The Princeton Guide to Evolution

III.4

Kin Selection and Inclusive Fitness

David C. Queller and Joan E. Strassmann

OUTLINE

1. The problem of altruism

2. Inclusive fitness and Hamilton’s rule

3. Kinds of social selection

4. Comparative evidence in social insects

5. Experimental evidence in microbes

6. Kin recognition

7. Challenges to kin selection

Kin selection is selection that operates via effects on relatives. It is the explanation for altruistic behavior, where an actor gives up fitness in order to help other individuals, because the trait can spread only through possession of the altruism gene by beneficiaries. It also applies to other forms of behavior toward relatives, including selfish behavior. Kin selection, and the associated but broader concept of inclusive fitness, is supported by many theoretical models and empirical studies.

GLOSSARY

Altruism. A behavior that is costly to its performer but that aids other individuals.

Eusocial. Societies with overlapping generations, cooperative care of young, and reproductive division of labor; usually applied to insects though some vertebrates and shrimp meet these criteria.

Hamilton’s Rule. A formulation that states that a trait is favored when it raises average inclusive fitness; for an altruistic behavior it is –c + rb > 0, where c is cost to self, b is benefit to partner, and r is their relatedness.

Haplodiploid Hypothesis. Hamilton’s hypothesis that the unusually high sister relatedness of ants, bees, and wasps was critical in the evolution of their eusociality.

Inclusive Fitness. The sum of all an individual’s fitness effects, on self and others, each multiplied by relatedness; it is a quantity that is maximized by selection.

Kin Recognition. A mechanism by which individuals identify their kin.

Kin Selection. Selection on genes causing behavior in one individual through the effects on fitness of other individuals who share the genes.

Phenotype Matching. Kin recognition mechanisms involving comparison of unknown individuals to a template, often learned from known kin.

Relatedness. Genetic similarity above the level expected by chance, usually due to pedigree ties.

1. THE PROBLEM OF ALTRUISM

Eusocial insect colonies are marvels of cooperation. Ants and termites follow chemical trails to good food sources, while honey bees communicate food locations via dance. Leaf-cutter ants can strip large trees bare of leaves, while blind army ants subdue prey many times their size and can even dismember vertebrates. Food is carried back to nests and distributed to growing young, often after storage, processing, and even cultivation. Nest structures can be built from leaves, mud, paper, and even, in army ants, their own bodies. Many ants excavate elaborate tunnel systems reaching far below ground, while some termites erect earthen towers stretching far above it. These colonies are carefully protected. Guards swarm out of the entrance to attack intruders with bites, stings, and venom, or occasionally employ more innovative defenses, such as termites that rupture their bodies to release glue to entangle enemies. Others simply block the entrance, sometimes using specially shaped heads as shields. Not one of these functions—foraging, building, feeding, or defense—is carried out in isolation; instead each is a regulated and finely coordinated group activity. For each of them, one could ask the same questions one would ask of any complex adaptation. Where do the right variations come from? How are the traits built by step-by-step selection?

But an additional, more special question arises. Beneath all this complex cooperation is a somewhat more subtle form that has been even more challenging to conventional evolutionary theory. All the tasks described above are carried out by the colony’s workers, a class of individuals that normally does not reproduce (see chapter VII.13). Their actions benefit the reproduction of one or a few queens, along with their mates (the latter existing only as the queen’s stored sperm in ants, but still present in termites). How are these adaptations inherited, given that the actors have no offspring? And going one step further, how could they evolve to have no offspring in the first place?

Darwin recognized this problem and viewed it as the “one special difficulty, which at first appeared to me insuperable, and actually fatal to my whole theory.” His solution was selection at the level of the family. He noted that animal breeders can select for tasty steaks, despite the death required for tasting, by breeding from the same stock, which would have the same characteristics.

2. INCLUSIVE FITNESS AND HAMILTON’S RULE

Darwin’s idea would work straightforwardly for clonal organisms. If we could breed from a clonemate of the cow who died to produce the desirable steak, it would be genetically just like breeding from the tasty cow herself. Indeed, we are familiar with this kind of explanation for the altruistic behavior at the lower level of cells in a body. A cow consists of a clone of cells, so though a cow’s liver cell or brain cell never reaches the next generation, the cells’ traits are inherited via the identical germ-line cells. At the level of the organism, individual cows do not have clonal relatives, but they do have siblings and more distant relatives. These too share genes with the tasty cow, but to a lesser degree.

Darwin did not know about genes, but in the 1960s W. D. Hamilton quantified the idea of gene sharing or relatedness, leading to the idea we now call kin selection. Hamilton’s math led to a useful new concept, inclusive fitness, and a quite simple result, which we call Hamilton’s rule. Inclusive fitness is the sum of all an individual’s fitness effects, on self and others, each multiplied by the individual’s relatedness to each party. Because relatedness to self is one, in the absence of effects on relatives, this reduces to the individual’s effect on its own fitness. Hamilton’s rule essentially says that a behavior will evolve when its average inclusive fitness is positive: Σsr > 0, where s is a fitness change caused by the actor, and r is the relatedness of the actor to the individual who experienced the fitness change. Returning to the cow, we would like to apply selection of intensity s to the tasty cows themselves, but since they are already dead, we apply selection intensity s to their full siblings and will get a response that is half as strong, because full siblings are related by one-half.

The relatedness that matters for kin selection is genetic similarity above random levels in the relevant population. Cows and pigs share some genes, but breeding a pig will not give us tastier cow steaks; the pig is not relevant to evolution in the cow population. More subtly, our cow shares genes with all other cows, but breeding a random cow will still not give us tastier steaks. Likewise, helping random individuals in the population does not increase the frequency of the helping gene. The main reason that genes are identical above random levels is pedigree kinship, although other possibilities can also be addressed using inclusive fitness.

3. KINDS OF SOCIAL SELECTION

Kin selection and inclusive fitness have been most famously applied to altruistic behavior, such as that of worker social insects. Altruists sacrifice personal fitness but increase the fitness of another. In this case, Hamilton’s rule says altruism will evolve if –c + rb > 0, where c is the fitness cost to the altruist, b the benefit to the beneficiary, and r their relatedness (the relatedness of 1 of the altruist to itself is omitted). There can be multiple rb terms if there are multiple beneficiaries. The equation makes it clear that altruism cannot evolve if it benefits only nonrelatives, and that altruism can evolve with some suitable combination of high relatedness, high benefits to relatives, and low costs to self.

It is important to remember, however, that kin selection and inclusive fitness apply not only to altruism but also to any effects on relatives. These effects are often organized into four classes according to the signs of the effects on self versus relatives. Altruism involves negative effects on self and positive effects on relatives. A positive effect on both parties is called mutual benefit. Selfish behavior involves a positive effect on self and a negative one on the partner. Note that relatedness reduces selfishness, but does not preclude it: one can evolve to harm relatives if b_self – rc_partner > 0. The last category, spite, which involves harm both to self and others, should not normally evolve. However, harm to self and to some partners can evolve if it sufficiently benefits other partners, as in the case of bacteriocins. These are bacterial poison-antidote systems, where release of the poison—sometimes suicidally by cell bursting—kills neighbors who lack the system, removing competitors of those who possess it.

For relatives that are not genetically identical, kin selection predicts the possibility of conflict. For example, when two or more honey bee queens hatch out in a queenless colony, they usually fight to the death, despite being sisters in a species that otherwise has extraordinary altruism. The reason is that neither can survive on her own, and each would do better genetically by taking over the colony herself than by letting her sister have it. Similar logic explains parent-offspring conflict, for example, over timing of weaning.

4. COMPARATIVE EVIDENCE IN SOCIAL INSECTS

In seeking evidence of kin selection, most attention has focused on the role of relatedness, because that was the most novel aspect of the theory. In addition, much of the early attention centered on a peculiarity of relatedness that Hamilton noticed. He knew that most social insects were ants, bees, and wasps of the order Hymenoptera, and that this order has a peculiar genetic system called haplodiploidy. Here the females are typical diploids and pass on a random half of their genome to each offspring; however, males are haploid, and they pass on their full genome, but only to daughters. The system is perpetuated by whether a mother uses stored sperm to fertilize her egg or not. If she does not fertilize the egg, the offspring is a haploid male, who is fatherless. If she does, the offspring is a diploid female. Full sisters—those who share the same father—will be unusually highly related because their paternally inherited parts are completely identical. When averaged with the relatedness of one-half for the maternally inherited parts, full sister relatedness turns out to be three-quarters, well higher than the corresponding relatedness of one-half in diploid organisms. Hamilton believed this extra dose of relatedness explained why sociality had evolved so much more often in the haplodiploid Hymenoptera than in diploids, and also why workers are always female in the Hymenoptera.

However, both the prediction and the evidence for this haplodiploid hypothesis are more complicated. Haplodiploid females are more related to their full sisters but worker females normally also rear brothers, to whom they are related by only one-quarter. Their average relatedness to an equal mixture of full sisters and brothers is one-half, exactly like diploids. They can still gain an advantage if they rear more sisters than brothers, but the advantage is considerably smaller than that first envisioned by Hamilton. On the empirical side, there are issues of what should be compared. Parental care is an important preadaptation for the allo-parental care seen in social insects. The Hymenoptera are the most parental of insect orders, so perhaps the ability to provide benefits, rather than haplodiploidy, is the real reason for the preponderance of social insects in this group. Moreover, hymenopteran parental care is nearly exclusively female, so it is females and not males that are preadapted for allo-parental care. This does not mean that kin selection was not involved; it simply means that the most relevant part of Hamilton’s rule for explaining helping by Hymenoptera females is their ability to provide benefits. In agreement with this, helping to rear siblings has evolved many times in diploid vertebrates, which also have parental care as a preadaptation (see chapter VII.10). The specific benefits provided in different groups may vary (see chapter VII.13).

This could mean that the extra dose of sister relatedness was not the main factor favoring eusociality, but it does not mean that kin selection and relatedness are unimportant. A central lesson of this history is that benefits and costs are just as important to kin selection as relatedness. But the importance of relatedness remains clear. Social insect colonies are essentially always groups of relatives. Phylogenetic analysis shows that eusocial groups always arose in monogamous settings, and never in situations where multiple fathers would lower relatedness (although this sometimes evolves secondarily in advanced social insects, where workers are so specialized that they would be very ineffective as reproductives).

The best evidence for the importance of kin selection in social insects comes from sex investment ratios. As noted above, female workers would pass on more genes by investing in full sisters related by three-quarters than in brothers related by one-quarter. They do just that by feeding sisters more, or even by killing brothers, so sex ratios are usually female biased. In species where different colonies have different relatedness structures, for example, with a singly mated queen versus a multiply mated queen, the colonies with more full sisters (singly mated) specialize in rearing females while the others tend to specialize in males.

5. EXPERIMENTAL EVIDENCE IN MICROBES

Kin selection is not just for relatively smart animals, such as social insects and vertebrates. The process applies to any organism that interacts with kin, and even microbes turn out to be very social (see chapter VII.9). The bacteriocin poisons that are secreted to harm competitors have already been mentioned. Bacteria also secrete many beneficial products, particularly for food digestion, and these products become available to neighbors, who may be kin. Microbes also cooperate for motility or dispersal. Even the primary energy pathways can have a social dimension because, compared to fermentation, respiration is slow but very efficient. Therefore, in a competitive environment, respiration can involve an energy cost to the individual but, by being less wasteful, leaves more carbon resources for neighbors.

Microbes bring two great advantages to the study of kin selection. First, the genes underlying social traits are more often known and easier to manipulate. Second, one can manipulate their population structure and follow selection over many generations (see chapter III.6). Many microbial studies have documented the importance of relatedness in favoring cooperation or altruism genes. For example, the bacterium Pseudomonas aeruginosa produces and secretes molecules that bind iron in way that allows the cell, or other cells, to take it up. High relatedness structures favor secretion over nonsecretion, because some of the benefit of secretion goes to neighbors.

Another example comes from the social amoeba Dictyostelium discoideum. These amoebas collectively produce a fruiting body in which 20 percent of the cells die to form a stalk, promoting the dispersal of the other 80 percent, which become spores. Genetic selection experiments have isolated numerous cheater mutants that produce excess spores, for example, by shirking on stalk production. When populations are maintained under low-relatedness conditions, it leads to the spread of mutants that cheat but cannot fruit on their own, even though the result is sharply declining spore production of the population. Under very high relatedness, however, groups are either nearly all wild-type cooperators, which do well in the absence of cheaters, or nearly all cheaters, which cannot produce spores without having cooperators to exploit.

6. KIN RECOGNITION

Kin selection requires some method to make kin the specific targets of behavior. The simplest mechanisms involve proximity. For example, birds growing up in a nest can generally count on each other being kin. In more continuous populations, limited dispersal can make neighbors close kin. This mechanism depends on the extent to which neighbors are also each other’s closest competitors, which is determined by factors like the timing of altruism and competition during the life cycle. This increased competition can sometimes cancel out the kinship effect.

The advantages potentially available through kin selection suggest that organisms should often have evolved mechanisms to explicitly recognize kin. This could be said to be one of the great predictions of kin selection theory, something that essentially started a new field. The prediction has been confirmed repeatedly across the tree of life, from microbes to mammals. Learning is often an important component of these mechanisms. A bird might learn who its nest mates are by proximity and then remember that information for later use after leaving the nest. In addition to remembering individuals, a more general mechanism of phenotype matching can be employed. Characteristic cues of known relatives (which could include self) are learned and remembered. Later individuals are compared against this stored template and scored as kin according to the degree of match. This mechanism can allow the identification of unfamiliar relatives, which is important in large social insect colonies where all the individuals cannot be learned individually. Social insects generally use learned colony odors to separate friend and foe, though some small-colony wasps recognize individuals by color patterns.

Some phenotype matching systems rely on genetic cues, either learned as in social insects, or innate, as in social amoebas. A significant puzzle, known as Crozier’s paradox, is how the required variability is maintained at these cue loci. In theory, the most common cue allele receives more benefits from those who match, or less harm from those who do not. This advantage drives the common cue to fixation, rendering the locus useless for discrimination. One possible solution is that genetic kin recognition systems rely on cues whose variability is maintained for other reasons, such has host-pathogen selection at immune-system genes.

Kin recognition can be challenging. Errors in recognition can have costs, both if nonrelatives are favored, and if relatives are disfavored. Thus we would predict that recognition would be most effective in organisms where cooperating was most important. In an insightful study, Griffin and West explored the relationship between the importance of helping in social vertebrates, and the extent to which it was preferentially directed toward relatives. Species like kookaburra and superb fairy-wrens, where helping was not particularly beneficial, were less likely to discriminate kin. Seychelles warblers and pied kingfishers show greater benefits of helping and had a correspondingly higher preference for kin.

7. CHALLENGES TO KIN SELECTION

Kin selection theory has revolutionized our understanding of cooperative social interactions. Ever more advanced models, phylogenetic comparative studies, and experimentation, including experimental evolution, support it. However, this does not mean kin selection has not faced challenges or that it can explain all forms of cooperation.

Kin selection is not refuted by benefits that go to nonrelatives. Some beneficial effects on others are simply by-products of self-interested behavior. For example, when one parent benefits its own fitness by caring for its young, it also enhances the fitness of the other parent, who is typically unrelated. Secretion by a bacterium of products that help unrelated neighbors could still be favored provided the bacterium itself gets a net gain.

Other beneficial effects are neither by-products nor kin-selected benefits. Much human cooperation occurs between nonrelatives who do not share breeding interests (see chapter VII.11). This cooperation can be explained as reciprocal altruism, which requires direct or indirect payoffs to the actors. The kinds of accounting and retaliation necessary for reciprocal altruism to work make it unlikely in most animals, but there are still other ways of cooperating for immediate direct benefits. For example, all individuals in a herd may benefit by saturating local predators. Huddling for warmth can help all involved. Another major kind of cooperation is mutualism. Even members of different kingdoms can benefit in mutualisms. For example, in a lichen mutualism, the fungus provides structure and protection, while the alga provides carbon. Cleaner shrimp eat parasites they pick off the fish they clean. There are many more such examples throughout the tree of life (see chapter VII.9).

The key difference between all these other forms of cooperation and kin-selection cooperation is that the former all require direct benefits to the actor. Only kin selection can lead to true altruism, where there is a fitness cost to the actor that is not repaid. Since inclusive fitness includes not just kin benefits but also direct benefits to self, it can account for all these forms of cooperation.

There is a kind of true altruism that involves some individuals helping others with no evident selfish gain to their soma, or to their genes. This is the kind of thing that kin selection predicts should not be favored by selection, so if it were pervasive, it would mean trouble for kin selection theory. Unicoloniality is a feature of sociality in a small number of ant species that appears to break all rules. There is a great deal of movement from one nest to another, resulting in workers rearing brood to which they are unrelated. In some species, such as the Argentine ant, single cooperative colonies can extend for hundreds of kilometers.

Unicoloniality appears to be the result of family recognition gone astray, through genetic bottlenecks and loss of kin recognition alleles during invasion. Kin selection does not preclude the existence of this kind of trait, but it does predict that the trait is maladaptive. The lack of relatedness is predicted to result in the loss of selection on adaptive worker behaviors, since genes for effective worker behavior will not be favored by selection and should degrade through mutation and drift. Unicoloniality is therefore expected to be evolutionarily short-lived and this appears to be true. The trait is scattered through the twigs of the ant phylogeny and does not give rise to successful clades, according to an analysis by Heikki Helantera and colleagues.

8. INCLUSIVE FITNESS AND OTHER APPROACHES

Inclusive fitness is a very easy way to think about kin selection. It is simple and yet quite general. It provides a quantity maximized by selection and therefore allows us to continue the very useful practice of imagining individuals as fitness-maximizing agents (now inclusive-fitness maximizing) that act as if they had goals. However, the process of kin selection can be modeled by a variety of approaches from population genetics, game theory, evolutionary dynamics, and quantitative genetics. The result of such models is generally a version of Hamilton’s rule, though the models may not be formulated to make this immediately apparent.

Indirect genetic-effects models extend the quantitative genetic tradition to social traits. They are becoming increasingly popular because they (like inclusive fitness) start with phenotypic traits for which the underlying genetics may be unknown, which is true for most traits biologists want to study. These models work with selection coefficients and heritabilities, and show that relatedness can be thought of as a ratio of heritabilities. For example, a trait that increases the fitness of full siblings is half as heritable through those effects as it is when it increases the fitness of self—exactly the same as relatedness of self to full siblings.

A related approach that yields compatible results is group selection (see chapter III.2). Here selection is decomposed into selection between groups and selection within groups. Selfish behavior is usually favored within groups, but cooperation may be favored when it sufficiently increases group reproduction. Relatedness among group members falls out as a crucial parameter here too. Older ideas of group selection and adaptation were lacking primarily in that they paid scant attention to relatedness or equivalently to the relative heritabilities of group and individual effects. Current group selection models do pay attention to this and give results compatible with kin selection.

To sum up, kin selection is an empirically well-supported form of selection that occurs through effects on relatives. It is most easily understood via inclusive fitness thinking, but can also be modeled by a variety of other approaches.