Evolution of Epistatic Networks and the Genetic Basis of Innate Behaviors

Abstract

Instinctive behaviors are genetically programmed behaviors that occur independent of experience. How genetic programs that give rise to the manifestation of such behaviors evolve remains an unresolved question. I propose that evolution of species-specific innate behaviors is accomplished through progressive modifications of pre-existing genetic networks composed of allelic variants. I hypothesize that changes in frequencies of one or more constituent allelic variants within the network leads to changes in gene network connectivity and the emergence of a reorganized network that can support the emergence of a novel behavioral phenotype and becomes stabilized when key allelic variants are driven to fixation.

Evolution of complex innate behaviors: a genetic conundrum

Innate (“instinctive’) behaviors are genetically programmed stereotypical behaviors that occur independent of previous experience and can occur spontaneously or be set in motion by environmental triggers. One example of an innate behavior is nest building by birds, where nest architectures are characteristic for each species, but distinct across species [1–3]. Web building spiders present a similar example, where web construction is stereotypical within a species, but shows variation among species [4,5]. A more complex example is nest building by ants. Many species of ants, such as the leaf cutter ant, Atta laevigata [6], and the Florida harvester ant, Pogonomyrmex badius [7], construct extensive and elaborate subterranean nests consisting of numerous chambers connected by corridors or vertical shafts.

The observation of widespread species-specific innate behaviors raises the question as to how complex genetic programs that regulate such species-specific stereotypical behaviors evolve. There are two alternative possibilities. One could imagine that specific genes have evolved to direct the genetic programs underlying each distinct behavior. To date, extensive genomic studies have presented scant evidence for such a hypothesis. For example, the FoxP2 gene, which encodes a forkhead-domain transcription factor has been implicated in brain development and synaptic plasticity. Mutations of the human FOXP2 gene cause severe speech and language disorders, which identified this gene as a critical gene for the development of language [8,9]. However, this gene did not arise de novo in humans or even in the primate lineage, but is also found in songbirds, fish and reptiles [10,11]. In mammals, the FOXP2 gene is highly conserved and shows a signature of recent positive selection [12]. Thus, it is the evolutionary trajectory of a preexisting gene and its genetic context that has enabled the expression of a novel quintessential human behavioral trait, communication through language. This supports a second more plausible theory that evolution of species-specific innate behaviors is accomplished through modification of genetic networks by altering expression patterns, frequencies and gene-gene interactions of naturally occurring allelic variants.

The transcriptional niche

The theory that modification of genetic networks underlies the evolution of complex behaviors is supported by extensive evidence. A critical assumption of this theory is that genes act as interdependent ensembles [13–21]. Studies on Drosophila have shown that disruption of expression of a focal gene will result in up- or down-regulation of a suite of coordinately expressed genes [22], which has been named its ‘transcriptional niche’ [23,24]. Transcriptional niches do not reflect single cellular pathways but may encompass transcripts with diverse gene ontology designations that together contribute to the manifestation of the phenotype. The extent of a transcriptional niche depends on the statistical criterion for significance of coordinated expression. When a polymorphism arises it affects not only the expression and function of the gene in or near which it occurs but will create a ripple effect that affects the organization of the focal gene’s transcriptional niche.

Epistasis

The transcriptional niche is not only defined by coordinate expression of its associated genes but is also influenced by epistatic interactions [22]. Epistasis, which in quantitative traits is defined as non-linear interactions between genes where the effect of one locus depends on the genotype of its interacting partner, is a common feature of the genetic architecture of complex traits [25–28], including behavioral traits [24,29]. Gene-gene interactions have been demonstrated in mice for nest building behavior [30], alcohol drinking behavior [31,32] and a variety of anxiety related traits, including hyperactivity, prepulse inhibition, and conditional fear [33]. Epistasis has also been identified in studies on zebrafish by crossing a standard laboratory strain to fish from a wild population and measuring boldness behavior [34]. The genetic underpinnings of foraging behavior of honey bees also displays epistasis [35,36], as does thermal preference behavior in Caenorhabditis elegans [37]. Epistatic interactions are plastic and can be modified by environmental changes [38,39].

The most comprehensive studies on epistasis come from the Drosophila model. Studies on Drosophila have shown that epistasis is a predominant feature of the genetic underpinnings of quantitative traits [27,40], including behaviors, such as olfactory behavior [38,41–43], climbing behavior [44], aggression [45,46] and startle behavior [47,48]. Furthermore, large-scale mutagenesis studies showed that behavioral phenotypes present large mutational targets, implying extensive pleiotropy [38,49–52]. In addition, studies on inbred mouse strains have provided evidence that epistatic interactions contribute to genetic variation in pleiotropy [53].

Epistatic networks of pleiotropic genes

Consistent with the notion of pleiotropy, genome-wide association studies on the Drosophila melanogaster Genetic Reference Panel [54,55] have identified hundreds of polymorphisms associated with variation in any behavioral phenotype [42,46,56,57]. Subsets of genes harboring these polymorphisms can be organized in genetic interaction networks. Gene ontology enrichment analyses shows that networks that underlie variation in behavioral phenotypes encompass many genes associated with development and function of the nervous system [42,46,56,57]. Thus, subtle variation in neural connectivity is likely to underlie phenotypic variation of any given behavior. As expected, considering the extensive pleiotropy, networks associated with different behaviors overlap in their composition, but are distinct in their connectivity. Thus, the network architecture determines the specificity of the behavioral phenotype.

Variation of epistatic networks in a population

Polymorphic gene networks are derived from genome-wide association studies based on phenotypic variation. Thus, these networks are not identical in every individual, but represent average representations of the genetic interactions in the population. Allelic differences among individuals generate variation in connectivity around the basic network structure (Figure 1A). Major network hubs, however, are likely to be preserved across individuals.

Figure 1. Evolution of a novel phenotype through remodeling of a genetic network.

A. A hypothetical genetic network that gives rise to variation in a behavioral phenotype. The diagram depicts subtle variations in network connectivity among different individuals within the population. Genes are indicated by pentagons and the strength of connectivity between interacting genes is reflected by the thickness of the edge. The gray and black symbols represent hub genes in the network. The purple pentagon represents an allelic variant that will drive remodeling of the network illustrated in B and C, which show the progressive increase in frequency of this allele, reflected by the increased color intensity. Drift or selection on the purple allele leads to remodeling of the network in which the purple allele becomes a novel hub gene. Concomitant with this process, connectivity of the gray hub diminishes until ultimately it is no longer recognizable as a hub gene. The network in C can drive a novel behavioral phenotype distinct from that associated with the network architecture of A. This novel network again is representative of a collection of related networks that harbor subtle variations, as depicted in A.

Changes in frequencies of one or more allelic variants due to genetic drift or under the influence of selective forces may modify transcriptional niches, which may result in changes in the connectivity of the genetic interaction network (Figure 1B) and lead to the emergence of a novel network (Figure 1C). Genetic interaction networks that regulate the expression of complex innate behaviors can become stabilized when positive selection that results in enhanced fitness drives key allelic variants in the network to fixation (Figure 2). Epistatic interactions between fixed alleles and interconnected partners would confer robustness to the network [58]. Fixation of allelic variants that represent multiple hubs in genetic interaction networks can thus serve as anchors for the stabilization of interconnected genes. Reduced connectivity of allelic variants of connected genes might result in destabilization of the network with a consequent deleterious effect on fitness. It is in this context of interest that suppressing epistasis buffers the effects of newly arising mutations [48].

Figure 2. Stabilization of a genetic network by positive selection on allelic variants at critical positions within the network.

Fixation of alleles (indicated by spheres) is illustrated by the transition from key to blue to red colors concomitant with the rise of a steep fitness peak associated with the network.

Stabilization of genetic networks and the fitness landscape

Stabilization of genetic networks by fixation of allelic variants at key positions in genetic interaction networks is reminiscent of the fitness landscapes described by Wright [59], where fixation of key alleles drives the genetic interaction network upward into a distinct peak surrounded by a fitness valley that has become so steep that it becomes virtually fixed in the evolutionary trajectory, i.e. in Waddington’s concept the genetic program becomes canalized [60] (Figure 2). A stabilized genetic network, represented by an exceptionally steep fitness peak in the genomic landscape, can support stereotypical behaviors with low evolvability.

The reorganized network can enable the manifestation of a novel behavioral phenotype or remain cryptic providing an Anlage that can support the future emergence of new behavioral phenotypes. For example, reorganizing the structure of an ancestral network could lay the genetic underpinnings to support nest building behavior within avian clades while allowing species-specific variations in nest architecture. Positive selection on gene interaction networks can occur by enabling more effective predation (e.g., spider web construction [5]), more effective protection from predation (e.g., arboreal nesting [1]), or through sexual selection [61,62].

Concluding remarks

The notion of behaviors as emergent manifestations of epistatic networks of pleiotropic genes [24] together with the concept of individual variation in network organization as the substrate for evolutionary change can explain the emergence of species-specific innate behaviors. Whereas the versatile Drosophila model enables the most comprehensive genetic studies to explore this premise, these principles apply widely across species to explain the genetic underpinnings of the evolution of fixed action patterns of behaviors from courtship and aggression in sticklebacks to egg retrieval behaviors of greylag geese, described in Tinbergen’s classic studies [63], as well as the evolution of human language, mentioned earlier. The theory proposed here also applies to canalization of developmental processes [60] where evolutionary consolidation provides the neurobiological foundation for the expression of innate behaviors. The theory that new behavioral phenotypes evolve through the combined effects of altered network connectivity in response to allelic variation and stabilization of the network by selection will have to be corroborated through comparative genomic studies across lineages of related species that exhibit similar or divergent innate behaviors. However, the hypothesis presented here provides a framework for such studies (see Outstanding Questions).

Outstanding Questions.

Behaviors are the ultimate expression of the nervous system. Thus, understanding the evolution and manifestation of innate behaviors requires a neurogenetic approach.

How do changes in genetic architecture and global gene expression result in changes in neuronal connectivity?

Does this involve changes in neural connectivity or changes in synaptic strengths?

What is the exact nature of the forces that result in changes in allele frequencies, e.g. what would be the contribution of drift relative to positive selection?

Could balancing selection contribute to the stabilization of allelic networks, while maintaining some variation to confer plasticity as environmental conditions evolve?

Whereas many species (e.g. birds or ants) build elaborate nests, the nest structures can differ widely among them. Could nest building per se be viewed as an overarching behavior that can be accomplished through the stabilization of multiple allelic networks within a common genomic signature?

Highlights.

• Innate behaviors are genetically programmed and occur independent of previous experience

• Genetic interaction networks form a framework for the evolution of innate behavior

• Changes in frequencies of one or more constituent allelic variants within genetic interaction networks lead to changes in network connectivity

• Stabilization of key allelic variants within networks through positive selection supports the emergence of novel behavioral phenotypes

Glossary

allelic variants

sequence variants of a gene at a specific locus

balancing selection

natural selection in which heterozygotes have increased fitness relative to both homozygotes

fitness

the genetic contribution of an individual to the next generation

fixed action patterns

instinctive behavioral sequences that are relatively invariant within the species and almost inevitably run to completion

genetic drift

random variation in allele frequency caused by sampling in finite populations

genome-wide association studies

studies which seek to correlate variation in a trait with polymorphisms across the entire genome in an outbred population

pleiotropy

affecting multiple phenotypes

polymorphism

naturally occurring DNA variant among individuals in a population as a result of a spontaneous mutation

positive selection

increase in frequency in the population of a beneficial allele