Chapter 2: Genetic Causes of Behavior

Barry Sinervo

Table of Contents

Introduction to Genetic Terms and Mendelian Traits

Introduction to Genetic Terms and Mendelian Traits

When we search for the proximate factors for behavioral causes, we naturally start with genes. When the fusion of sperm and egg produces a zygote, the stage is set for a cascade of developmental events that lead to the production of the phenotype. The central dogma of biology holds that DNA from alleles at a genetic locus translates into proteins. Structural proteins are used to contract the cells, and enzymatic proteins provide the basis for cellular machinery. Cells in the developing zygote interact mechanically or chemically with adjacent cells, or cells can produce "messengers" that facilitate the cell-cell interaction at distance. Cell-cell and organ-mediated interactions cause a cascade of events that we collectively refer to as development. The embryo is gradually organized into a functioning animal with nerves and organ systems that begin to regulate behaviors. Those behaviors have functional or selective consequences. Individuals selectively live, breed and die according to their phenotype. Gene frequencies change across the generation and the species evolves.

The goal for this chapter is to develop an understanding of the genetic factors that govern the expression of morphological traits associated with alternative behaviors. Techniques of classical genetics such as genetic crosses and pedigree analyses are now being combined with tools of molecular biology to identify specific genes that appear to govern the expression of complex behaviors. Before we begin this quest, let us arm ourselves with a few of the terms of genetics (distilled from Keller and Lloyd, 1992). In particular, we focus on those genetic processes that deal with genetic variation within and among individuals.

The Genotype and Phenotype

Genotype -- the sum total of all the alleles at all the loci in a organism. While precise and concise, this definition is not very useful. The genotype may also apply to genetic material at a single genetic locus and describe the alleles that an individual possesses at a particular locus. The two alleles at a genetic locus come from the male and female parents and form the basic unit of genetic variation in an individual. The process of meiosis, mendelian segregation, and recombination among all the myriad of genetic loci in an individual effectively makes each sexually-produced individual unique (see Side Box 2.1. Mutation, Segregation and Recombination). In contrast, an asexual organism is a genetic clone of its female parent. The genotype is largely static for an individual during its lifetime, except when a mutation occurs. A mutation is a lesion of DNA that changes the genetic material in one allele at a locus. If the mutation occurs in a somatic cell, there will be no consequence for transmission of material across generations. However, if the mutation occurs in a cell of the germ line, which produces sperm or eggs for sexual reproduction, then the mutation can be transmitted across generations. Mutations are the ultimate source of all genetic variation.

Phenotype -- the external expression of the genes, and the result of a gene's interaction with the environment. This includes mechanisms of development. Because the stage of development depends on age, the phenotype can be highly labile, or can change dramatically during its lifetime. The organism develops, it learns, it acclimates and its phenotype changes accordingly. For simplicity of analysis, we break the whole organism into phenotypic traits that are largely functional units. For example, the number of offspring that an animal produces in one bout of reproduction, termed fecundity, is a phenotypic trait related to reproduction. Likewise, the kind of parental care is a second phenotypic trait related to reproduction. However, phenotypic traits are usually correlated with other phenotypic traits and such correlations arise from proximate mechanism. For example, the number of offspring that a parent produces is related to the kind or quality of care that the parent can provide to the offspring. Usually we refer to such relationships as fitness trade-offs because an increase in one fitness trait (fecundity) has an impact on a correlated trait (quality of care). The proximate mechanisms that link these particular traits are related to energy and metabolism. Rearing large numbers of offspring requires more total energy to keep the level of care constant for each offspring. Alternatively, less energy (or lower quality care) is available for each offspring.

Environment -- the environment is anything external to the genetic material. For example, food availability in part determines body weight. In animals that lay eggs without parental care, the eggs are subjected to environmental variation in the form of temperature, hydration, or perhaps salinity. In animals which lay eggs and have extended parental care in a nest, the environment is a function of the parents and perhaps the other sib-mates in the nest. In animals with internal fertilization and some kind of gestation, the mother's physiology per se becomes an important component of the offspring's environment. Regardless of whether or not an organism requires extended parental care as a juvenile, the environment still plays a major role in that individual's development. Of course, there is a second sense in which the environment interacts with the genotype and phenotype: the environment causes natural selection. While this natural selection acts on phenotypes within a generation (e.g., the parents, its effects are transmitted to the next generation (e.g., offspring).

 

A simple expression describes the relationship between the variation in a particular phenotypic trait among individuals found in a single population:

where P, G, and E are phenotypic, genotypic, and environmental variation in the trait. If the environmental variation is large, then little phenotypic variation arises due to genetic sources of variation. Conversely, if genetic variation is large, then the phenotypic trait is largely determined by genetic factors. As we have seen, heritable or genetic variation is central to Darwin’s theory of natural selection, and evolutionary change will occur more rapidly for traits that are strongly determined by genetic factors.

Side Box 2.1: Mutation, Segregation, and Recombination

The process of natural selection is often characterized as "blind" and this is largely because the source of all genetic variation is largely a stochastic process. The ultimate source of genetic variation is mutations; however, segregation, and recombination provide an stochastic mechanisms for randomizing genetic variation in a population. While mutations alter DNA by changing base pairs, segregation and recombination do not alter the material content of DNA. Instead they provide powerful mechanisms for mixing up the DNA during sexual reproduction.

Mutation

The process of mutation is probabilistic. We describe the process in terms of the probability of a mutation occurring in an individual during its lifetime. Typical rates of mutation are between 1 in 10,000 (10^-4) and 1 in 1,000,000 (10^-6) for many organisms. Either this means that a long time must pass before a mutation occurs in a population, or the population must be very large. The population must number in excess of 10⁶ or one million members in order to see an average of one mutation per generation for a gene with a mutation rate of 10^-6. Most mutations are detrimental, and perhaps only 1 in 1,000 is beneficial. Thus, in this population of a million we might have to wait for one thousand years for a specific genetic locus to throw us a beneficial mutation (of course there are thousands of possible loci in an organism so the waiting time for any beneficial mutation is less). Even if a mutation arises, there is no guarantee that natural selection will act on it.

We can calculate the probability that a beneficial mutation makes it through meiosis and segregation. A beneficial mutation occurs within a diploid parent which has two copies of each gene. The beneficial mutation has a 50% chance of being transmitted to any one offspring and a 50% chance of not being passed on. With an organism that has four progeny, the probability of a single offspring not getting the beneficial mutation is independent of the other offspring not getting the mutant allele. By using the laws of probability we can multiply successive independent events to compute the probability that none of the four progeny gets the beneficial mutation:

1/2 x 1/2 x 1/2 x 1/2 = 1/16.

If the unique beneficial mutation is not passed on to the progeny, then more time will be required for a new beneficial mutation to arise in the population. At this point, natural selection has a chance of promoting its spread through the population.

Segregation

Genes are found on chromosomes. The first random event in sexual reproduction occurs during meiosis when homologous pairs of chromosomes line up on the equatorial plate in preparation for the cell division that reduces a diploid (2N) cell of the germ line to an haploid (N) gamete. During the process of segregation different homologues will randomly be distributed between the two daughter chromosomes, gametes end up with very different chromosome complements. Given that the total number of chromosomes is C, then there are 2 raised to the power of C different gametes that could be produced from segregation of chromosomes (in humans that would be 2⁴⁶). Segregation can produce a vast number of different gametes and is a potent mixer of variation in sexual organisms.

Recombination

Recombination between homologous pairs of chromosomes during meiosis can generate even more gamete types. The chromosome is loaded with recombinational hotspots. Any given chromosome will typically recombine with its homologue at one or two points during a given meiotic event (recombination rate on a given chromosome is a function of chromosome length). However, the recombination points for two different meiotic events can be different. Given that gametes such as sperm are generated by millions of different meiotic events, the number of potential gamete types from a single parent is vast. If one considers all possible parents in a modest-sized breeding population such as humans (5 billion), there are easily more potential recombination products than molecules in the known universe.

<END SIDE BOX>

Components of Phenotypic and Genetic Variation

We can follow a roadmap from genotype to phenotype, but the route is rarely a one to one mapping between genes and phenotypic traits. Environment further obscures the relationship between genotype and phenotype. We have several hierarchical concepts that describe how genotype can be related to phenotypic traits. These additional terms describe different levels of genetic interaction and their effect on genotype and phenotype:

1. the effect of one allele on another allele at the same genetic locus (e.g., additive effect versus dominance),
2. the effect of one genetic locus on another independent locus (e.g., polygenic and epistasis),
3. the effect of a single gene or two or more phenotypic traits (e.g., pleiotropy),
4. the effect of two or more genes on a single phenotypic trait (e.g., additive, epistasis),
5. the interaction between genetic factors and the environment (e.g, norm of reaction).

Pleiotropy -- a single gene that has an effect on the expression of two or more phenotypic traits is said to have a pleiotropic effect on the traits. For example, testosterone controls the development of what are referred to as secondary sexual characteristics (e.g., a lion's mane), but testosterone also relates to behavioral traits like aggression. Thus, a gene that controls the levels of testosterone would have a pleiotropic effect on the expression of many secondary sexual traits which are morphological, as well as behavioral. The concept of pleiotropy is intimately related to the concept of trade-off (Stearns 1976). Pleiotropy describes the proximate genetic source for many phenotypic trade-offs. If one gene controls the expression of two or more traits and those traits are related to a fitness trade-off, then we have identified the proximate source of the trade-off.

Polygenic -- if two or more genes are responsible for a single trait, the phenotypic trait is said to be governed by polygenic factors. For example, growth rate is undoubtedly caused by a number of genes that act in a complex cascade. Body size, which is the result of a large number of genes, is polygenically determined. Genes that control growth hormone have a large effect on body size. Likewise, genes that control sex steroids like testosterone have some effect on body size, especially during maturation phases of growth. Undoubtedly, many genes that influence metabolism have a small, but measurable effect on body size.

Additive effects within and between loci. The simplest additive genetic relationship occurs between two alleles at the same genetic locus. If two alleles are co-dominant, then the heterozygote is exactly intermediate in phenotype relative to the two homozygous types. In this idealized case, "the effect of substituting one allele for another is additive in its effect on phenotype". Just like alleles acting in an additive fashion at a single locus, polygenic genetic loci can also interact in an additive fashion to produce a phenotype. If two or more genes have a simple effect on the phenotype they are generally thought of as having an additive effect. If a trait results from two or more genes then an additive relationship between them would lead to the simplest kind of polygenic inheritance. Additive genetic variance is what underlies the notion of heritability and is responsible for the similarity between parents and offspring. Natural selection operates on additive genetic variation because additive genetic variation is largely responsible for the resemblance or heritability between parents and offspring (see Side Box 2.2: Additive Genetic versus Dominance Variation).

Dominance. Interactions between alleles at a single locus are termed dominance interactions. For example, if an allele is said to be recessive to another allele, then an individual that possesses a copy of the dominant allele and a copy of the recessive allele (e.g., heterozygote) will be phenotypically identical to an individual that possesses two copies of the dominant allele (e.g., homozygous). The "recessive phenotype" is only expressed if the individual is homozygous for two of the recessive alleles. The rover and sitter gene of Drosophila larvae discussed below is a classic example of the effect of a dominant gene on behavior. If two genes are co-dominant then the heterozygote is intermediate between the two homozygous genotypes and the alleles are additive in their effect. While dominance variation can be a very large component of the total genetic variation of a phenotypic trait, dominance variation is entirely non-additive (see Side Box 2.2).

Epistasis. If two different genetic loci interact in any way that is not additive, then they are said to act epistatically. Many pigmentation genes act in an epistatic fashion. The simplest example of epistasis relates to genes that control pigmentation. This example illustrates how a HREF=#anchor214836 between alleles at the same locus is conceptually similar to an epistatic interaction between two genetic loci. Consider a single locus such as the coat color of a popular breed of dog the Labrador Retriever. The color yellow (y) is recessive to black (b). One copy of the black allele and production of the black pigment overwhelms the expression of the yellow coated phenotype and a heterozygous yb individual is black. The effect of the black pigment is non-additive, in that one copy completely masks the recessive yellow. A second coat color locus codes for production of brown pigment. The brown locus interacts in a non-additive fashion in that an individual with an allele at the brown locus can be brown regardless what pigment is expressed at the black locus. The brown locus overwhelms the expression of the black locus in a non-additive fashion, but in this case, brown interacts epistatically with the black locus.

Genotype and Environment Interaction and Phenotypic plasticity. If the expression of a gene depends on the environment in any way, then the phenotype is said to be due to an interaction between the genotype and environment. This idea is central to behavior and we will explore it in great detail in upcoming chapters. For example, birds undoubtedly have genes for learning, and indeed some species of birds may differ in how they learn. A famous example of genotype environment interactions relates to how birds learned to open milk bottles in England. For years milkman in England left milk bottles outside on the porch with no problems. One day a single bird learned how to open the top of the milk bottles and drink the milk. This trait was passed on to other birds by learning. Even different species of birds learned the task. Here the environment changed, with the advent of a single bird learning to opening milk bottles. However, only some species of birds learned how to open milk bottles, other species did not. Learning was presumably contingent on both the environment and on the genes for learning in each species. Both the presence and absence of milk bottles, and presence or absence of a tutor bird to teach naive birds was necessary for learning this task.

Phenotypic plasticity -- the expression of different alternative phenotypes within a constant genetic background. While, there are few genetic differences among individuals some environmental cue (predictable) causes development to proceed along different pathways. From a single genotype, many different phenotypes can be produced.

 

Side Box 2.2: Additive Genetic Variation versus Dominance Variation

What is the source of heritable variation? The resemblance between parents and offspring is entirely due to the additive effect that genes have on phenotype. Not all genetic variation is additive. For example, dominance variation arises from the degree to which one allele at a locus alters the effect of the allele on the complementary chromosome. A useful way to think about additive genetic variation is to consider the proportion of the phenotype that can be predicted from a knowledge of the number of A versus a alleles that are present in the genotype -- or the genotypic value. Consider an aa individual to have zero A alleles, Aa has one A allele, and AA has two A alleles. Only in the idealized case when alleles at a genetic locus are co-dominant, are the genes said to act in a purely additive fashion. Dominance reduces the additive genetic variation, increasing the dominance variation at the same time.

<END SIDE BOX>

  Mendelian Traits

We will explore many of these issues of genetics through examples. I have chosen a diverse set of behavioral traits to introduce many concepts in animal behavior which will be covered in greater detail in later chapters. Each example, illustrates one of the genetic terms discussed above. In many cases model organisms studied in the laboratory facilitate our study of behavioral causes because of the large sample sizes possible in the laboratory allow for necessary statistical power required to assess genetic bases of behavioral traits. Other genetic examples are drawn from natural populations. While it is more difficult to collect genetic data on animals in the wild, these larger animals provide answers to natural selection because we can follow them through their lives. Moreover, it is essential to study animals in the wild so that we can infer how their actions are reflected in the process of natural or sexual selection. By long-term observation of animals in the wild, we are able to reconstruct present day patterns of selection (Lande, 1983) that may have been historically significant during the origin of behavioral traits. A first step in selection analyses is determining the genetic basis of behavioral traits. Classic genetic crosses and pedigree analyses are the standard tools of a behavioral geneticist.

Behavioral Mutants in the Laboratory: Roving versus Sitting in Drosophila larvae

Many of the basic issues of mendelian inheritance of animal behavior are found in the example of roving versus sitting larvae in Drosophila. Marie Sokolowski and her colleagues (Sokolowski 1985; de Belle and Sokolowski 1987) generated two strains of larvae and then carried out some illuminating crosses. If you cross parental rover males and sitter females, each of the ensuing strains is pure and homozygous for either the rover or sitter allele. Crossing two pure parental lines (P0) produces progeny which are referred to as the F1. In theory, if the phenotype rests entirely on a single gene with two alternative alleles (a dominant rover allele and a recessive sitter allele). The F1 cross produces all rovers with no sitters. In practice, this cross produces nearly all rovers, suggesting that sitter is recessive to rover (or rover is dominant to sitter). The telling bit of evidence arises from a cross between F1 progeny, which should produce 3 genotypes (see Punnett Square in Table 2.1.A), but only two phenotypes (see Table 2.1.B), owing to dominance of the R allele over the S allele. These offspring produce rovers and sitters in nearly a 3:1 ratio -- exactly the pattern one expects if the F1 are all heterozygous for rover and sitter alleles.

Beak Size and Seed Preference in African Finches

Like the behavioral morphs of fruitflies, species of African Finch that inhabits Cameroon, Pyrenestes ostrinus, displays a discrete set of bill morphs with a simple mendelian pattern of inheritance (Smith, 1993). In particular, a small-beaked (s) and large-beaked (L) form co-occur throughout much of their geographic range. The small- and large-beaked forms are often found building nests together. These two forms differ dramatically in the feeding behaviors and preferences on large versus small seeds. In the wild, the small-billed morph has a very strong preference for the seeds from a species of sedge that produces small seeds while the large-billed morph prefers the seeds from a large-seeded sedge species. Given their propensity to feed on seeds, the birds common name is the seed-cracker. Thomas Smith initiated breeding studies on the P. ostrinus to determine the genetic basis of the behavioral and morphological differences between bill morphs. High levels of nest predation in the wild precluded the possibility of obtaining pedigrees from the wild. Smith circumvented this problem by importing breeding pairs for his study in a collaborative effort with the Riverbanks Zoo in South Carolina. In order to study the inheritance of beak morphology which is strongly associated with feeding behaviors, he first had to get the animals to breed in the laboratory.

Finding the environmental factor that triggers reproduction in undomesticated animals is often a daunting task. Many temperate birds are triggered to initiate reproduction with a change from short photoperiods to long photoperiods. The early attempts by the Riverbank Zoo to trigger reproduction by manipulating the photoperiod met with failure. However some tropical Finch aficionados suggested that they try an unusual environmental trigger. In the wild, the finches can rely on a very time honored cue, they experience two dry seasons and two wet seasons every year. One of the wet seasons is a little more damp than the other. The workers at the Zoo played audio tapes with load thunder, simultaneously drenching the aviaries with water. The finches, happy with the beginning of a simulated wet season, began to breed. This example, illustrates some of the challenges of non-model systems, but also the creative solutions used to solve practical problems of science.

Over the years, Smith and his colleagues developed a relatively large pedigree (Fig. 2.1) from crosses: L • L, L • s, s • s. The best fit genetic model for the control of beak size based on the genetic data in the pedigree, is a single mendelian locus with an allele for the small beaked form (s) recessive to an allele for the large beaked form (L). (see Side Box 2.4: Hardy-Weinberg Law and Random Mating). A simple mendelian gene causes large differences in morphology in the seed crackers, and this has cascading effects on a suite of foraging behaviors. This illustrates that a single gene can influence a large number of morphological and behavioral traits. As we shall see in the next Chapter on selection, the gene for beak morphology also influences survival in the wild.

<Fig. 2.1. Pedigree of seed crackers from the riverbanks zoo, see lecture>

Side Box 2.4: Random Mating and Hardy-Weinberg Gene Frequencies

The Hardy-Weinberg Theorem describes what happens to gene frequencies when no evolutionary forces act on phenotypes. This assumes that no selection, migration, mutation, or genetic drift alters the gene frequencies from generation to generation. Another important assumption is that individuals in the population breed randomly with respect to genotype and phenotype. In a randomly mating population gametes combine in proportion to the frequency of each gametic type taking part in the union. For a single locus with two alternative alleles, A which occurs at a frequency of p, and a which occurs at frequency q (or 1-p) the proportion of genotypes is given by multiplying the frequency of each gametic type:

or the more familiar: p²+ 2 • p • q + q² = 1. These calculations describe how gametes pair up randomly during fertilization. Another expression of random mating occurs at the level of the phenotype. If the frequency of each phenotype is: p², 2pq, and q², and phenotypes pair up randomly we would expect to see the following crosses to occur with the following frequencies:

Random Mating Finches

Smith (1987) observed the following frequencies of matings in Pyrenestes ostrinus: S_males • S_females = 34, S_m • L_f = 14, L_m • S_f = 14, L_m • L_f = 6. A few simple computations are in order to learn how the birds are breeding. The birds may mate randomly, assortatively by phenotype (like breeds with like) or perhaps by disassortative mating (birds seek out a more dissimilar partner). While we cannot distinguish between all the genotypic classes in P. ostrinus, we can ask whether the phenotypes are breeding randomly. What frequency of matings would we expect by chance? We need to compute the following items.

What is the probability that a small-billed male pairs randomly with a small-billed female? The probability that a small breeds with small is given by multiplying the overall frequency of each:

0.71 • 0.71 = 0.54 and, we would expect to see a total of 68 • 0.54 = 34.3.

By the same logic, we can compute our random expectations for the other three kinds of matings to derive an expected number of matings if the birds were randomly mating. By inspection alone we can see that the observed and expected random frequencies are nearly identical. We could carry out a formal test, the Chi-square, which is based on observed versus expected frequencies. I leave this as an exercise for the reader.

Mendelian Inheritance of Three Alleles

For a single locus with three alleles, the formula for frequency of each genotypic class is slightly more complicated as we add a few more heterozygous classes and one homozygous class.

The two allele and multi-allele Hardy-Weinberg Law really only implies that gametes achieve union randomly with respect to genotype. Given the observed Ams gene frequencies in isopods (see text), what is the frequency of observed mating phenotypes that we would expect under random mating?

<END SIDE BOX>

Molecular Tools of Behavioral Geneticists

The search for genetic factors in natural populations and laboratory stock depends largely on the pre-existence of genetically-based morphs. Examples include the case of the pedigree analysis of crosses between bill morphs in finches or in the screenings for alternative behaviors carried out in Drosophila. Many human genetic disorders are uncovered by analyses of pedigrees that are obtained from isolated populations where the affliction occurs at a relatively high frequency. When we see a maladaptive genetic disorder genetic drift is the probable cause. In small isolated populations, a single copy of a genetic disorder can drift to high frequency when inbreeding occurs. Inbreeding is much more likely in a small population where consanguineous matings occur at high frequency versus those in a large population where encountering a relative is quite unlikely.

Genetic sleuths refine their search for genetic factors by screening the genetic material of key families in which the affliction is particularly well characterized. If researchers find a perfect match between transmission of a certain piece of DNA and the transmission of the malady, they can then isolate where the gene coding for the behavior is to be found on a genetic map of the human genome. The number of behavioral genes discovered in humans is increasing at a rapid rate using inference from pedigrees and the transmission of specific regions of DNA.

Another important technique for isolating genes that control behavior entails a mutant screen in which a parental generation in a colony of animals is subjected to a mutagen, and then the progeny in the colony are screened for behavioral disorders. The researchers then search for the genetic basis by molecular methods that are similar to those described above in the analysis of human pedigrees. The next set of examples illustrate how molecular methods aid in the determination of genetic factors underlying behaviors.

Mutation Analysis of a Gene for Parental Care in Mice: fosB

Brown et al (1996) have recently identified a gene in mice, fosB, that is essential for the correct expression of maternal behaviors. They induced a mutation in a single gene, fosB, that when disabled appears to extinguish nurturing behaviors in female mice. Evidence from their study suggests that a very simple neural pathway may be involved. Lesion experiments have shown that the hypothalmus in female mice is critical for nurturing behavior. By deleting a gene whose gene product is expressed in the hypothalmus, Brown et al (1996) have isolated a key genetic factor involved in nurturing. It appears that the gene products of fosB are expressed in the small part of the hypothalmus called the preoptic area of the brain. Fos B deficient mothers do not show the following nurturing behaviors:

1. they do not retrieve their young, despite normal maze running ability.
2. they do not nurse the young, despite normal mammary gland development.

By validating that FosB deficient moms are not incapacitated in basic organismal functions like spatial ability or hormone physiology, Brown et al have shown that fos B is important to nurturing per se and not merely a pleiotropic effect of fos B's effects on other non-nurturing traits. Indeed, fosB deficient mothers have normal expression of the reproductive hormones Estrogen and Progesterone which change during the course of the development of nurturing behaviors of the mother. In addition, fosB deficient mothers also have normal olfactory abilities based on olfactory discrimination tests. fosB deficient mothers simply do not nurture their young, there is little else wrong with their phenotype. When normal female mice, intact virgins, and even males, are exposed to pups the expression of fosB genes are triggered in the preoptic area of the brain. Olfactory neurons in this neural pathway appear to be particularly important. The presence of fosB is necessary to induce normal parenting behavior, but it is not the only gene required -- it alone is not sufficient to elicit all aspects of parental care. Nevertheless, its correct function is crucial for the expression of normal nurturing. The FosB gene represents a crucial link to the chain of proximate factors that lead to a complex suite of parenting behaviors in mice, and perhaps many other mammals.

Paternity Analysis and a Gene for Alternative Male Strategy in the Ruff

Molecular methods are also useful for determining pedigrees of animals which have bred in the wild. In such cases the mother is usually known with certainty as she laid the eggs. However, the female could have mated with a large number of putative sires. DNA paternity analysis has become a standard for determining the identity of an offspring’s sire.

Ruffs are a shore bird that breed and nest in northern Europe and they come in two plumage and behavior morphs. "Independent" male ruffs have a territory and defend females against other independents. The "territories" are very small and localized in "leks", which are areas where many males aggregate and display to attract visiting females. Non-territorial "satellite" males move among the independent males and obtain copulations from females on the independent's territory. Ruffs are fixed in their plumage color and behavior throughout their lives. Eighty-four percent of ruffs are independents, and sixteen percent are satellite males.

David Lank (1995) and his colleagues used molecular probes called "mini-satellites" to determine which male sired the chicks on ruff breeding grounds in Finland. By comparing the alleles in chicks and alleles in the mother and putative sires, they were able to determine the father and reconstruct the father's morphology (Fig. 2.1). They also scored the morphology of the female parent's brother and father to determine the likely phenotype that the female would have expressed had she been male. Such genetic sleuthing is particularly important with sexually dimorphic traits which have a sex-limited expression (e.g., traits only seen in males or only seen in females). Females do not express the alternative male morphology, yet they might carry genes for the morphology that they pass on to their male offspring. Lank and his colleagues then reared the field caught chicks to maturity when they could score the breeding morphology of the male progeny. They also reared an additional generation of chicks in captivity that were derived from the field collected cohort of chicks. They were certain of the father's identity in captive bred birds and did not need to use molecular probes to determine paternity.

® Figure 2.2. Variation in minisatellite alleles in male ruffs (ST, AR, BT, TH) that are used to determine the sire of progeny (0, N, M) of the female (K:22). Allelic variation at three microsatellite loci are shown (cPpu11, cPpu11, cPpu3). Given the mothers known contribution to offspring, male BT is the only male that could be could contribute alleles at all three loci and he is the likely sire.

Because only two phenotypes are seen in the ruff, a single genetic locus with two alleles must be controlled by a completely dominant allele and a recessive allele, otherwise we would expect to see three phenotypes. The data on ruff pedigrees can discriminate between several simple mendelian patterns of inheritance. Mendelian genes are either found on sex chromosome or on autosomal chromosomes. Their genetic crosses rule out the possibility that the gene is a sex-linked dominant (see below). Their pedigree data is consistent with an autosomal gene with Satellite dominant.

While the example of male ruffs provides solid evidence of major genes which affect the phenotype, the sample size in these studies is still inadequate to test for the existence of other genetic loci that control development of male behaviors. Testing for the effects of two or more factors requires sample sizes in the thousands, particularly if those factors interact epistatically or in a non-additive fashion.

Epistatic Genes for Alternative Male Strategy and Sex Transformation in Marine Isopods

A final example of genes that have a major effect on behavior involves the genetic control of alternative male behaviors in a marine isopod, Paracerceis sculpta. Stephen Shuster and his colleagues have characterized three alternative male morphs in the marine isopod (Shuster and Wade, 1992, Shuster and Sassaman, 1997):

1. a large alpha male with elongate posterior appendages that they use to defend harems of females
2. a medium-sized beta male that can invade female harems by mimicking female behavior and morphology, and
3. a small-sized gamma male that invades female harems by virtue of its secretive behaviors.

In a large breeding study, based on hundreds of genetic crosses, Shuster has developed a genetic model that explains patterns of inheritance of the three morphs. Three alleles (a , b , g ) at the Alternative male strategy (Ams) locus provide a reasonable explanation of the general pattern of inheritance of the male morph. The three alleles have the following dominance relations:

1. b is dominant to both g and a ,
2. g is dominant to a ,
3. a is of course recessive to both b and g .

The a allele occurs at a very high frequency in the population (93%) compared to either b (1%) or g (6%). An alpha male phenotype must be homozygous (e.g, a |a ) at the Ams locus because a is recessive to the other to Ams alleles. Gamma males can be either a |a or g |g . However, the heterozygous form of gamma g |a is very common while the g |g form is rare in a natural populations. Likewise, beta males can be b |g , b |a , or b |b . However, as a is the most common allele in the population, the b |a is the most common genotype for a beta male. As an exercise, compute the frequency of each male genotype assuming Hardy-Weinberg gene frequencies and convince yourself that a |a , b |a , and g |a are the most common male genotypes in the isopod population.

Shuster and Sassaman (1977) found that a careful inspection of genetic crosses between females and the predominant male genotypes revealed a significant departure from a 50:50 sex ratio of the progeny. The existence of a second locus termed transformer (Tfr-two alternative alleles, 1 and 2) is required to adequately explain the aberrant sex ratio found only in certain genetic crosses. The Tfr locus causes males to transform into females, or females to transform into males.

However, the direction of sex change depends on the genotype at the Ams locus. Homozygous alpha males transform from male to female if they bear at least one copy of the Tfr - 2 allele (e.g., genotypes 1|2 and 2|2). Alpha males with the 1|1 genotype at the Tfr locus are not sex-transformed. Conversely, beta males transform from female to male if they bear at least one copy of the Tfr - 1 allele (e.g., genotypes 1|1 and 1|2). Beta males with the 2|2 genotype at the Tfr locus are not sex-transformed. Gamma males transform from female to male if they are homozygous for the Tfr-1 allele, but not if they carry one or more copies of the Tfr-2 allele. The epistatic interaction between two loci governs the expression of male and female behaviors in this marine isopod. The gene for Alternative male strategy interacts in a very non-additive fashion with the gene for Transformer.

Finally, we find an added complexity in this interesting genetic system. An extrachromosomal factor (ECF) accentuates the effect of the Tfr locus, but only in the Tfr heterozygotes of alpha and beta males. The exact nature of the ECF remains obscure but cytoplasmic effects on that determine phenotype are common in the animal kingdom. They could be transmitted from mother to egg (as are many cell organelles such as mitochondria).

Table 2.2. The interaction between three alleles at the Alternative male strategy locus (Ams) and two alleles at the Transformer locus (Tfr) in a marine isopod, Paracerceis sculpta. See text for the dominance relations at the Ams locus. A plus symbol with arrow (+) indicates a sex transformation owing to the alleles at the Tfr locus. Alpha males that bear one or more copies of the Tfr-2 allele are transformed into females during embryogenesis (pink). Conversely, Beta males that bear a one or two copies of the Tfr-1 allele are transformed from male to female (blue). The action of the Tfr gene can be accentuated by the presence of an extrachromasomal factor ECF (* in blue). However, gamma males are only transformed from female to male if they are homozygous for the Tfr allele.

Polygenic Inheritance

Needless to say, analysis of the genetic control of behaviors becomes a daunting task for anything but the simplest genetic systems. As the number of genetic loci increases beyond three or four it becomes increasingly difficult to isolate the expression of a phenotypic trait to specific genes by using standard genetic crosses. The sample sizes required become far to large in practice. Consider two loci each of which has two alleles (A,a and B,b). Let us assume that alleles at both loci additively act in the following fashion. Whereas the allele a adds nothing to phenotype, the alternative allele A increments the phenotype by one unit. Likewise, allele b adds nothing to phenotype, the alternative allele B increments the phenotype by one unit. A Punnet square which describes the union of all possible gametic types yields 16 genotye combinations and five phenotype combinations (see Table 2.3).

Table 2.3. Punnett Square for the effect of two loci which have purely additive effects.

Figure 2.2. Phenotype distribution of a trait governed by two mendelian loci that act additively to produce the phenotype. If the phenotype is not affected by environmental factors then 5 distinct modes are present. As phenotype becomes more and more affected by random environmental factors, the phenotype distribution becomes more and more normally distributed.

Rather than isolate the action of single genes, quantitative genetic analyses assume that traits are due to many genes of small effect, each of which act additively to produce the phenotype. As the number of mendelian factors increase, and as more of the phenotype is determined by environment, the distribution of many traits resembles a normal distribution. The inheritance pattern of such polygenic traits is succinctly described by a simple parameter known as heritability. The heritability reflects the proportion of a phenotypic trait that is due to the additive genetic effect of loci. The most straightforward way to estimate the heritability of a trait is to measure the resemblance or correlation between relatives such as parents and offspring, or between sibs.

A major caveat of the quantitative genetic approach to behavioral genetics is that it is quite difficult to build in epistatic effects such as those that are seen in alternative male morphs of isopods. This is because additive effects are easy to model. In contrast, epistatic models can take hundreds of possible combinations for even a simple two or three locus system.

Side Box 2.5: Heritability

A central aspect of Darwin's theory of evolution includes a key statement: natural selection acts on traits that are heritably transmitted between parents and offspring.

Darwin's cousin, Sir Francis Galton coined the term for regression based on the observation that the regression line which predicts offspring phenotype from a parent's phenotype always has a slope less than one. The slope of the parent-offspring regression line is also known as heritability. The heritability for any phenotypic trait describes the proportion of the offspring's phenotype that we can predict from a knowledge of the phenotype of both parents. The prediction is not exact and includes some error about the regression line because offspring phenotype includes environmental effects in addition to polygenic factors inherited from parents. Finally, another formulation of heritability is derived from Eqn. 2.1:

heritability = h² = G / P = G / (G + E).

Heritability is the proportion of phenotypic variation (P = G + E) that is due to genetic causes (G). Because phenotypic variation is larger than genetic variation, heritability be less than one.

<END SIDE BOX>

It is possible for measured heritability to be confounded with environment (Falconer, 1981). For example, if parents and offspring share a common environmental factor which makes them more similar by upbringing than genes alone heritability could be inflated. In principle, a regression based upon any related individuals can be used to estimate heritability. Sibs share the additive effects of alleles, much like parents resemble offspring because of the additive effect of alleles. We could use a correlation between sibs to predict h², however, sibs have an even greater tendency to share common environmental factors owing to their common rearing environment, and h² derived from sibs is likely to be inflated owing to shared environment. Sibs not only share additive effects of alleles, a common environment, but they also share another component of variation referred to as dominance variation that makes them resemble each other more so than they do their own parents.

Heritability of IQ

Few arguments in behavior and genetics are as contentious as those that rely on heritability estimates for IQ. Periodically, a popular book arises that argues for genetically based differences among racial or ethnic groups. These arguments invariably rely on estimates of the heritability of IQ derived from twin studies. In a sense, it is perplexing that something as well studied as IQ can remain so fiercely debated. For example, in a recent study Devlin et al (1997) amassed 212 separate analyses of familial resemblance which comprised 50,740 distinct pairings of varying degrees of relations. Estimates of heritability were derived from the correlation between: mono-zygotic twins, fraternal twins, siblings, parent and offspring, adoptive parents and offspring.

If IQ were determined by a large number of purely additive genes and did not include environmental influences, then the coefficient of relatedness could be used to predict IQ. Because twins share identical genes, a correlation between their IQ's should be close to the theoretical maximum of 1.0 which assumes no environmental influences on IQ, only additive genetic influences. The correlation for identical twins reared together is 0.85, which suggests that at least part of IQ is environmental in origin due to the common household. Perhaps, a better estimate for IQ would be provided by using the IQ of 0.75 for twins that were "separated-at-birth" and reared in different household environments. This still suggest a very strong genetic component to intelligence as indexed by IQ tests.

However, the heritability estimate based on twins still confounds a number of genetic and environmental factors. Many popular arguments that rely on twin studies do not isolate the portion of the correlation between relatives that is crucial to evolutionary arguments -- the additive genetic variation. Another component of genetic variation that is found between sibs does not contribute to evolutionary change -- dominance variation. This dominance component is often inappropriately included in the estimate of the heritability of IQ.

In order to clarify these issues, Devlin et al (1997) used a technique called meta-analysis in which results from a large number of individual studies of familial resemblance were used to estimate various genetic and environmental influences on IQ. The model which best fit the data included an additive genetic factor, a factor for dominance variation, a factor for rearing household, and factor for pre-separation environment. The pre-separation environment, which has been neglected in previous twin studies, includes the common womb environment that twins would simultaneously share prior to separation by adoption at or shortly after birth. Devlin et al (1997) found that 20% of the variation in IQ was explained by a common womb environment. While non-twin siblings share the same womb, they do not share it at the same time, but sequentially share the womb. Only 5% of the overall resemblance in IQ between ordinary siblings could be linked to this interesting maternal factor. Devlin et al's study shows that variation in IQ has more to do with development in the womb than previously thought. They suggest that many environmental agents are likely causes of this environmental effect. For example, nutritional state, smoking, and alcohol consumption by the pregnant mother are all known to lower IQ.

From their results (1997) it is clear that previous estimates based on twins have greatly overestimated the heritability of IQ (e.g., as high as 60-80%). The heritability estimate which excludes the confounding factors of maternal womb environment and common household environment is on the order of 0.48. Finally, many IQ studies do not make an important distinction between the total genetic variance in IQ and the additive genetic variance. Any estimate of IQ from siblings confounds the additive genetic component of IQ with the genetic component in IQ that arises from dominance variation. All full siblings are expected to resemble one another a little more than other family members owing to the possibility that they inherit a similar dominance configuration from their parents (see side box on dominance). While full siblings share this additional genetic source of resemblance, it is not transmitted between parent and offspring. Thus, dominance variation is not an important source of genetic variation underlying evolutionary change in intelligence. The amount of additive genetic variation for IQ is estimated at 0.34 (excluding dominance variation), which is still substantial but far lower than the amounts generally purported for this interesting human behavior trait. Moreover, the identified environmental factors such as the womb environment (5-20%), and the common household environment (17%) total 22-37%, nearly the same as additive genetic factors. Needless to say, improvement in human IQ could be brought about by pre-natal and post-natal intervention.

Table 2.3. Correlations for IQ among various degrees of familial resemblance (table modified from Devlin et al, 1997). Note* : While the coefficient of relatedness is 0.5 for each parent, if IQ were entirely due to additive genetic factors we could exactly predict IQ of offspring from parents so we have the following total quantity of information 0.5 + 0.5 = 1.0 (Li, 1965).

  Side Box 2.6: Additive versus Dominance Variation

Relatives vary in the proportion of alleles that they share in common, and the coefficient of relatedness is a measure of the probability that an allele is in common between relatives. For example, the coefficient of relatedness is 0.5 for one parent and offspring, 0.5 for sibs, 0.25 for half sibs, and 1.0 for dizygotic twins. By sharing alleles, relatives end up sharing the additive effects of individual alleles. While the relationship for the proportion of shared alleles between a parent and an offspring is exact, the relationship between sibs is probabilistic. Offspring get exactly one half of their genes from a single parent. However, it is theoretically possible that full sibs could share none of their alleles, one of each pair of alleles, or both pairs of the alleles (Li, 1965). How could this be? First we can label alleles that parents could potentially give to their children: 1, 2 from their mother and 3, 4 from their father. Likewise, the situation where sibs share no genes occurs with probability one-quarter for any given gene. The most likely situation is when sibs share a single allele of a pair of alleles, which occurs with probability of one-half. Finally, if offspring happen to share both copies of their alleles in common they also share any dominance relations between the two pairs of alleles that they inherit. This occurs with probability one quarter for any given gene.

If sibs happen to share both pairs of alleles 1 and 4, and 1 happens to be dominant to 4, then the sibs still inherit the additive effect of the genes as well as the dominance variation. Sibs still resemble each other largely because they share the additive effects of genes. However, sharing a similar dominance configuration also increases the resemblance between sibs relative to the resemblance between parents and offspring. Sibs resemble each other more so than they do their parents. This additional genetic component of variance is why some sibs appear to be identical twins. The additional dominance variation that sibs share (no other relationship shares this dominance variation) is also shared by twins. While full sibs share dominance variation, offspring do not share dominance variation with their parents, only additive genetic variance.

<END SIDE BOX>

Environmental Effects on Phenotype

We will end our discussion of genetics on the importance of environmental factors. Up to this point in our consideration of the genetic causes of behavior, we have treated the environment as if it were some random factor that obscures the genetic transmission of behavior. However, the environment can interact with genetic causes in interesting ways. In particular, the genotype can be relatively fixed (e.g., little variation among individuals), but the genetic machinery of development can still allow the phenotype to develop into alternative types. We have already seen how genetic differences among individuals can lead to alternative male phenotypes in marine isopods and the ruffed grouse. The environment per se can also trigger alternative developmental pathways that transform a juvenile into different morphologies which have alternative behaviors (Smith-Gill 1983).

Condition-dependent strategies and alternative male types

Many alternative male types are thought to be condition-dependent strategies in which an organism's internal physiological (e.g., nutrition, age, or body size) interacts with environmental factors (e.g., food availability, hatching date, etc.) to alter development of morphology and behavioral. Differences in morphology and behavior are not governed by genetic factors such as alternative alleles. Blue-gill sunfish have three alternative male phenotypes, however, the attempt to isolate alternative alleles that control such behaviors have not been successful (Gross, 1984; Gross, 1991). It is thought that when males mature into three different-sized male types with different behaviors, the trigger for the three states, is the overall body-size at maturation. Males in good condition may mature later, and grow into a large territorial male, that defends a small area in which females deposit their eggs. The male is also parental and takes care of the brood. Males that somewhat smaller, mature earlier, and develop into a female mimic that attempts to fertilize the females eggs as the female attempts to oviposit them on a territorial holding males nest. Finally, the smallest male type is thought to mature rapidly, and at very small size. This sneaker male darts into a breeding pairs mating ritual, and relies on confusion to obtain close access to the female and squirt sperm on the run. Evidence that these types are environmentally determined is provided by experimental manipulations of the availability of nest sites. If nesting sites are restricted, the frequency of medium and small-sized males increase, and the frequency of large parental males decreases (Gross, 1991)

<Drawing of Alternative Male Strategies from M. Gross, see lecture>

Condition-dependent strategies are also referred to as phenotypic plasticity. The phenotype of constant genotypes are responding plastically to the environment. While the alternative phenotypes are not controlled by alternative alleles, the development of the phenotype is still governed by a cascade of events that are under some kind of genetic control. The distinction is important because presumably all individuals in a population are capable of developing into the alternative phenotypes, if their environment had been conducive to these alternative developmental pathways from the outset.

Alternative Larval Types in Spadefoot Toads and Tiger Salamanders

A clear example of alternative behaviors is seen in the development of larval amphibians. The interaction between environment and the behaviors governing energy acquisition are central to the growth and development of individuals. A basic dichotomy in feeding strategies entails the distinction between carnivory and omnivory. Carnivores largely feed on animals while ominvores tend to have a more catholic diet consisting of animals, plants, and in some cases decaying plant and animal remains or detritus.

If you were to wander into the deserts of the southwest during dusk after the hot July day has been chilled by a torrential thunderstorm, you will find a small toad calling for mates. Clustered around ephemeral pools filled with water, choruses of singing male spadefoot toads try to attract females that are hopping towards the water to lay their eggs. Water is a scarce resource for the toads and they emerge with the first moistening of soil. Successful adult males clasp the back legs of females in a tight amplectic embrace to stimulate the female to oviposit their eggs. Fertilized eggs are left behind in the pool to develop. As you might have already guessed, water does not last long in the glare of the summer desert sun. Tadpoles that hatch from egg masses must develop rapidly into the terrestrial adult form if they are to survive.

<Drawing of spadefoot toad lifecycle from Pfennig, see lecture>

If a tadpole, happens to hatch in a pond that is rich in a key resource, the fairy shrimp (a.k.a, Artemia salina or sea monkeys), the tadpole might ingest enough shrimp and its development can be profoundly altered (Pfennig 1990, 1990a). Tadpoles that eat a lot of the shrimp are triggered to develop into a carnivore with specialized keratinized tooth and large jaw musculature that is efficient for feeding on more shrimp. Carnivores patrol the pond in a solitary fashion searching for their next meal. If however, the tadpoles lands in a pool with few or no shrimp they develop into an omnivore with a rasping jaw structure, a huge gut that allows for efficient digestion of plant remains. Omnivores tend to swim in schools and the schooling behaviors is thought to allow the group to stir up food more efficiently. Individuals are capable of either the carnivorous or omnivorous morphology and behavior. They only need to ingest enough shrimp to develop into the carnivorous form.

<picture of carnivore and omnivores from Pfennig, see lecture>

What is the key environmental factor in the shrimp? Shrimp are naturally loaded with the potent hormone thyroxine and shrimp also have a heavy does of iodine a key component of thyroxine (Pfennig 1992b). Thyroxine is a key metabolic hormone found in all vertebrates, and also governs development of juvenile amphibians, and controls the metamorphosis from tadpole larva to terrestrial adult. By ingesting large amounts of this hormone, the tadpoles precociously develop structures that allow for a more carnivorous lifestyle, structures that are incidentally more reminiscent of the adult form. David Penning performed critical experiments to show that high levels of thyroxine precociously triggers development of the carnivores.

Carnivores develop much more rapidly than the omnivore and carnivores metamorphose into toadlets in a shorter period of time. The omnivore must consume greater quantities of food to reach the size and stage required for metamorphosis to a terrestrial toadlet that can escape the desiccating pool. Carnivores have a tremendous time advantage in small ephemeral pools. Omnivores take longer to develop, but the tend to metamorphose at a slightly a larger size, and with greater fat reserves. Omnivores are successful in less ephemeral ponds. Moreover, the fairy shrimp is much more abundant in the smaller faster drying ponds. The plasticity in development of behavior and morphology is adaptive in that a female can leave progeny in a small fast-drying pond or a larger long-standing pond, and the tadpoles environment, presence or absence of shrimp determines the offspring's development.

Study Questions for Chapter 2

1. What additional genetic factors increase the resemblance between sibs besides additive genetic variation?

2. Describe a mutation experiment that was used to isolate a gene. Why is it important to show that mice defective in this gene can perform non-nurturing tasks like maze running or olfaction tests? What kinds of effects do these additional tests detect? Why are we interested in a lack of effect of a gene on other traits?

3. Why are heritability estimates for sibs raised in different households preferable to sibs raised in the same household? What additional confounding factors does this design not remove from the heritability estimates?

4. You overhear someone in the coffee shop citing the following evidence regarding heritability of IQ: "IQ must be very heritable, the correlation between fraternal twins reared by the same family is 0.59." "Hah!" the other combatant exclaims, "The correlation between an adopted child and their unrelated sib is 0.20."

5. Compare and contrast genetic determination and condition-dependent determination of as proximate causes of alternative male strategies.