 and A47 benefited A96 by nearlycompletely preventing A47 from being present in fruiting bodies onthe A96 side of the colony interface that A47 would have entered inthe absence of kin discrimination. Additionally, because A96 didproduce substantial numbers of spores in forced mixes (despitemaking fewer than A47, SI Appendix, Table S1), chimerism patternsin self–self controls would predict that, in the absence of kin dis-crimination, A96 should penetrate into some fruiting bodies on theA47 side of A47–A96 colony borders. However, no such chimerismwas found among 21 fruiting bodies sampled.These results demonstrate that naturally evolved social in-

compatibilities in M. xanthus can reduce interactions betweennonkin during a cooperative process, irrespective of their evo-lutionary origins. The natural isolates examined here and othersfrom the same locale diverged relatively recently (20), indicatingthat kin discrimination evolves over short evolutionary periods innature, but the relative contributions of various evolutionaryforces to maintaining kin discrimination traits at high frequenciesin natural populations are unknown. We thus turned to a moredefined evolutionary system and tested whether similar forms ofkin discrimination might evolve rapidly among laboratory pop-ulations of M. xanthus with known evolutionary histories and de-fined selective regimes. Specifically, we asked whether replicateexperimental populations that evolved in several distinct envi-ronments independently evolved colony-merger incompatibilitiestoward their common ancestor and/or toward one another.

Independent Adaptation Pervasively Generated Kin Discrimination.One hundred and four populations of M. xanthus were estab-lished from independent clones of two laboratory ancestors[GJV1 (25), and GJV2, a rifampicin-resistant derivative ofGJV1] and each population was allowed to evolve in one oftwelve distinct agar environments that varied in surface type(hard or soft agar), nutrient level (high or low Casitone), and/or

Fig. 1. Kin discrimination between natural isolates reduces chimerismamong fruiting bodies near territory borders. (A) Self–self encounter controlof oncoming colonies of the same natural isolate (A47) that differ only intheir antibiotic-resistance marker (K and R indicate kanamycin- and rifam-picin-resistant strain variants, respectively). No interface demarcation linewas visible for any self–self encounters. Dark spots are individual fruit-ing bodies. (B) Kin discrimination phenotype between distinct natural iso-lates. A visible line of demarcation formed between colonies of A47 andA96. Similar barriers formed during encounters between A47 and A23. Ar-rows indicate the colony-interface line. (C and D) Percentage of fruitingbodies (FB) chimeric for A23 and A47 (C) or A47 and A96 (D) sampled fromnear the interfaces of genetically distinct colonies (center, light gray) and fromself–self encounter controls [white, A23 (C) or A96 (D); dark gray, A47]. Av-erage chimerism across independent experimental replicates is shown. Errorbars represent 95% confidence intervals. *P < 0.05; two-tailed t test.

Rendueles et al. PNAS | July 21, 2015 | vol. 112 | no. 29 | 9077

EVOLU

TION

nutrient type (Casitone and/or prey bacteria) (SI Appendix,Table S2). In each treatment, either eight or twelve independentreplicate populations of swarming colonies were allowed to ex-pand outward for two week intervals, after which a sample fromthe leading colony edge was transferred to the center of a newplate, as described in SI Appendix, Methods and Fig. S2. Thistransfer protocol was shared by all treatments and selected fornumerical dominance at the leading swarm edge.We screened for kin discrimination (hereafter also “KD”)

phenotypes among 88 of these evolved populations in the formof merger incompatibilities between swarming colonies growingon soft nutrient-rich agar (Fig. 2). Evolved populations were firstscored for kin discrimination phenotypes in encounters with theirown experimental ancestor (hereafter “KD-A”). A majority ofcolonies grown from samples of evolved populations [55 of 88examined (63%), SI Appendix, Table S2] did not merge withancestral colonies in the same manner as paired colonies of theancestor in self–self encounters that freely merge into a contin-uous swarm lacking territorial demarcation (Fig. 2). KD-Aphenotypes evolved independently in at least half of all replicatepopulations examined within every evolutionary treatment ex-cept one (TPM hard agar, predation on Escherichia coli), inwhich KD-A evolved in only two populations (Fig. 2 and SIAppendix, Table S2). No significant effects of evolutionary sur-face type, nutrient level, nutrient type, or prey type on the fre-quency of KD-A evolution were evident (P > 0.05 in all cases).We tested whether, like natural isolates, experimentally evolved

populations also exhibit kin discrimination phenotypes at colonyborders during multicellular fruiting body development under low-nutrient conditions. Out of 28 populations examined, 15 (54%)were found to exhibit clear KD-A phenotypes under these con-ditions (e.g., Fig. 3). For several evolved populations examined,fruiting bodies were absent from an area along the territorial in-terface of evolved vs. ancestor colonies, whereas fruiting bodiesdid form in the same respective regions of self–self encountercontrols with both ancestral and evolved populations (Fig. 3).

Temporal Patterns of Kin Discrimination Emergence. To determinewhether temporal patterns of initial KD-A evolution were simi-lar or divergent across populations, we tested for the presence ofKD-A in temporally intermediate population samples in 20populations. The timing of KD-A emergence was found to varygreatly, including across populations from the same evolutionarytreatment (Fig. 4). In some cases, KD-A traits were already presentwithin three cycles (6 wk) of selection (e.g., P60 and P67), whereasin other populations KD-A was not evident until after 32 cycles(64 wk, e.g., P10 and P58). Of the 20 populations examined

for temporal dynamics of KD-A appearance, one populationtransitioned directly from merger compatibility to a visuallystriking form of incompatibility (P10), whereas nine first exhibitedincompatibility phenotypes of intermediate visual strength (butthat were nonetheless clear and consistent) before evolving morevisually conspicuous incompatibilities. Eight populations transi-tioned just once to incompatibility phenotypes of intermediatestrength, and two populations reevolved merger compatibility withtheir ancestor after having initially evolved intermediate pheno-types but then subsequently transitioned to strong incompatibilityphenotypes (P11 and P67) (Fig. 4). Variation in the visual con-spicuousness of incompatibility phenotypes indicate that theyshould not be considered merely in terms of discrete presence-or-absence states, but as phenotypes that vary in both population levelmorphology (Fig. 2) and visual prominence (Fig. 4).Within any given population over time, a KD-A incompati-

bility might be caused by a single mutation and then remainrelatively constant over time or might rather be strengthenedover time by multiple successive mutations. To distinguish be-tween these possibilities, for a subset of populations we pairedtemporal samples from all evolutionary time points in all possi-ble combinations and screened for kin discrimination betweenthose samples (“KD-T” for kin discrimination among time-pointsamples). As observed for KD-A, distinct temporal patterns ofKD-T were strikingly evident. In roughly half of the populations,an abrupt evolutionary transition occurred such that all temporal

Fig. 2. Kin discrimination phenotypes evolved de novo in all evolutionary treatments. Examples of kin discrimination phenotypes between colonies ofevolved populations from each evolutionary treatment and their ancestor (KD-A) on high-nutrient agar are shown. White inset text depicts respective labelsof evolved populations. Evolutionary treatments are stated above each picture. The two first panels show colony-merger phenotypes of self–self encountercontrols for the two ancestral variants. Ratios show the proportion of examined populations within each evolutionary treatment that exhibited consistentKD-A phenotypes (SI Appendix, Table S2). B.s., B. subtilis; CTT, media containing Casitone; E.c., E. coli; TPM, starvation buffer.

Fig. 3. Kin discrimination phenotypes between colonies of evolved pop-ulations and their ancestor (KD-A) on low-nutrient plates. Fruiting bodies failedto form along evolved-ancestor interface zones but do form along the interfacezones of self–self controls. “c” (cycle) indicates evolutionary time point.

9078 | www.pnas.org/cgi/doi/10.1073/pnas.1502251112 Rendueles et al.

samples from after the transition were incompatible with allsamples stored before the transition (Fig. 5A and SI Appendix,Fig. S3). In other populations, KD-T phenotypes evolved grad-ually such that their presence and strength among intermediatetime-point samples correlated with the temporal distance be-tween samples (and thus, presumably, genetic distance also; Fig.5B and SI Appendix, Fig. S3). These contrasting temporal pat-terns suggest that kin discrimination in M. xanthus can evolveboth in the form of all-or-nothing “kind discrimination” that isdetermined by distinct allelic states at a single locus and in moregraded forms in which the degree of social incompatibility scalesmore gradually with overall genomic relatedness (5–7).

Kin Discrimination Evolved by Diverse Genetic Mechanisms. To de-termine whether populations evolved KD-A by similar or distinctmechanisms, we staged colony encounters for 117 within-treat-ment pairs of terminal evolved populations (i.e., paired pop-ulations were descended independently from the same ancestorin the same environment). If the failure of independently evolvedpopulations to each merge with their common ancestor is due tothe same molecular mechanism, those populations are likely tobe compatible with one another. Any failure of colonies of in-dependently evolved populations to merge with each other(hereafter “KD-B” for between-population kin discrimination)implies that either they evolved distinct KD-A mechanisms thatalso cause the KD-B phenotype or they evolved distinct KD-Aand KD-B mechanisms. More than half of all such population

pairings (62 of 117, 53%) revealed clear KD-B phenotypes (SIAppendix, Table S3). From the pervasiveness of KD-B evolutionacross treatments, we infer that mere independence of the adaptiveprocess is sufficient to generate diverse molecular forms of kin dis-crimination (SI Appendix, Table S3).To document this molecular diversity at the genomic level, we

sequenced clones from all 24 terminal populations from the CTThard and soft agar treatments and analyzed mutation patterns (SIAppendix, Tables S4 and S5 and Fig. S4). Seventeen of these 24populations (67%) evolved KD-A (SI Appendix, Table S2). Ex-cluding the P29 clone (which is an apparent “mutator” carrying 435mutations; ref. 26), evolved clones had accumulated an average of13 mutations, a large majority of which altered amino acid se-quences (SI Appendix, Table S4). No mutation was found in traA, agene that encodes a protein implicated in outer membrane ex-change (OME) among M. xanthus cells and has been hypothesizedto have evolved as an adaptive greenbeard system for kin recogni-tion in natural populations (27). However, the absence of traAmutations in our populations and the occurrence of colony-mergerincompatibilities between natural isolates that carry the same traAallele (e.g., A47 and A96 in Fig. 1B, and ref. 27) together suggestthat TraA functional incompatibilities evolved indirectly in evolu-tionary isolation after other social barriers were already establishedand do not represent a directly selected greenbeard system (27).Fourteen genes were mutated in at least four populations and four

genes were mutated in eight or more (SI Appendix, Table S5).Among the latter, only one locus, the CRISPR-associated gene cmr4,was mutated only in populations that had evolved KD-A (SI Ap-pendix, Table S5). We thus sequenced cmr4 in all terminal pop-ulations and found two significant patterns. First, all cmr4polymorphisms were identical (cmr4-P72H, see SI Appendix), andsecond, this mutation was present only in populations descendedfrom four of the twelve ancestral subclones used to initiate our ex-periments (GJV1.1, 1.4, 1.5, and 1.6, SI Appendix, Table S2). Se-quencing of cmr4 in ancestral subclones revealed that the mutationwas already present in these four ancestors but not in any other. This

Fig. 4. Temporal kin discrimination phenotype-emergence patterns forsample populations. (A) Kin discrimination phenotypes between evolvedpopulations and their ancestor (KD-A) were classified into four qualitativecategories: freely merging (green), inconsistent (gray), incompatible (lightred), and strongly incompatible (red). White boxes indicate that the cyclecould not be analyzed. (B) Ancestral colonies (Left) encountering P10 or P35samples (Right) from different evolutionary time points.

Fig. 5. Kin discrimination phenotypes evolved both abruptly and gradually.Kin discrimination between temporal samples of the same population (KD-T)evolved abruptly in P10 (A) but gradually in P35 (B). Colony-encounterphenotype classifications for all possible pairs of time points within eachevolved population are shown.

Rendueles et al. PNAS | July 21, 2015 | vol. 112 | no. 29 | 9079

EVOLU

TION

result indicates that the source GJV1 culture from which the an-cestral subclones were derived was polymorphic for this crm4 mu-tation. However, despite the fact that cmr4-P72H did not evolve viaselection after the initiation of experimental evolution, populationscarrying this mutation showed a statistically disproportionatetendency to evolve KD-A (binomial test, P = 0.002). Thus, al-though the cmr4-P72H mutation clearly does not directly generateKD-A itself, it may increase the probability of evolving KD-A.(See SI Appendix for additional discussion of the cmr4 mutation.)As noted previously, the recurrent evolution of KD-B (i.e.,

multiple incompatibility allotypes among independent pop-ulations that share a common ancestor and evolved in the sameselective conditions) implies that multiple distinct mechanisms ofcolony incompatibility evolved. Patterns of mutation among the24 genome-sequenced populations are consistent with this im-plication. First, there is no single locus that was mutated in all 17of the populations that evolved KD-A. Second, among sevengenes that mutated independently in five or more populationsafter the initiation of experimental evolution (SI Appendix, TableS5), all of them were mutated in multiple populations that didnot evolve KD-A as well as in populations that did. These mu-tation patterns further indicate that kin discrimination traitsevolved due to mutations in different loci across populations.To demonstrate genetically divergent evolution of kin discrimi-

nation with finer temporal resolution, multiple clones from each ofthree populations showing KD-A (P10, P33, and P34) were isolatedindependently from both the generational time point at which KD-Afirst appeared (cycles 32, 9, and 9, respectively, Fig. 4) and the priortime point. [For P34, cycle 9 was the earliest cycle at which someindividual sampled clones exhibited clear KD-A, although the KD-Aphenotype of whole-population samples was not fully consistent untilcycle 15 (Fig. 4A).] These clones were then tested for the presence orabsence of all mutations detected in the terminal clone from therespective populations. Two mutations each first appeared at thefirst KD-A time points in P33 and P34, whereas four did so in P10(SI Appendix, Table S6). Importantly, none of the eight mutationsthat are candidates for direct causation of KD-A in these threepopulations occurred in the same gene, further indicating that KD-Aevolved by diverse molecular mechanisms across independent pop-ulations (SI Appendix, Table S6). As the genetic bases of kin-dis-criminatory behaviors become defined in an increasing number ofsocial species (28–30), it will be of interest to determine how fre-quently kin discrimination is caused by a high level of allelic richnessat only a single locus or variation at multiple loci.

Evolutionary Causation. The many KD-B phenotypes documentedin this study are indirect byproducts of evolutionary processesthat occurred independently within fully isolated populations.Whether by direct selection, hitchhiking, or genetic drift, distinctsocial-gene alleles that rose to high frequency in completely allo-patric populations reduce social compatibility of those populationsupon secondary contact. In this regard, KD-B incompatibilities areconceptually analogous to Bateson–Dobzhansky–Muller (BDM)allele incompatibilities that evolve indirectly and reduce repro-ductive compatibility between divergent populations of sexual eu-karyotes upon secondary contact (31, 32).Just as BDM incompatibilities strengthen incipient species

boundaries in sexual organisms, early barriers to social merger andcooperation in M. xanthus may constitute first steps in a long-termprocess of increasing social incompatibility. This view is supportedby those populations in our experiments in which the likelihood ofcompatibility between time-point samples was found to be a func-tion of evolutionary distance (e.g., P35, Fig. 5). Although BDMreproductive incompatibilities can result from differential selectiveforces, they can also result from populations following distinct ge-netic pathways of adaptation to an identical environment (33) justas indirectly generated forms of KD-B arose in our populations.

Similarly, incompatibilities among distinct lineages of otherbiological and cultural systems are proposed to have originatedindirectly, including hybrid necrosis in plants (34), somatic cellincompatibilities in animals (16), and linguistic incompatibilitiesin humans (35). The commonality in our analogy between the KD-Bincompatibilities that evolved in experimental populations of anasexual prokaryote and BDM incompatibilities in sexual eukary-otes or incompatibilities in many other complex systems (e.g.,human language) lies in the indirectness of their evolution (36).The rapid, pervasive and indirect evolution of KD-B pheno-

types documented here among experimental populations ofM. xanthusmay help explain the great diversity of social allotypesfound in natural populations (20). Crozier proposed that some“extrinsic” factor other than direct selection on kin-discriminatoryinteractions is necessary for the maintenance of diverse social al-lotypes because positive frequency-dependent selection on co-operative traits is expected to favor the most common socialallotype (11, 16, 37, 38). Our results suggest that, at least in somebiological systems, stochastic variation in adaptive trajectories maygenerate novel allotypes more rapidly than the most common al-lotype in a local population can increase toward fixation.Direct selection favoring biological incompatibilities per se can

also occur and adaptive explanations for interaction incompatibilitiesamong other noncognitive organisms have been proposed, includingallorecognition specificity among colonial invertebrates (15, 39) andvegetative fusion incompatibilities in fungi (40–43). In sexual sys-tems, reproductive barriers can be directly reinforced by selectionwhen hybridization upon secondary contact is maladaptive (44, 45).In our experiments, it is possible that some mutations causing

KD-A might have been adaptive in the social environment inwhich they first arose specifically because of kin-discriminatoryeffects. In this scenario, a KD-A mutation would be adaptivebecause it both (i) generated a fitness advantage requiring pref-erential interaction among clone-mates sharing the mutation and(ii) itself mechanistically caused such preferential interaction (46).Due to the large number of independent KD-A origins (n = 55)and the diversity of their molecular mechanisms, we consider itunlikely a priori that all 55 initial KD-A mutations rose to highfrequency specifically by selection on kin-discriminatory effects ofthose mutations, although this hypothesis may apply in some cases.Importantly, we note that initially latent colony-incompatibility

traits that do not first evolve because they are specifically favored byselection may later be subject to direct selection in newly encoun-tered social contexts. Some indirectly evolved kin discrimination traitsmay represent exaptations (47) that were initially nonadaptive (withrespect to their potential to cause kin discrimination) but later provedadaptive during interactions with genotypes that are more competi-tive in mixed groups than in segregated groups. Parsing out tempo-rally variable evolutionary forces acting on a given kin-discriminationtrait will often be exceedingly difficult. However, given both the easewith which latent kin discrimination traits can arise indirectly (Fig. 2)and the strong effects that such traits can exert on the spatio-geneticdistribution of social interactions among local neighbors (Figs. 1–4),we expect that such temporal variation in the forces maintaining kindiscrimination in natural populations is common.Studies examining interactions between conspecific natural

isolates of social microbes under laboratory conditions are in-creasingly common (6, 20, 48, 49) and adaptive hypotheses forthe origin of interaction phenotypes are often proposed, withvarying degrees of support. Our results empirically demonstratethat kin discrimination phenotypes that appear to be consistentwith the hypothesis of a directly adaptive origin mediated by kinselection can instead readily originate indirectly, simply as a re-sult of differential independent adaptation to a common envi-ronment. In a similar vein, quantitatively striking examples ofsocial cheating have previously been shown to evolve as indirectbyproducts of adaptation to an unstructured habitat rather thanas a result of selection for social cheating per se (50). Distinct

9080 | www.pnas.org/cgi/doi/10.1073/pnas.1502251112 Rendueles et al.

natural strains of social microbes that are experimentally forcedto compete during a focal social process (e.g., fruiting body de-velopment in Myxococcus bacteria or Dictyostelium amoebae) of-ten exhibit unequal fitness (20, 48, 51, 52). Our results suggest thatsuch fitness inequalities may often be indirect byproducts of al-ternative evolutionary forces rather than the result of the winningstrains having undergone selection for increased within-groupcompetitiveness during the focal social process.Difficulties inherent to inferring what evolutionary forces have

shaped naturally evolved traits in general (53) apply with equalforce to social traits specifically. Our results highlight that thehypothesis that kin-discriminatory behaviors or other social in-teraction phenotypes originated by indirect processes should notbe excluded without compelling positive evidence that theyevolved primarily as directly selected social adaptations. Manyextant social interaction traits are likely to have been shaped bycomplex combinations of indirect causes and direct selection.

MethodsTo assess whether evolved populations discriminate between themselves andtheir ancestor (KD-A), samples from other evolutionary time-points from

the same population (KD-T), or other evolved populations from the sameevolutionary environment (KD-B), we staged colony-encounter assays asfollows. One day before each assay, cells were inoculated in liquid media andplates were prepared by pouring 20 mL of CTT soft agar (0.5% agar) into9-cm-diameter Petri dishes unless otherwise specified. To start each assay, 10 μLof each culture (previously adjusted to ∼5 × 109 cells per mL) were spotted1 cm apart from each other. In experimental assays testing for KD-A, onespot on a plate contained the ancestor and the other contained an evolvedpopulation. In experimental assays testing for KD-B or KD-T, the two spotson a plate were from distinct evolved populations or different evolutionarytime points within the same population, respectively. In control assays, twospots of the same genotype (or population sample) were tested for colonymerger. After spotting, culture samples were allowed to dry in a laminarflow hood and plates were then incubated for three days, after which col-onies were examined for the presence or absence of a clearly discernableline of demarcation between swarms.

For more detailed methods, see SI Appendix.

ACKNOWLEDGMENTS. We thank the Genomic Diversity Center (GDC) at ETHZürich and especially Jean-Claude Walser for his help in whole-genome se-quence analysis. This work was supported in part by National Institutes ofHealth Grant GM079690 (to G.J.V.) and an EMBO Long-Term Fellowship(to O.R.).

1. Sherman PW, Reeve HK, Pfennig DW (1997) Recognition systems. Behavioural Ecol-ogy: An Evolutionary Approach, eds Krebs JR, Davies NB (Blackwell Scientific Publi-cations, Oxford, UK), 4th Ed.

2. Wu HM, Holmes WG, Medina SR, Sackett GP (1980) Kin preference in infant Macacanemestrina. Nature 285(5762):225–227.

3. Buss LW, Grosberg RK (1990) Morphogenetic basis for phenotypic differences in hy-droid competitive behavior. Nature 343(6253):63–66.

4. Keller L, Ross KG (1998) Selfish genes: A green beard in the red fire ant. Nature394(6693):573–575.

5. Kuehne HA, Murphy HA, Francis CA, Sniegowski PD (2007) Allopatric divergence, sec-ondary contact, and genetic isolation in wild yeast populations. Curr Biol 17(5):407–411.

6. Ostrowski EA, Katoh M, Shaulsky G, Queller DC, Strassmann JE (2008) Kin discrimi-nation increases with genetic distance in a social amoeba. PLoS Biol 6(11):e287.

7. Strassmann JE, Gilbert OM, Queller DC (2011) Kin discrimination and cooperation inmicrobes. Annu Rev Microbiol 65:349–367.

8. Karban R, Shiojiri K, Ishizaki S, Wetzel WC, Evans RY (2013) Kin recognition affectsplant communication and defence. P R Soc B 280:20123062.

9. Brusini J, Robin C, Franc A (2013) To fuse or not to fuse? An evolutionary view of self-recognition systems. J Phylogen Evolution Biol 1:103.

10. Cadavid LF (2005) Self/non-self discrimination in basal metazoa: Genetics ofallorecognition in the hydroid Hydractinia. Integr Comp Biol 45(4):623–630.

11. Rousset F, Roze D (2007) Constraints on the origin and maintenance of genetic kinrecognition. Evolution 61(10):2320–2330.

12. Axelrod R, Hammond RA, Grafen A (2004) Altruism via kin-selection strategies thatrely on arbitrary tags with which they coevolve. Evolution 58(8):1833–1838.

13. Hamilton WD (1964) The genetical evolution of social behaviour. I. J Theor Biol 7(1):1–16.14. West SA, Griffin AS, Gardner A, Diggle SP (2006) Social evolution theory for micro-

organisms. Nat Rev Microbiol 4(8):597–607.15. Grosberg RK (1988) The evolution of allorecognition specificity in clonal inverteb-

rates. Q Rev Biol 63(4):377–412.16. Crozier RH (1986) Genetic clonal recognition abilities in marine invertebrates must be

maintained by selection for something else. Evolution 40(5):1100–1101.17. Grant V (1966) Selective origin of incompatibility barriers in plant genus Gilia. Am Nat

100(911):99–104.18. Pfennig DW, Collins JP, Ziemba RE (1999) A test of alternative hypotheses for kin

recognition in cannibalistic tiger salamanders. Behav Ecol 10:436–443.19. Smith J, Van Dyken JD, Velicer GJ (2014) Nonadaptive processes can create the ap-

pearance of facultative cheating in microbes. Evolution 68(3):816–826.20. Vos M, Velicer GJ (2009) Social conflict in centimeter-and global-scale populations of

the bacterium Myxococcus xanthus. Curr Biol 19(20):1763–1767.21. Dienes L (1946) Reproductive processes in Proteus cultures. Proc Soc Exp Biol Med

63(2):265–270.22. Munson EL, Pfaller MA, Doern GV (2002) Modification of dienes mutual inhibition

test for epidemiological characterization of Pseudomonas aeruginosa isolates. J ClinMicrobiol 40(11):4285–4288.

23. Ho HI, Hirose S, Kuspa A, Shaulsky G (2013) Kin recognition protects cooperatorsagainst cheaters. Curr Biol 23(16):1590–1595.

24. Rajagopalan R, Sarwar Z, Garza AG, Kroos L (2013) Developmental Gene Regulation.Myxobacteria: Genomics, Cellular and Molecular Biology, eds Yang Z, Higgs P (CaisterAcademic Press).

25. Velicer GJ, et al. (2006) Comprehensive mutation identification in an evolved bacterialcooperator and its cheating ancestor. Proc Natl Acad Sci USA 103(21):8107–8112.

26. Schofield MJ, Hsieh P (2003) DNA mismatch repair: Molecular mechanisms and bi-ological function. Annu Rev Microbiol 57:579–608.

27. Pathak DT, Wei X, Dey A, Wall D (2013) Molecular recognition by a polymorphic cellsurface receptor governs cooperative behaviors in bacteria. PLoS Genet 9(11):e1003891.

28. Benabentos R, et al. (2009) Polymorphic members of the lag gene family mediate kin

discrimination in Dictyostelium. Curr Biol 19(7):567–572.29. Gibbs KA, Urbanowski ML, Greenberg EP (2008) Genetic determinants of self identity

and social recognition in bacteria. Science 321(5886):256–259.30. Hirose S, Benabentos R, Ho HI, Kuspa A, Shaulsky G (2011) Self-recognition in social

amoebae is mediated by allelic pairs of tiger genes. Science 333(6041):467–470.31. Coyne JA, Orr HA (2004) Speciation (Sinauer Associates, Sunderland, MA).32. Presgraves DC (2010) The molecular evolutionary basis of species formation. Nat Rev

Genet 11(3):175–180.33. Unckless RL, Orr HA (2009) Dobzhansky-Muller incompatibilities and adaptation to a

shared environment. Heredity (Edinb) 102(3):214–217.34. Bomblies K, et al. (2007) Autoimmune response as a mechanism for a Dobzhansky-

Muller-type incompatibility syndrome in plants. PLoS Biol 5(9):e236.35. Pagel M (2009) Human language as a culturally transmitted replicator. Nat Rev Genet

10(6):405–415.36. Wright KM, Lloyd D, Lowry DB, Macnair MR, Willis JH (2013) Indirect evolution of hybrid

lethality due to linkage with selected locus in Mimulus guttatus. PLoS Biol 11(2):e1001497.37. Gilbert OM (2015) Histocompatibility as adaptive response to discriminatory within-

organism conflict: A historical model. Am Nat 185(2):228–242.38. Holman L, van Zweden JS, Linksvayer TA, d’Ettorre P (2013) Crozier’s paradox revisited:

Maintenance of genetic recognition systems by disassortative mating. BMC Evol Biol 13:211.39. Hughes RN, Manríquez PH, Morley S, Craig SF, Bishop JDD (2004) Kin or self-recog-

nition? Colonial fusibility of the bryozoan Celleporella hyalina. Evol Dev 6(6):431–437.40. Bégueret J, Turcq B, Clavé C (1994) Vegetative incompatibility in filamentous fungi:

Het genes begin to talk. Trends Genet 10(12):441–446.41. Dettman JR, Jacobson DJ, Turner E, Pringle A, Taylor JW (2003) Reproductive isolation

and phylogenetic divergence in Neurospora: Comparing methods of species recog-nition in a model eukaryote. Evolution 57(12):2721–2741.

42. Glass NL, Kuldau GA (1992) Mating type and vegetative incompatibility in filamen-

tous ascomycetes. Annu Rev Phytopathol 30:201–224.43. Saupe SJ (2000) Molecular genetics of heterokaryon incompatibility in filamentous

ascomycetes. Microbiol Mol Biol Rev 64(3):489–502.44. Noor MAF (1999) Reinforcement and other consequences of sympatry. Heredity

(Edinb) 83(Pt 5):503–508.45. Turner E, Jacobson DJ, Taylor JW (2011) Genetic architecture of a reinforced, post-

mating, reproductive isolation barrier between Neurospora species indicates evolu-

tion via natural selection. PLoS Genet 7(8):e1002204.46. Ackermann M, Chao L (2004) Evolution of cooperation: Two for one? Curr Biol 14(2):

R73–R74.47. Gould SJ, Vrba ES (1982) Exaptation - a missing term in the science of form. Paleo-

biology 8(1):4–15.48. Buttery NJ, Rozen DE, Wolf JB, Thompson CR (2009) Quantification of social behavior in D.

discoideum reveals complex fixed and facultative strategies. Curr Biol 19(16):1373–1377.49. Kraemer SA, Velicer GJ (2014) Social complementation and growth advantages pro-

mote socially defective bacterial isolates. P B Sci R Soc 281(1781):20140036.50. Velicer GJ, Kroos L, Lenski RE (2000) Developmental cheating in the social bacterium

Myxococcus xanthus. Nature 404(6778):598–601.51. Strassmann JE, Zhu Y, Queller DC (2000) Altruism and social cheating in the social

amoeba Dictyostelium discoideum. Nature 408(6815):965–967.52. Rendueles O, Amherd M, Velicer GJ (2015) Positively frequency-dependent interference

competition maintains diversity and pervades a natural population of cooperative mi-

crobes. Curr Biol, 10.1016/j.cub.2015.04.057.53. Williams GC (1966) Adaptation and Natural Selection. A Critique of Some Current

Evolutionary Thought (Princeton University Press, Princeton, NJ).

Rendueles et al. PNAS | July 21, 2015 | vol. 112 | no. 29 | 9081

EVOLU

TION

 

SUPPLEMENTARY INFORMATION

SUPPLEMENTARY FIGURES

Figure S1. Colony-encounter assay. A. Initially separate colonies (solid circles) grow and

swarm outward (arrows) and freely merge (as depicted here) in the absence of kin

discrimination barriers. B. Self-self encounter control of oncoming colonies of the same

natural isolate (A47) that differ only in their antibiotic-resistance marker. ‘K’ and ‘R’

indicate kanamycin- and rifampicin-resistance marked variants of A47, respectively. No

interface demarcation line between colonies was visible for any self-self encounters.

Dark spots are individual fruiting bodies.

1

 

Figure S2. Experimental evolution. Independent clones from differentially marked

ancestor variants (either rifampicin sensitive (GJV1) or resistant (GJV2)) were grown in

liquid and used to start experimental evolution populations. A total of 104 independent

populations were established across twelve different agar-plate environments (see Table S2).

Each population was allowed to grow and swarm outward for two weeks at 32 °C and 90%

rH. At two-week intervals, a small rectangle (~3 mm x 5 mm) was cut out from the point

along the swarm perimeter furthest from the colony center (or from a random point if no

deviation from circularity was evident) and placed upside-down on the center of a new plate.

This process was repeated for either 18 or 40 cycles, depending on the environmental

treatment.

2

Figure S3. Kin discrimination patterns between temporal samples from the same

population (KD-T). Colony-encounter phenotypes were examined between time-point

samples from within the same population. Colors represent classifications described in Fig. 4.

Axis numbers represent evolutionary time in selection cycles. Data for P10 and P35 are not

shown but are included in Fig. 5. “Gradual” indicates that the probability of incompatibility is

largely a function of evolutionary time whereas “abrupt” indicates a relatively discrete

transition between compatibility states. “E.c.” indicates E. coli and “B.s.” indicates B. subtilis.

3

 

Figure S4. Venn diagram summarizing all mutated loci in clones from 23 evolved

populations. Mutations analyzed from genome sequences of clones from 23 populations

(excluding the clone from P29), sixteen of which displayed KD towards the ancestor (KD-A)

and seven of which did not (nKD-A). Superscripts indicate the number of independently

evolved populations that accumulated a mutation at the specified locus (i.e. instances of

convergent evolution), whereas loci without superscripts depict singleton mutations unique to

one population. A total of 173 independent mutations are represented, the majority of which

(134) occurred in loci that were mutated in only one population (45 for nKD-A populations

and 89 for KD populations). Eighteen loci were mutated in clones from both KD-A and nKD-

A populations. Loci are identified either by their published gene names or respective “MXAN”

locus tag number in case of uncharacterized loci.

4

SUPPLEMENTARY TABLES

Table S1. Fitness effects of forced mixing at a 1:1 ratio for two pairs of antagonistic natural isolates during multicellular development.

Paired two-tailed t tests were performed to compare appropriate parameters. Ni, total spores produced in pure culture (log10) per 5 x 108 initial

cells; Ni (j), spores produced by i in forced mixture with j per 5 x 108 initial cells; W*ij : Log10-difference in pure-culture spore production of

strains i and j; Wij , relative fitness, or the observed difference in (log10) spore production of strains i and j during forced mixing. Values

presented are the average of at least three independent replicates and corresponding upper and lower bounds of the 95% confidence intervals.

Forced Mix Clone Ni W*ij Ni (j) Wij Ni vs. Ni(j) W*ij vs. Wij

A47 vs. A23 A47 7.91 ± 0.36 0.42 ± 0.34 7.45 ± 0.43 3.42 ± 1.38 p = 0.064 p = 0.015

A23 7.49 ± 0.42 -0.42 ± 0.34 4.02 ± 1.48 -3.42 ± 1.38 p = 0.009

A47 vs. A96 A47 7.75 ± 0.19 0.21 ± 0.23 6.85 ± 0.12 2.19 ± 0.85 p = 0.043 p = 0.032

A96 7.71 ± 0.36 -0.21 ± 0.23 4.66 ± 0.74 -2.19 ± 0.85 p = 0.009

5

Table S2.   Experimental evolution treatments, populations and patterns of KD-A evolution. Odd-numbered populations descend from the

rifampicin-sensitive ancestor GJV1 and the even-numbered populations descend from the rifampicin-resistant ancestor GJV2. Bold blue text indicates

terminal populations that evolved a clear and consistent KD-A phenotype (55 populations), italicized black text indicates inconsistent KD-A

phenotypes (seven populations) and standard black text indicates the absence of a KD-A phenotype (26 populations). *Asterisks indicate ancestral sub-

clones found to have a shared mutation in cmr4 (cmr4-P72H, see main text and Supplemental Results).

Treatment Description Cycles Examined populations (organized by ancestral GJV1 or GJV2 sub-clone)

GJV 1.1*

GJV 2.1

GJV 1.2

GJV 2.2

GJV 1.3*

GJV 2.3

GJV 1.4

GJV 2.4

GJV 1.5*

GJV 2.5

GJV 1.6*

GJV 2.6

CTT hard agar (HA) 1% Casitone, 1.5% agar 40

CTT soft agar (SA) 1% Casitone, 0.5% agar 40

Low nutrient CTT HA 0.1% Casitone, 1.5% agar 40

Low nutrient CTT SA 0.1% Casitone, 0.5% agar 40

CTT HA E. coli E. coli grown on CTT HA 40

P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12

P29 P30 P31 P32 P33 P34 P35 P36 P37 P38 P39 P40

P57 P58 P59 P60 P61 P62 P63 P64

P65 P66 P67 P68

P89 P90 P91 P92 P93 P94 P95 P96

CTT HA B. subtilis B. subtilis grown on CTT HA 40 P97 P98 P100 P102 P103 P104

CTT SA E. coli E. coli grown on CTT SA 40 P105 P106 P107 P108 P110 P112

CTT SA B. subtilis B. subtilis grown on CTT SA 40 P113 P115 P116 P117 P118 P119

TPM HA E. coli E. coli overlaid on TPM HA 18 P121 P122 P123 P124 P125 P126 P127 P128

TPM HA B. subtilis B. subtilis overlaid on TPM HA 18 P129 P130 P131 P132 P133 P134 P135 P136

TPM SA E. coli E. coli overlaid on TPM SA 18 P137 P138 P140 P142 P144

TPM SA B. subtilis B. subtilis overlaid on TPM SA 18 P145 P146 P148 P149 P151

6

Table S3. Kin discrimination between independent replicate populations from the same evolutionary treatment (KD-B).

Treatment Examined pairs (N) KD-B % KD-B

CTT HA 28 15 54

CTT SA 22 15 69

0.1% CTT HA 12 7 59

0.1% CTT SA 3 2 67

CTT HA E. coli 8 4 50

CTT HA B. subtilis 6 3 50

CTT SA E. coli 6 6 100

CTT SA B. subtilis 10 3 30

TPM HA E. coli 10 2 20

TPM HA B. subtilis 1 1 100

TPM SA E. coli 7 1 15

TPM SA B. subtilis 4 3 75

Total 117 62 53

7

Table S4. Numbers of mutations accumulated in evolved clones after 40 two-week cycles. The genomes of single clones from each of 24

distinct populations were sequenced and the number of mutations in total and in four categories are shown.

Population KD-A

phenotype Evolutionary

treatment Total # of mutations Genic Coding Synonymous Intergenic

P1 + CTT HA 13 12 11 1 1 P2 - CTT HA 13 10 8 2 3 P3 + CTT HA 10 10 10 0 0 P4 - CTT HA 10 9 9 0 1 P5 - CTT HA 14 13 11 2 1 P6 + CTT HA 12 11 10 1 1 P7 + CTT HA 7 6 5 1 1 P8 + CTT HA 9 9 8 1 0 P9 + CTT HA 10 10 9 1 0

P10 + CTT HA 10 10 7 3 0 P11 + CTT HA 13 11 10 1 2 P12 - CTT HA 19 15 14 1 4 P29 + CTT SA 435 373 ND ND 62 P30 + CTT SA 13 13 10 3 0 P31 Undefined CTT SA 16 15 12 3 1 P32 Undefined CTT SA 12 12 11 1 0 P33 + CTT SA 11 10 9 1 1 P34 + CTT SA 13 13 11 2 0 P35 + CTT SA 23 21 19 2 2 P36 + CTT SA 14 14 13 1 0 P37 + CTT SA 16 14 13 1 2 P38 - CTT SA 10 10 9 1 0 P39 + CTT SA 17 15 13 2 2 P40 + CTT SA 13 12 11 1 1

8

Table S5. Genes mutated in more than two populations. Seventeen of the sequenced clones from 24 populations exhibited clear and

consistent KD-A (71%) and seven did not. The third through sixth columns show the number of populations mutated at the respective locus, the

number and percentage of those populations that show a KD-A phenotype and the number of distinct mutation sites, respectively.

Mutated locus Description No.

mutated clones

No. in KD-A clones

% KD-A

No. distinct

mutations Mutated populations

MXAN_6012 response regulator 17 11 65 10 P1, P2, P3, P5, P7, P8, P12, P30, P31, P32, P33, P34, P35, P36, P38, P39, P40

frzF protein methyltransferase FrzF 17 13 76 13 P1, P3, P5, P8, P9, P10, P11, P29, P30, P31, P32, P33, P35, P37, P38, P39, P40

MXAN_5852 sensory box histidine kinase 9 5 56 6 P5, P30, P31, P32, P33, P35, P36, P37, P38 cmr4 CRISPR‑associated RAMP protein Cmr4 8 8 100 1 P1, P7, P9, P11, P29, P35, P37, P39

MXAN_5032 efflux transporter, HAE1 family, inner membrane component 6 4 67 6 P5, P34, P35, P36, P37, P38 MXAN_4798 hypothetical protein 5 3 60 4 P10, P30, P31, P32, P36

hsfB response regulator/sensor histidine kinase HsfB 5 3 60 5 P1, P5, P9, P12, P36 MXAN_7214 RNA polymerase sigma‑70 factor, ECF subfamily 5 2 40 5 P1, P2, P6, P32, P38 MXAN_7216 ICE‑like protease (caspase) p20 domain protein 4 4 100 4 P3, P7, P9, P11

MXAN_5030 efflux transporter, HAE1 family, outer membrane effluxprotein 4 2 50 3 P31, P32, P33, P40

MXAN_0289 putative membrane protein 4 2 50 4 P3, P5, P12, P40 MXAN_6704 acetyltransferase, GNAT family 4 4 100 4 P6, P30, P34, P39

lon ATP‑dependent protease La 4 4 100 5 P29, P33, P35, P37 frzCD frizzy aggregation protein FrzCD 4 2 50 4 P2, P7, P12, P29 rpoC DNA‑directed RNA polymerase subunit beta 3 2 67 3 P32, P34, P35

MXAN_5031 HAE1 family efflux transporter MFP subunit 3 3 100 3 P9, P30, P39 frzE gliding motility regulatory protein 3 2 67 3 P6, P12, P36

MXAN_3952 sigma‑54 dependent transcriptional regulator, Fis family 3 1 33 1 P4, P12, P37

9

Table S6. Mutations that first appeared in three independently evolved populations at the same transfer cycle that KD towards the ancestor (KD-A) first appeared.

Population Gene Function

P10

MXAN_1574 TfoX domain protein MXAN_4006 peptidase, S1C (protease Do) subfamily

agmK adventurous gliding motility protein AgmK

MXAN_5837 bacterial Ig-like domain (group 1)/fibronectin type III domain protein

P33 MXAN_5852 sensory box histidine kinase lon ATP-dependent protease La

P34 rpsB ribosomal protein S2 MXAN_6704 acetyltransferase, GNAT family

10

METHODS

Semantics. We adopt a broad definition of kin discrimination, namely any ‘alteration of

social behavior as a function of genetic relatedness among interactants’. This definition is

merely phenomenological and is thus entirely decoupled from the evolutionary, behavioral

and molecular causes of kin discrimination traits, whatever those may be. Such causes must

be determined independently of the mere demonstration that kin discrimination, as defined

here, occurs in any given biological system. Thus, the definition encompasses both social

adaptations per se and indirect byproducts of non-adaptive processes (or alternative adaptive

processes) at the level of evolutionary causation, as well as organismal behaviors and

molecular mechanisms of all degrees of complexity.

We define ‘colony-merger incompatibility’ as a reduced degree of merger by oncoming

colonies each composed of a distinct genotype into one social group, relative to oncoming

colonies that are composed of the same genotype. Colony-merger incompatibilities are

phenotypically variable rather than a single discrete phenomenon, and can result in

phenotypes ranging from a complete absence of contact between the cells at the leading edge

of colony expansion to a slight but detectable reduction in the degree of colony inter-

penetration relative to controls.

Strains, experimental evolution and growth conditions.

(i) Experimental evolution. Parallel evolving populations were initiated from

independently isolated sub-clones of the two ancestral strains GJV1 (1), a rifampicin-

sensitive clonal derivative of DK1622 (2) and GJV2, a rifampicin-resistant clonal derivative

of GJV1. Six sub-clones each of GJV1 and GJV2 (GJV1.1 – GJV1.6 and GJV2.1 – GJV2.6)

were stored frozen. All twelve sub-clones were used to initiate evolutionary the CTT hard

and soft agar treatments (populations P1 - P12 and P29 - P40) whereas GJV1.1 – GJV1.4 and

11

GJV2.1 – GJV2.4 were used to initiate all other treatments (Table S2). Odd and even

numbered populations derived from GJV1 and GJV2 sub-clones, respectively. The

numerical order of population designators corresponds with the numerical order of sub-clone

designators. For example, sub-clones GJV1.1, GJV2.1, GJV1.2, GJV2.2, GJV1.3, GJV2.3,

etc. founded populations P1, P2, P3, P4, P5, P6, etc.; P29, P30, P31, P32, P33, P34, etc. and

so on (Table S2).

All evolving populations were initiated and propagated as described previously (3) (see

Fig. S2 and Table S2 for summary). Briefly, culture samples from ancestral clones (~5 x 107

cells in 10 µl) were placed in the center of agar plates (InvitrogenTM Select Agar) and allowed

to grow and swarm outward for two weeks at 32 °C and 90% rH. After two weeks, a sample

of ~3 mm x 5 mm (~15 mm2) from the leading edge of each swarming colony was harvested

with a sterile scalpel and transferred upside down at the center of a new plate. If a colony

was not circular at the time of transfer, the sample was taken from the point along the colony

edge farthest from the center. This transfer protocol was repeated every two weeks for either

18 or 40 cycles. Evolution experiments were carried out at the Max-Planck Institute for

Developmental Biology in Tübingen, Germany from 2001-2003. All post-evolution assays

were performed at ETH Zürich from 2012-2014 in at least three temporally independent

replicates.

(ii) Evolution environments. Experimental evolution was performed in twelve distinct

laboratory environments with either eight or twelve replicate populations each. Environments

varied in nutrient source (bacterial prey vs. media substrate), nutrient level, surface viscosity

(thus affecting motility evolution) and other parameters (Table S2).

Non-prey treatments: One day prior to colony transfer, 50 ml of either CTT (8 mM

MgSO4, 10 mM Tris pH 8.0, 10 g/L Casitone, 1 mM KPO4) or 0.1% Casitone CTT (identical

to CTT except with only 1 g/L Casitone) agar (hard or soft: 1.5% or 0.5% agar, respectively)

12

were poured into 14 cm-diameter petri dishes. Plates were allowed to solidify uncovered in a

laminar flow hood (for 15-20 minutes) and then stored overnight at room temperature.

Prey treatments: Two days prior to transfer, agar plates were poured as previously

described, including with TPM buffer, which is identical to CTT medium except without any

Casitone. The next day, prey (Bacillus subtilis PY79 (4) or Escherichia coli REL607 (5))

were inoculated into three flasks containing 900 ml of CTT liquid and grown overnight at 32

°C, 300 rpm. On the transfer day, 200 µL of grown prey were spread out on CTT plates for

CTT + prey environments. For TPM + prey plates, prey cultures were centrifuged in 500 mL

tubes at 4 °C, 10,000 rpm for 10 minutes. The supernatant was discarded and pellets were

resuspended in 3 mL of TPM by shaking at 400 rpm. Resuspended pellets of each prey type

were pooled and 1 mL of prey suspension was spread out over the entire surface of TPM

plates and allowed to dry. Cycles transfers were then performed as described above. Plates

were kept upright for one night and then turned upside down. (iii) Growth conditions. All

evolution and post-evolution agar-plate cultures were incubated at 32 °C, 90% rH. For post-

evolution experiments, cultures initiated from frozen stocks were grown on CTT hard agar

plates for three or four days. Prior to the start of each experiment, culture samples were

transferred from plates to 8 mL CTT liquid (in 50 mL flasks) for 24 hours at 32°C with

constant shaking at 300 rpm until they reached mid exponential phase (OD600 = ~0.5).

Colony-encounter assays. (Some text in this section is replicated from Methods for clarity.)

To assess whether evolved populations discriminate between themselves and their ancestor

(KD-A), samples from other evolutionary time-points from the same population (KD-T) or

other evolved populations from the same evolutionary environment (KD-B), we staged

colony-encounter assays as follows. One day prior to each assay, cells were inoculated in

liquid media and plates were prepared by pouring 20 mL CTT soft agar (0.5% agar) into 9

13

cm-diameter petri dishes unless otherwise specified. Plates were allowed to solidify

uncovered in a laminar flow hood (for 15-20 minutes) and stored overnight at room

temperature. To start each assay, 10 µL of each culture (previously adjusted to ~5 x 109

cells/mL) were spotted 1 cm apart from each other. In experimental assays testing for KD-A

(see main text), one spot on a plate contained the ancestor and the other contained an evolved

population. In experimental assays testing for between-population KD-B or KD-T, the two

spots on a plate were from distinct evolved populations or different evolutionary time-points

within the same population, respectively. In control assays, two spots of the same genotype

(or population sample) were tested for colony merger. Such self-self encounter controls were

performed for all assayed strains and populations simultaneously with experimental

treatments. After spotting, culture samples were allowed to dry in a laminar flow hood and

plates were then incubated for three days, after which colonies were examined for the

presence or absence of a clearly discernable line of demarcation between swarms. Colony-

interface phenotypes were photographed and classified into four qualitative categories: a)

freely merging (no visually detectable difference from self-self encounter controls of the

ancestor, green in Figs. 4, 5 and S3), b) consistently reduced merger of intermediate

phenotypic strength (light red), c) consistently and greatly reduced merger or complete non-

merger (red), and iv) inconsistent phenotypes across replicate experiments (grey). Such

inconsistencies may have been due to differential behavior of genetically heterogeneous

populations across experimental replicates. The small minority of pairings that yielded highly

inconsistent results was excluded from statistical analysis. Colony pairings not performed, or

not analyzed due to contamination events are represented by white matrix cells in Figs. 4 and

S3.

14

Fruiting body chimerism assays.

To test whether colony-merger incompatibilities reduce developmental co-aggregation of

distinct genotypes along fruiting bodies near the inter-colony borders, we assessed the

frequency of chimerism across fruiting bodies. Cultures were prepared as described above. A

day prior to the assay, plates were prepared by pouring 9 mL of CF agar (a low-nutrient

medium that allows some growth and swarming before development is initiated upon nutrient

depletion; ref. 6; 10 mM Tris pH 8.0, 1 mM KH2PO4, 8 mM MgSO4, 0.02% (NH4)2SO4,

0.2% citrate, 0.1% pyruvate, 150 mg/L Casitone) into 5 cm-diameter petri dishes. Ten

microliters of each culture adjusted to ~5 x 109 cells/mL were spotted one centimeter apart

from each other, allowed to dry in a laminar flow hood and incubated at 32 ºC, 90% rH.

After six days, four to eight individual fruiting bodies adjacent to the interface of oncoming

swarms were harvested and their respective locations relative to the interface documented for

each. Individual fruiting bodies were incubated in 500 µL sterile ddH2O at 50 ºC for two

hours to select for viable, heat-resistant spores. Samples were sonicated with a sterile

microtip, diluted in sterile ddH2O and plated in CTT soft (0.5%) agar containing the

appropriate antibiotics. All mixes were performed at least three times in temporally

independent blocks. Experimental treatments and controls were performed simultaneously

with each mixed culture assay. M. xanthus natural clones A23, A47 and A96 were isolated

from a 16 x 16 cm soil plot in Tübingen, Germany as described previously (7). Rifampicin-

and kanamycin-resistant variants of natural isolates (specified by ‘R’ and ‘K’, respectively, in

Figs. 1 and S1) were previously characterized (8) and had no significant defects in pure

culture spore production.

15

Whole-genome sequencing and mutation identification.

Genomic DNA (>30 µg) was extracted from exponentially growing cells using Qiagen’s

Genomic DNA Isolation Kit and 100G Genomic-Tip. Illumina HiSeq sequencing was

performed by BGI Tech Solutions Co., Ltd. (Hong Kong, China) and yielded a total of ~109

bp of sequence per genome, or ~100-fold average coverage. Small genomic changes were

identified for each genome by mapping reads against the reference genome of M. xanthus

DK1622 (NC_008095.1) using breseq v0.21 (9). Five known SNPs between the DK1622

derivative used as the ancestor in these studies (GJV1 (1)) and the published sequence of

DK1622 (2) as well as a known rpoB mutation in GJV2-derived (rifampicin-resistant)

populations served as positive controls for the reliability of mutation identification. All five

polymorphisms shared by all evolved clones were detected in all 24 sequenced genomes and

the rpoB mutation was detected in all twelve of the rifampicin-resistant clones. Additionally,

a subset of polymorphisms in evolved clones was checked by sequencing PCR products.

SUPPLEMENTARY RESULTS

While several genes evolved convergently (at the gene level) among the 24 genome-

sequenced clones, only one locus, the CRISPR-associated gene cmr4, was mutated

exclusively in KD-A populations (Tables S2 and S5). CRISPR genes protect bacterial cells

against mobile genetic elements such as viruses and plasmids and also regulate social traits

such as virulence and biofilm formation in Campylobacter jejuni and Pseudomonas

aeruginosa, respectively, as well as multicellular development in M. xanthus (10, 11).

Surprisingly, the same nonsynonymous mutation in cmr4 (amino acid substitution P72H) was

found in all populations mutated at this locus, whereas for all other loci that were mutated in

four or more populations, the precise mutations in those genes differed across most

16

populations (Table S5). This pattern suggested that the cmr4 mutation might be causally

related to the evolution of kin discrimination in at least some populations.

To investigate this possibility further, all populations were tested for cmr4 mutations at

their terminal time point. The cmr4 mutation P72H was found in approximately one third of

all populations, but during the process of tracing the temporal origins of this mutation we

noted that it was only present in populations derived from four specific ancestral sub-clones

and not in populations derived from the other eight sub-clones. We thus hypothesized and

subsequently confirmed that this mutation was present in four sub-clone ancestors used to

initiate experimental populations (GJV1.1, GJV1.4, GJV1.5 and GJV1.6, see Table S2), but

was not present in GJV1.2, GJV1.3 or in any of the six GJV2 sub-clones. This finding

implies that the source culture of GJV1 used to isolate the GJV1.1-1.6 sub-clones was

polymorphic for this mutation prior to sub-clone selection and the onset of experimental

evolution.

Our tracing of the cmr4-P72H mutation to a subset of ancestral sub-clones is of interest

for both methodological and evolutionary reasons. Methodologically, for microbial

experimental evolution studies this finding highlights the importance of initiating replicate

populations of a clonal ancestral strain from independently isolated sub-clones of that strain

that are themselves stored frozen for future reference. Given that all microbial cultures

larger than a few thousand individuals are expected to be polymorphic due to spontaneous

mutation (12), curation of ancestral sub-clones allows determination of whether identical

mutations found in distinct replicate populations evolved convergently or not.

Evolutionarily, among the 55 populations that evolved clear and consistent KD-A

phenotypes, the proportion that derived from a sub-clone ancestor that carried the cmr4-P72H

mutation (20/55, 36%) is improbable under the null hypothesis that the ancestral presence of

this mutation and the probability of evolving KD-A are causally unrelated, relative to the

17

much lower frequency of ancestral cmr4-P72H carriage among the remaining populations

(4/29, 14%; two-tailed p = 0.042, Fisher’s exact test). This outcome suggests the hypothesis

that carriage of cmr4-P72H causally increases the probability of evolving of KD-A relative

the absence of this mutation (even though KD-A did evolve in many populations not carrying

this mutation).

A second evolutionary pattern is also statistically associated with the ancestral presence

of cmr4-P72H. Among the 18 populations for which KD-T phenotype patterns were analyzed

temporally, seven exhibited relatively gradual KD-T emergence patterns whereas nine

showed abrupt KD-T appearance and two showed an intermediate pattern (Fig. 4 and Figs.

S2 and S3). Of these 18 populations, all six that carried the ancestral crm4 mutation (P1,

P29, P35, P65 and P119) exhibited gradual KD-T emergence, an outcome with very low

probability under the null expectation that carriage of cmr4-P72H does not affect the

temporal pattern of KD evolution (two-tailed p = 0.0009, Fisher’s exact test with the two

intermediate-pattern populations excluded). The absence of cmr4-P72H in one gradual-KD-

T population shows that this pattern can evolve without the mutation, but the overall

distribution of KD-T evolution patterns suggests that carriage of cmr4-P72H may promote

the gradual evolution of KD-T rather than its abrupt appearance.

18

REFERENCES

1. Velicer GJ, et al. (2006) Comprehensive mutation identification in an evolved bacterial cooperator and its cheating ancestor Proc Natl Acad Sci USA 103(21): 8107–8112.

2. Goldman BS, et al. (2006) Evolution of sensory complexity recorded in a myxobacterial genome. Proc Natl Acad Sci USA 103(41):15200-15205.

3. Velicer GJ & Yu YT (2003) Evolution of novel cooperative swarming in the bacterium Myxococcus xanthus. Nature 425(6953):75-78.

4. Youngman P, Perkins JB, & Losick R (1984) Construction of a cloning site near one end of Tn917 into which foreign DNA may be inserted without affecting transposition in Bacillus subtilis or expression of the transposon-borne Erm gene. Plasmid 12(1):1-9.

5. Lenski RE, Rose MR, Simpson SC, & Tadler SC (1991) Long-term experimental evolution in Escherichia coli. I. Adaptation and divergence during 2,000 generations. Am Nat 138(6):1315-1341.

6. Hagen DC, Bretscher AP, & Kaiser D (1978) Synergism between morphogenetic mutants of Myxococcus xanthus. Dev Biol 64(2):284-296.

7. Vos M & Velicer GJ (2006) Genetic population structure of the soil bacterium Myxococcus xanthus at the centimeter scale. Appl Environ Microbiol 72(5):3615-3625.

8. Vos M & Velicer GJ (2009) Social conflict in centimeter-and global-scale populations of the bacterium Myxococcus xanthus. Curr Biol 19(20):1763-1767.

9. Deatherage DE & Barrick JE (2014) Identification of mutations in laboratory-evolved microbes from next-generation sequencing data using breseq. Methods Mol Biol 1151:165-188.

10. Viswanathan P, Murphy K, Julien B, Garza AG, & Kroos L (2007) Regulation of dev, an operon that includes genes essential for Myxococcus xanthus development and CRISPR-associated genes and repeats. J Bacteriol 189(10):3738-3750.

11. Westra ER, Buckling A, & Fineran PC (2014) CRISPR-Cas systems: beyond adaptive immunity. Nature Rev Microbiol 12(5):317-326.

12. Bromham L (2009) Why do species vary in their rate of molecular evolution? Biol Letters 5(3):401-404.

   

19