Corrections

CELL BIOLOGY. For the article ‘‘The NF1 tumor suppressor criti- BIOCHEMISTRY. For the article ‘‘Engineered single-chain dimeric cally regulates TSC2 and mTOR,’’ by Cory M. Johannessen, streptavidins with an unexpected strong preference for biotin- Elizabeth E. Reczek, Marianne F. James, Hilde Brems, Eric 4-fluorescein,’’ by Filiz M. Aslan, Yong Yu, Scott C. Mohr, and Legius, and Karen Cichowski, which appeared in issue 24, June Charles R. Cantor, which appeared in issue 24, June 14, 2005, of 14, 2005, of Proc. Natl. Acad. Sci. USA (102, 8573–8578; first Proc. Natl. Acad. Sci. USA (102, 8507–8512; first published June published June 3, 2005; 10.1073͞pnas.0503224102), the authors 6, 2005; 10.1073͞pnas.0503112102), the footnotes appeared in note the following. On page 8574, the last sentence of the first the wrong order, due to a printer’s error. The ** footnote on paragraph in the left column, ‘‘Clarified lysates were normalized page 8507 should have read: ‘‘Throughout this paper, we use the for protein levels and analyzed by Western blotting with the terms monomer(ic), dimer(ic), and tetramer(ic) to refer to the following antibodies: phospho-p70S6K (T-389), phospho- number of biotin-binding domains present in a molecule, re- tuberin (S-939), phospho-tuberin (T-1462), phospho-Akt (S- gardless of the number of polypeptide chains it has.’’ In addition, 473), tuberin (C-20) from Santa Cruz Biotechnology, and the †† footnote on page 8512 should have read: ‘‘We note, protein kinase B␣͞AKT1, actin, and ␤-tubulin from Sigma- however, that extremely high-affinity mutants might not be Aldrich,’’ should read: ‘‘Clarified lysates were normalized for recovered in this procedure due to near-irreversible binding of protein levels and analyzed by Western blotting with the follow- biotin from the culture medium.’’ ing antibodies: phospho-p70S6K (T-389), phospho-tuberin (S- ͞ ͞ ͞ ͞ 939), phospho-tuberin (T-1462), phospho-Akt (S-473) from Cell www.pnas.org cgi doi 10.1073 pnas.0508348102 Signaling Technologies, tuberin (C-20) from Santa Cruz Bio- technology, and protein kinase B␣͞AKT1, actin, and ␤-tubulin from Sigma-Aldrich.’’ On page 8574, the second sentence of the PERSPECTIVE. For the article ‘‘Eusociality: Origin and conse- second paragraph in the left column, ‘‘For lentiviral production, quences,’’ by Edward O. Wilson and Bert Ho¨lldobler, which HEK293T cells were transfected with the ⌬8.2 lentiviral con- appeared in issue 38, September 20, 2005, of Proc. Natl. Acad. Sci. USA (102, 13367–13371; first published September 12, 2005; struct (encoding gag, pol, rev), VSVG, and either empty pKO ͞ vector or the pKO vector containing the following sequence 10.1073 pnas.0505858102), the authors note that Donat Agosti transcribing short-hairpin RNAs specific to NF1: 5Ј-TTATA- was omitted from the list of editors in ref. 48. The corrected reference appears below. AATAGCCTGGAAAAGG-3Ј,’’ should read: ‘‘For lentiviral CORRECTIONS production, HEK293T cells were transfected with the ⌬8.2 lentiviral construct (encoding gag, pol, rev), VSVG, and either 48. Ward, P. S. (2000) in Ants: Standard Methods for Measuring and Monitoring empty pLKO.1puro vector or the pLKO.1puro vector containing Biodiversity, eds. Agosti, D., Majer, J. D., Alonso, L. & Schultz, T. R. the following sequence transcribing short-hairpin RNAs specific (Smithsonian Institution Press, Washington, DC), pp. 99–121. Ј Ј to NF1: 5 -TTATAAATAGCCTGGAAAAGG-3 .’’ In addi- www.pnas.org͞cgi͞doi͞10.1073͞pnas.0508220102 tion, the authors would like to add the following statement to the Acknowledgments: ‘‘We would also like to thank Dr. William C. Hahn (Harvard Medical School and Dana–Farber Cancer In- stitute, Boston) and the Broad Institute (Cambridge, MA) for the pLKO.1 lentiviral constructs.’’

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0508351102

www.pnas.org PNAS ͉ November 1, 2005 ͉ vol. 102 ͉ no. 44 ͉ 16119–16120 Downloaded by guest on September 27, 2021 ANTHROPOLOGY. For the article ‘‘Descent of the larynx in chim- panzee infants,’’ by Takeshi Nishimura, Akichika Mikami, Juri Suzuki, and Tetsuro Matsuzawa, which appeared in issue 12, June 10, 2003, of Proc. Natl. Acad. Sci. USA (100, 6930–6933; first published May 29, 2003; 10.1073͞pnas.1231107100), the authors note that the symbols for Cleo and Pal were switched in the legend for Fig. 2. The figure and its corrected legend appear below. These errors do not alter the conclusions of the article.

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0508487102

Fig. 2. SVT growth in the three chimpanzee infants: Ayumu (male, ࡗ), Cleo (female, ■), and Pal (female, Œ). (a) Growth of the oral (solid line) and oropharyngeal (dotted line) components of the SVTH.(b) Growth of the SVTH (solid line) and SVTV (dotted line). (c) Age-related changes in the ratio of SVTH to SVTV lengths.

16120 ͉ www.pnas.org Downloaded by guest on September 27, 2021 PERSPECTIVE

Eusociality: Origin and consequences

Edward O. Wilson*† and Bert Ho¨ lldobler‡§ *Museum of Comparative Zoology, Harvard University, 26 Oxford Street, Cambridge, MA 02138-2902; ‡School of Life Sciences-LSC 274, Arizona State University, Tempe, AZ 85287-4501; and §Theodor-Boveri-Institut fu¨ r Biowissenschaften (Biozentrum), Universita¨t Wu¨ rzburg, Lehrstuhl fu¨ r Zoologie II, Am Hubland, D97074 Wu¨ rzburg, Germany

Contributed by Edward O. Wilson, July 12, 2005

In this new assessment of the empirical evidence, an alternative to the standard model is proposed: group selection is the strong binding force in eusocial evolution; individual selection, the strong dissolutive force; and kin selection (narrowly defined), either a weak binding or weak dissolutive force, according to circumstance. Close kinship may be more a consequence of eusociality than a factor promoting its origin. A point of no return to the solitary state exists, as a rule when workers become anatomically differenti- ated. Eusociality has been rare in evolution, evidently due to the scarcity of environmental pressures adequate to tip the balance among countervailing forces in favor of group selection. Eusociality in ants and termites in the irreversible stage is the key to their ecological dominance and has (at least in ants) shaped some features of internal phylogeny. Their colonies are consistently superior to solitary and preeusocial competitors, due to the altruistic behavior among nestmates and their ability to organize coordinated ac- tion by pheromonal communication.

n eusociality, an evolutionarily ad- of groups will spread if the positive in- theory are difficult to relate to the com- vanced level of colonial existence, tergroup component of the altruists’ fit- plexities of tangible social phenomena. adult colonial members belong to nesses exceeds the negative within-group We suggest that each force can be two or more overlapping genera- component of the altruists’ fitnesses heuristically classified as binding or dis- I solutive in its effect on colony cohesion tions, care cooperatively for the young, (4–6). Altruism is defined as behavior and are divided into reproductive and that benefits others at the cost of the and either strong or weak in its relative nonreproductive (or at least less-repro- lifetime production of offspring by the power. The degree of relatedness, the ductive) castes. The phenomenon is well altruist. The forces that determine this similarity across the whole genome of marked and nearly confined to insects, outcome are group selection, the differ- individuals as a result of recent common especially ants, bees, wasps, and ter- ential survival and reproduction of en- ancestry, is a factor that biases the di- mites, where it has been subject to a tire cooperative groups as a result of the rection and strength of the forces. When large body of mostly specialized research frequency of the eusociality alleles in elevated, say by lower individual dis- scattered across disciplines from genet- each; individual direct selection, accru- persal rates, relatedness can bring alleles ics to paleontology. It has long been the ing from the differential personal sur- for presociality and eusociality together conviction of researchers on social in- vival and reproduction of each of the more quickly. If it brings such genes to- sects that common principles exist at the colony members; and kin (indirect) se- gether more frequently due to shared organismic and superorganismic levels, lection, which we define here in the microhabitat preference, mate choice, or thus between individual insects and the original narrow sense as the favoring or other biological traits, it can have the tightly integrated colonies they compose disfavoring by individuals of collateral same effect. Relatedness can also in- (1, 2). Parallels have been persuasively relatives, i.e., relatives other than per- crease variance in presocial and eusocial drawn between the self-construction of sonal offspring. The inclusive fitness of alleles among groups, thus quickening organisms from molecules and tissues the prescribing genotype, of individual the pace of colony selection. But relat- and that of superorganisms from inter- colony members and hence statistically edness is relevant only insofar as it acting entire organisms (3). The princi- the colonies they compose, is the nonad- affects the frequency of alleles that pre- ples can be further parsed into two ditive product of the three forces. The scribe social behavior. Eusociality arises segments of the time scale: the develop- definition of kin selection in the com- by the superiority of organized groups mental decision rules that assemble or- monly used broad sense, which folds in over solitaires and cooperative preeuso- ganisms and colonies in each generation individual direct selection as well, pro- cial groups. It can, in theory at least, be and the origin of these rules through duces results that are consistent with initiated by group selection in either evolutionary time. those from the narrow sense but intu- the presence or absence of close related- Focusing here on the second princi- itively less clear and less practicable in ness and, when close relatedness exists, ple, evolutionary process, we suggest application. It also leads to such con- also in the presence or absence of kin how three seemingly disparate evolu- founding statements as ‘‘solitary species selection. Conversely, eusociality cannot tionary phenomena can be causally evolve by kin selection.’’ When the arise without the driving force of group linked: the selection forces that generate broad definition of kin selection is used selection, regardless of the degree of and shape eusociality, the rareness of (offspring included), colony selection relatedness within local populations or the origin of eusociality, and the ecolog- and kin selection are reduced to just cooperative aggregations. ical hegemony of eusocial insects. alternative ways of viewing change in Group selection which with the addi- the frequency of eusociality alleles in tion of cooperative behavior becomes The Forces of Natural Selection a population structured by relatedness colony selection, is the result of the Research during the past half century (4, 5). This formulation raises basic interaction of entire groups with their has incrementally clarified the nature of theory to a high level of mathematical environments (6, 7–11), conditioned by the collective forces that create and inclusiveness and abstraction, but it de- the efficiency of their internal opera- shape eusociality. At the most basic parts from the earlier and very heuristic level, an allele or ensemble of alleles narrow definition of kin selection; fur- †To whom correspondence should be addressed. E-mail: prescribing phenotypic plasticity that ther, the dynamical terms of the pop- [email protected]. includes self sacrifice of some members ulation genetic models composing the © 2005 by The National Academy of Sciences of the USA www.pnas.org͞cgi͞doi͞10.1073͞pnas.0505858102 PNAS ͉ September 20, 2005 ͉ vol. 102 ͉ no. 38 ͉ 13367–13371 tions (12). It promotes harmony among group members and genetic fitness of the group as a whole, once colony level is attained, and its effect is thereby in- herently binding. Individual direct selec- tion and kin selection play out in the social interactions of the colony mem- bers. By pitting individual members and nepotisms against one another, individ- ual direct selection tends to reduce col- ony harmony and genetic fitness. De- pending on circumstance, kin selection can be either binding or dissolutive, the latter by competition among nepotistic subgroups. In all of this multilevel the- ater of evolution, the gene remains the unit of selection, whereas the target of selection can be the individual, the group, or nepotistic alliances or other- wise biased actions within the group. The Point of No Return An abundance of evidence suggests that the strength of the biasing role of relat- edness within a species depends on the stage of its social evolution. The key transition occurs at a point in colony evolution that can be conveniently Fig. 1. The two competing hypotheses of the origin of eusociality in insects and thence before the point called the point of no return (13). of no return. The first (A) holds that in the earliest stage, kin selection is binding, making close relatedness a key feature; if combined with group selection, kin selection favors primitively eusocial colonies in a Beyond this level, it is impossible, or population of solitary or preeusocial insects (far left). The second hypothesis (B), newly presented here, at least difficult and uncommon, for a holds in contrast that group selection is paramount as the binding force, and kin selection is minor or species to regress from the eusocial to absent as a binding force for the group as a whole and weakly dissolutive if it forms competing groups. a more primitively eusocial, presocial, Relatedness, in hypothesis B, is increased as group selection cleanses the dissolutive nepotisms effect of kin or solitary condition. selection. The empirical evidence appears to favor, but does not conclusively prove, B. When in evolution does eusociality become irreversible? We infer that this comes very early in the evolution of that The Origin of Eusociality stage and circumstance, but it does not condition, in particular when an ana- A key question remaining is whether, in join group selection from environmental tomically distinct worker caste first ap- fact, relatedness is an important biasing pressure in driving the system. Close pears, hence when a colony can most factor and hence kin selection an impor- genetic relatedness arises from eusocial- meaningfully be called a superorganism. tant force in the origin and maintenance ity but is not a necessary precondition. Three lines of solitary halictine bees and of eusociality. For the origin and evolu- In the second model, alleles exist that one of allodapine apid bees are known tion of eusociality before the point of no induce cooperation and possess pheno- to have originated from primitively eu- return, a new and alternative model typic plasticity, which includes a nonge- social lines, in which the worker caste should be considered. The two compet- netic worker caste. If these alleles are was not yet anatomically distinct (14, ing hypotheses can be compared as fol- favored by sufficiently powerful colony 15). In contrast, not a single such rever- lows (see also Fig. 1). selection, eusociality can originate when sal is known among the Ͼ11,000 de- founding members of the primitive colo- scribed species of ants (family Formici- Model A. In the standard model, altruism nies have low relatedness or, in theory dae) or 2,000 described termites (order originates by selection of its prescribing (and albeit unlikely), none at all (18, Isoptera). Both of these taxonomic alleles through a kin network. Kin selec- 19). What counts is the common posses- groups contain many species with rela- tion is thus strongly binding, whereas sion of eusociality alleles, not related- tively primitive social organizations, but individual direct selection is dissolutive. ness. However, the probability that whose worker caste is anatomically dis- Group selection from environmental others share such alleles is obviously tinct or else are social parasites derived pressure and kin selection drive the sys- higher in closely related than in unre- from lines in which this trait previously tem. Close genetic relatedness precedes lated individuals (after all, when such an existed and that now depend on host the origin of eusociality. allele arises, it most likely spreads in a workers. The same evidence concerning population of relatives). the lack of evolutionary reversal is of- Model B. In the alternative model, which Thus, if the second, opposing hypoth- fered by the scores of lines of polistine we propose here, altruism originates by esis is correct, the high relatedness and vespine wasp species that possess an group selection of prescribing alleles generally observed in even the evolu- anatomical worker caste. The only whose phenotypic expression is flexible, tionarily primitive colonies could be known exceptions to this rule are sev- directing individuals to become repro- more a consequence than a cause of eral lines of thrips (16) and aphids (17) ductives or workers according to physio- eusociality and not a necessity for its that have lost the nonreproductive logical or behavioral cues. Kin selection origin. Preeusocial groups formed of soldier caste and hence reverted from can be absent, weakly binding, or weakly parents and offspring would automati- eusociality to cooperative breeding. dissolutive, according to evolutionary cally have close kinship. In the hypothe-

13368 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0505858102 Wilson and Ho¨ lldobler sis we offer here, groupwide relatedness (Polistes), females act less aggressively sion of labor by genetic proneness to is raised in the following two steps: toward genetic relatives than toward specialization by the workers. group selection leads to eusociality with nonrelatives but do not favor close over At the same time, a great many stud- little or no binding kin selection, then distant relatives when founding colonies ies have implicated kin selection as a further group selection raises the degree (23). Similar results obtained from the weak dissolutive force arising from nep- of relatedness by reducing the dissolu- primitively eusocial wasps of the genus otism and conflict among colony mem- tive effects of selfish individual behavior Ropalidia demonstrate that relatedness bers. Its principal documented effect is and (in the event binding kin selection plays no significant role in eusocial the perturbation of sex investment ratios occurs) nepotistic competition arising organization (24). in the social Hymenoptera away from from the kin selection that variously Also arguing against a strong biasing the expected 1:1 Fisherian balance. The favors different genetic subgroups. role of relatedness and binding force of perturbation stems from the haplodip- Another way of expressing the process kin selection in the origin of eusociality loid mode of sex determination in the is that as the degree of relatedness is the collapse of the ‘‘haplodiploid hy- Hymenoptera, which causes ratios in drops, the dissolutive effect of individual pothesis,’’ an early and once persuasive singly mated mother queens (1:1) to dif- selection and kin selection increases, stanchion of the standard model, due to fer from those of her daughter workers and as the degree of relatedness rises the discovery in recent years of enough (more investment in new virgin queens through group selection, the dissolutive phylogenetically separate lines (9) to than in males) (4, 5). effects of individual selection and kin render the association of haplodiploidy A second phenomenon possibly biased selection decline. Close relatedness at and eusociality originations statistically by relatedness and established in the the outset accelerates the concentration independent. later, irreversible stage of eusocial evo- of eusociality genes but does not work Finally, the rarity of the evolutionary lution is policing, the use of harassment as a selective force. Although quickly origin of eusociality also appears to or selective egg removal to restrict re- favored by group selection, close relat- argue against the standard model and production to the reproductive individ- edness is not a form of kin selection, favor its alternative, a point to be elabo- ual. Kin selection has been strongly indi- which is the preferential favoring of col- rated shortly. cated as a binding force in one species lateral relatives (i.e., not including off- of social wasps, where policing decreases spring) within groups according to their Beyond the Point of No Return with the relatedness of the workers (30). degree of relationship. In the later stages of eusocial evolution, On the other hand, the role of kin selec- Evidence from living presocial and past the point of no return, the favoring tion has been eliminated altogether in primitively eusocial species that might of close collateral kin has been depicted favor of group selection in the Cape distinguish between the two hypotheses as dissolutive in some respects but also, race of the honey bee (31) and several and shed light on the importance of ge- and much more importantly (with some species of ponerine and formicine ants netic relatedness is still tenuous. But it writers demurring to variable degree), as (32–36). is at least consistent with the second a strong binding force crucial to the hypothesis, namely that relatedness is maintenance of altruism and eusociality. The Rareness of Eusociality less powerful a factor and kin selection A growing body of evidence of several The colonial insects thus seemingly often more dissolutive in the early kinds now suggests otherwise. It in- command us to turn back to the exter- stages of colonial evolution than previ- cludes the rarity of male production by nal environment and group selection ously suspected. Xylocopine carpenter workers in colonies of social hym- and away from a preoccupation with bees, which are facultatively semisocial enopterans (ants, bees, and wasps), with abstract models of kin selection if we and possibly examples of an early stage one single-mated queen, contrary to the are to best understand the origin and in the origin of eusociality, often form prediction from models of narrow kin maintenance of eusociality. That frame pairs, which then divide into queen and selection (25); the lack of favoring bias shift is consistent with a remarkable worker roles by means of a dominance by workers of their respective mothers phenomenon that has been in plain sight order. Subordinates prefer to stay if un- in colonies with multiple queens, also at but largely overlooked by investigators: related but to leave if related. The result variance with traditional expectation the phylogenetic rarity of eusociality. Of is consistent with kin selection, but the (26, 27); and the existence of unex- the Ϸ2,600 living taxonomic families of effect is dissolutive rather than binding pected low degrees of relationship, in insects and other arthropods currently (20). Thus, group selection appears not some cases approaching background val- recognized, only 15 are known to con- to be yet strong enough to counteract ues, in many species of ants (27). More- tain eusocial species (13, 24, 37). Be- dissolutive kin selection. Groups of bees over, nestmate recognition in the social cause six of the eusocial families are in the allodapine genus Exoneura have Hymenoptera is by colony scent, which termites, the living species of which ap- higher per capita productivity when the turns out to be a complex gestalt of hy- parently derived from a single phylad, members are closely related but individ- drocarbons absorbed into the outer cuti- whereas eusociality of halictid bees orig- ual reproduction approaches uniformity, cle of the exoskeleton, shared by food inated in three independent events (15), as expected from group selection, and, if exchange and grooming, learned by im- the total number of known origins of sustained in the absence of sufficiently printing, and largely independent of eusociality in arthropods is 12. strong colony selection, would seem to kinship in composition (25, 28). In an It follows that some property of na- preclude the origin of an altruistic ant checked for such correlations ture has set a very high bar for the at- worker caste (21). In the very primi- (Pogonomyrmex occidentalis), colonies tainment of eusociality. The inhibiting tively eusocial wasp Eustenogaster with low relatedness among the workers property is most likely to be the dissolu- fraterna, founding females have lower have overwhelmingly higher growth and tive force of individual direct selection. relatedness than sisters and, due to high reproduction rates than those with high Even the large number of phyletic lines adult turnover, the relatedness of un- worker relatedness (29). This effect, fa- where individuals nest in aggregates and mated females to female brood is far voring group selection as opposed to kin disperse to a limited degree, including lower (r ϭ 0.210 Ϯ 0.171, 95% confi- selection as the binding force, may be those that reproduce by cloning (r ϭ 1), dence intervals) (22). Among somewhat due to improved genetic resistance to have not in the vast majority of cases primitively eusocial paper wasps disease or to the enhancement of divi- vaulted the bar. It seems to follow that

Wilson and Ho¨ lldobler PNAS ͉ September 20, 2005 ͉ vol. 102 ͉ no. 38 ͉ 13369 only some additional extraordinary cir- eusocial, and in none extant do adults mined in larval life or even in part cumstance or set of circumstances in build nests or provision their young. genetically (13, 37). their prior history and in the environ- Other eusocial species, those outside the Among the Ͼ11,000 known species of mental challenge they faced lifted them Hymenoptera, have descended from ants, mature colony size, varying by six over the bar. When those rare condi- species adapted to unusually well pro- orders of magnitude, is correlated with tions exist, it is not difficult to imagine tected microhabitats that can be most sociobiological trends parallel to those how the plasticity of a single genome effectively used and defended by groups. of bees and wasps. It is also manifestly can be molded by group selection first They include green-wood borings (the adapted to nest sites and food habits to create cooperative breeding and then ambrosia beetle Austroplatypus) (41), particular to each species in turn. At caste systems (the defining property of plant galls (aphids and thrips) (42, 43), one extreme are the tens of workers of eusociality). The adaptive programming and, in the one known example among some predatory ponerine colonies that of phenotypic plasticity to fit environ- crustaceans, cavities in sponges (Synal- occupy preformed cavities in decaying mental contingency is a well established pheus snapping shrimps) (44). vegetation and specialize variously on principle of genetics (38). So even if col- A second rule is consistent with the polyxenid millipedes, centipedes, beetles, ony selection renders some individuals first: the key adaptation that led to eu- and other restricted prey. At the other sterile drudges or suicidal defenders, it sociality is defense against enemies, extreme are the gigantic colonies of does not matter, because neither devi- specifically predators, parasites, and African driver ants, with Ͼ20 million ants who leave to try their luck as soli- competitors (9, 45). Numerous studies, workers, that live in large excavated soil taires nor colonies weakened by their especially those on ants, have shown cavities and hunt in armies for a wide deviancy can compete against well inte- that groups are more effective in de- range of arthropod and small vertebrate grated colonies of the same species. fense than solitary individuals and large prey (25). Even if workers retain the ability to re- groups more effective than small groups. The origin of colonies with mature produce and dissolutive individual selec- In more advanced species, with larger colony populations of 105 and higher, tion remains potentially strong, their mature colonies and recruitment com- bolstered by major innovations in colony prudent cooperation could be enhanced munication, colonies gain further advan- organization, is a rare event, like the by programming into the hereditary so- tage from their superior foraging origin of eusociality itself. The weaving cial repertory through the acceptance of methods (25, 46). of arboreal nests from larval silk has ‘‘bribes’’ of small amounts of reproduc- Along with the establishment of the originated in ants only three times tion bestowed by the dominant queens fortress has been the evolutionary devel- worldwide. Army-ant behavior, marked or by the dominant collectivity of nest- opment of the factory. The division of by mass foraging, has appeared perhaps mates, or else the prospect of assuming labor has grown stronger along phyletic six times. Within this ensemble, swarm- the role of primary reproductive in the lines, first between reproductives and raiding, in which the foraging army ad- event of the death of the queen (39). workers and then within the worker vances over a broad front, has appeared caste. Communication has grown in three times, in Dorylus, Eciton, and Phei- Colony Selection as the Driver complexity. Judging by comparison dologeton, respectively. Fungus garden- The defining nature of the programs of among living species, signals of fertility ing has evolved only once in ants, in the developmental plasticity in insect super- status and colony recognition were first; New World Attini, and within this tribe, organisms can be more clearly visualized then alarm signals; next, recruitment to it has advanced to the use of fresh foli- by the metaphor of the colony as a fac- food sources and new nest sites by suc- age only once also, in the monophyletic tory inside a fortress. The factory is the cessful foragers; then mass foraging and, leafcutter complex of Acromyrmex and egg-laying queen together with the nurse depending on the timing and spatial dis- Atta (25). workers who rear the young to maturity, tribution of resources, long-lasting trunk whereas the fortress is the nest and the trails; and finally, territorial defense by Eusociality and Preemptive workers who build, repair, and guard it territorial pheromone marking, ritual- Ecological Success (12, 40). ized tournaments, and other sophisti- The breakthrough by two of the lines to Several evolutionary rules have cated exchanges (25). eusociality has conferred on them spec- emerged from decades-long studies of Correlative studies across large num- tacular ecological success. Although ants the insect fortress and factory. All point bers of eusocial insect species provide and termites together compose only 2% to ecological colony selection as the strong evidence that the colony life cy- of the Ϸ900,000 insect species known strong binding force of eusocial evolu- cle of each species, along with the de- globally, they make up more than half tion. First is the necessity of preadapta- tails of its caste and communication the insect biomass. Their dominance is tion. The known eusocial phyletic lines system, has been shaped by the particu- ecological in origin. Colonies control with primitive species still living stand lar environments in which the species nest sites and foraging grounds in com- out among closely related solitary phyl- occurred through evolutionary time. The petition with solitary insects. They use etic lines in adaptations that evidently most important predictor of evolved so- chemical communication to assemble predisposed them to forming colonies. cial complexity is mature colony size nestmates and organized maneuvers to In the Hymenoptera, for example, at (25, 46). Among species of eusocial bees defeat adversaries. The disposition of least seven independent lines among the and wasps, for example, an evolutionary workers to risk or surrender their lives 60,000 or so known species of wasps and increase in the populations was gener- enhances the genetic fitness of the bees (with ants added, composing the ally accompanied by elaboration of the mother colony. In general, ants in par- division Aculeata of the hymenopteran division of labor, together with in- ticular dominate the central, more stable suborder Apocrita) have evolved eusoci- creased complexity in the modes of areas of habitats, whereas solitary in- ality, evidently from ancestral species caste determination and changes in the sects are best able to flourish in the pe- that built nests and fed their larvae with modes of fertility signaling. Caste ripheral, more ephemeral areas (25, 45). prey or pollen. In contrast, not one of shifted from a flexibly behavioral rela- How did this hegemony arise? The the Ͼ70,000 known sawflies (suborder tionship among anatomically similar fe- modern ant family (Formicidae) is dom- Symphyta) and parasitic hymenopterans males to complete dimorphism in which inated by four subfamilies, including the (division Parasitica of the Apocrita) are the status of colony members is deter- Ponerinae, a member of the anatomi-

13370 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0505858102 Wilson and Ho¨ lldobler cally and behaviorally relatively primi- many lines) as granivores and collectors with increasing clarity, the relations be- tive poneromorph group of subfamilies, of homopteran sugary excreta. During tween levels of biological organization and the more advanced Myrmicinae, the Eocene, the dietary versatility of the and the forces of natural selection that Dolichoderinae, and Formicinae. Each myrmicines promoted their advance into formed and shaped them. We have also of these ‘‘big four’’ far exceeds in num- other habitats. The same expansion oc- begun to glimpse, albeit still dimly and ber of genera, global range, and local curred even more disproportionately in in fragments, connections between ma- abundance any of the 17 other currently the formicines and dolichoderines, with jor features of the sociobiology, ecology, recognized living subfamilies, including a large number of species spreading and biogeography of these insects. five other poneromorph subfamilies. The both upward into the tropical forest can- If the conclusions drawn here about origin of the contemporary ecological opies, which they now dominate in good eusociality in insects and other arthro- success of the ants is essentially the his- part as cryptic herbivores on enriched pods are correct, they could have impli- tory of these four groups. Although still sap transmitted as excrement by ho- cations for advanced social behavior largely speculative, the following ecolog- mopterans (49), and outward into tem- outside the arthropods. Rarity and the ical history of the dominant groups can perate forests and xeric habitats around preeminence of group selection in be drawn from fossil evidence and phy- the world. By the mid-Eocene, ants had unusual environments that favor cooper- logenetic reconstructions of the modern become numerically dominant insects. ation are shared by the bathyergid ro- fauna (47), combined with the natural The history of the ants thus appears to dents, the only highly eusocial phylad history of the dominant groups (25, 48). have started with ecological preemption known in the vertebrates. Rarity of oc- Ants arose in the ground layer of the based on eusociality in the late Creta- currence and unusual preadaptations tropical Cretaceous angiosperm and an- ceous or early Paleogene and to have characterized the early species of Homo giosperm–gymnosperm mixed forests, or continued with subsequent dynastic ex- and were followed in a similar manner else they penetrated this habitat early in pansions into environments beyond the during the advancement of the ants and their evolution and began to diversify tropical forest ground layer (47). termites by the spectacular ecological within it. The poneromorphs, including Broader Implications success and preemptive exclusion of the stem clade of the present-day competing forms by Homo sapiens. Ponerinae, gained an early advantage as In the ants and other social insects, we predators of arthropods and flourished are thus privileged to see not only how We are very grateful to A. F. G. Bourke, thereafter in the broad array of niches complex societies have evolved indepen- J. T. Costa, D. Haig, P. Langer, D. C. Quel- opened by this specialization. The myr- dently of those of humans and in a dif- ler, G. E. Robinson, M. J. West-Eberhard, micines also proliferated in the ground ferent sensory modality (mostly chemo- and D. S. Wilson for extended discussion and layer, mostly as predators but also (in sensory versus audiovisual) but also, critiques related to the present article.

1. Wheeler, W. M. (1911) J. Morphol. 22, 307–325. 16. Crespi, B. J. & Worobey, M. (1998) Evolution 33. Kakuta, N. & Tsuji, K. (1999) Behav. Ecol. Socio- 2. Seeley, T. D. (1995) The Wisdom of the Hive: The (Lawrence, Kans.) 52, 1686–1696. biol. 46, 180–189. Social Physiology of Honey Bee Colonies (Harvard 17. Stern, D. L. & Foster, W. A. (1997) in The 34. Gobin, B., Billen, J. & Peeters, C. (1999) Anim. Univ. Press, Cambridge, MA). Evolution of Social Behavior in Insects and Arach- Behav. 58, 1117–1122. 3. Camazine, S., Deneubourg, J.-L., Franks, N. R., nids, eds. Choe, J. C. & Crespi, B. J. (Oxford Univ. 35. Endler, A., Liebig, J., Schmitt, T., Parker, J. E., Sneyd, J., Theraulaz, G. & Bonabeau, E. (2001) Press, New York), pp. 150–165. Jones, G. R., Schreier, P. & Ho¨lldobler, B. (2004) Self-Organization in Biological Systems (Princeton 18. Wilson, D. S. & Dugatkin, L. A. (1997) Am. Nat. Proc. Natl. Acad. Sci. USA 101, 2945–2950. Univ. Press, Princeton). 149, 336–351. 36. Hartmann, A., Wantia, J., Torres, J. A. & Heinze, J. 19. Riolo, R. L., Cohen, M. D. & Axelrod, R. (2001) 4. Bourke, A. F. G. & Franks, N. R. (1995) Social (2003) Proc. Natl. Acad. Sci. USA 100, 12836–12840. Nature 414, 441–443. Evolution in Ants (Princeton Univ. Press, Prince- 37. Michener, C. D. (1974) The Social Behavior of the 20. Dunn, T. & Richards, M. H. (2003) Behav. Ecol. ton). Bees (Harvard Univ. Press, Cambridge, MA). 14, 417–424. 5. Crozier, R. H. & Pamilo, P. (1996) Evolution of 38. West-Eberhard, M. J. (2003) Developmental Plas- Social Insect Colonies: Sex Allocation and Kin 21. Langer, P., Hagendoorn, K. & Keller, L. (2004) Nature 428, 844–847. ticity and Evolution (Oxford Univ. Press, New Selection (Oxford Univ. Press, New York). York). 6. Korb, J. & Heinze, J. (2004) Naturwissenschaften 22. Landi, M., Queller, D. C., Turillazzi, S. & Strass- Insect Soc. 50, 39. Reeve, H. K. & Jeanne, R. L. (2003) Proc. R. Soc. 91, 291–304. mann, J. E. (2003) 262–267. 23. Strassmann, J. E. (1996) in Natural History and London Ser. B 270, 1041–1045. 7. Sille´n-Tullberg, B. (1993) Evolution (Lawrence, Evolution of Paper-Wasps, eds. Turillazzi, S. & 40. Wilson, E. O. (1968) Am. Nat. 102, 41–66. Kans.) 47, 1182–1191. West-Eberhard, M. J. (Oxford Univ. Press, New 41. Kent, D. S. & Simpson, J. A. (1992) Naturwissen- 8. Costa, J. T. & Ross, K. G. (2003) Proc. R. Soc. York), pp. 190–201. schaften 79, 86–87. London Ser. B 270, 1697–1702. 24. Gadagkar, R. (2001) The Social Biology of Ropa- 42. Stern, D. L. & Foster, W. A. (1996) Biol. Rev. 72, 9. Choe, J. C. & Crespi, B. J., eds. (1997) The lidia marginata: Toward Understanding the Evolu- 27–79. Evolution of Social Behavior in Insects and Arach- tion of Eusociality (Harvard Univ. Press, Cam- 43. Crespi, B. J. (1992) Nature 359, 724–726. nids (Cambridge Univ. Press, New York). bridge, MA). 44. Duffy, J. E., Morrison, C. L. & Rios, R. (2000) 10. Abe, T., Bignell, D. E. & Higashi, M., eds. (2000) 25. Ho¨lldobler, B. & Wilson, E. O. (1990) The Ants Evolution (Lawrence, Kans.) 54, 403–516. Termites: Evolution, Sociality, Symbioses, Ecology (Harvard Univ. Press, Cambridge, MA). (Kluwer, Amsterdam). 45. Wilson, E. O. (1990) Success and Dominance in 26. Bernasconi, G., Krieger, M. J. B. & Keller, L. Ecosystems: The Case of the Social Insects (Ecology 11. Thorne, B. L., Breisch, N. L. & Muscedere, M. L. (1997) Proc. R. Soc. London Ser. B 264, 1331–1336. Institute, Oldendorf͞Luhe, Germany). (2003) Proc. Natl. Acad. Sci. USA 100, 12808– 27. Keller, L. (1995) Trends Ecol. Evol. 10, 355–360. 46. Bourke, A. F. G. (1999) J. Evol. Biol. 12, 245–246. 12813. 28. Boulay, R., Hefetz, A., Soroker, V. & Lenoir, L. 47. Wilson, E. O. & Ho¨lldobler, B. (2005) Proc. Natl. 12. Oster, G. F. & Wilson, E. O. (1978) Caste and (2000) Anim. Behav. 59, 1127–1133. Acad. Sci. USA 102, 7411–7414. Ecology in the Social Insects (Princeton Univ. 29. Cole, B. J. & Wiernasz, D. C. (1999) Science 285, Press, Princeton). 891–893. 48. Ward, P. S. (2000) in Ants: Standard Methods for 13. Wilson, E. O. (1971) The Insect Societies (Harvard 30. Foster, K. R. & Ratnieks, F. L. W. (2003) Nature Measuring and Monitoring Biodiversity, eds. Ma- Univ. Press, Cambridge, MA). 407, 692–693. jer, J. D., Alonso, L. & Schultz, T. R. (Smithso- 14. Wcislo, W. T. & Danforth, D. (1997) Trends Ecol. 31. Pirk, C. W. W., Neumann, P. & Ratnieks, F. L. W. nian Institution Press, Washington, DC), pp. Evol. 12, 468–474. (2002) Behav. Ecol. 14, 347–357. 99–121. 15. Danforth, D. (2002) Proc. Natl. Acad. Sci. USA 99, 32. Liebig, J., Peeters, C. & Ho¨lldobler, B. (1999) 49. Davidson, D. W., Cook, S. S., Snelling, R. R. & 286–290. Proc. R. Soc. London Ser. B 266, 1865–1870. Chua, T. H. (2003) Science 300, 969–972.

Wilson and Ho¨ lldobler PNAS ͉ September 20, 2005 ͉ vol. 102 ͉ no. 38 ͉ 13371