Bridging the Gap Between Developmental Genetics and Paleontology



The mechanisms by which variation within populations is translated into differences between populations, species, and higher taxa is one of the central problems in evolutionary biology. According to neo-Darwinian theory, natural selection and other population processes can alter the frequencies of phenotypes within populations. Combined with long periods of time, these processes provide a sufficient explanation for the divergence between related groups. However, the time scale at which species differences evolve is generally too long to study using contemporary populations and too short to be resolved in the fossil record. Additionally, intraspecific variation is usually small compared to interspecific divergence. The disjunction between the scales of time and phenotypic disparity in studies of populations and patterns of change in the fossil record or divergence among species has led some to propose that neo-Darwinian theory is incomplete.

The threespine stickleback fish, Gasterosteus aculeatus, exhibits major morphological differences both within and among populations. These differences are comparable to interspecific differences in many groups of organisms and clearly can evolve over short periods of time. Thus, the threespine stickleback provides an excellent opportunity to study the evolution of "species-level" phenotypic differences using contemporary populations in which genetic and phenotypic properties, demography, and ecological context can be assessed. Plus, its extended geographical range and fossil record permit placement of process at the population level into a larger evolutionary context.


The threespine stickleback is widespread in coastal marine and lowland fresh waters of the boreal and temperate northern hemisphere. It comprises a complex of morphologically divergent populations and biological species that occur in ecologically diverse habitats. Threespine stickleback are small and abundant, and large samples can be collected without adversely affecting local populations. Populations exhibit striking phenotypic differences that are correlated with habitat type. Their husbandry was developed during the first half of the 20th century by European ethologists, and they are relatively easy to maintain in the laboratory. They can be crossed using in vitro fertilization, so a variety of cross designs can be executed to investigate their genetics. Their generation time in the lab is about nine months to a year, and crosses between parents from even the most highly divergent populations are viable and fertile. Thus, experiments involving multiple generations and members of any pair of populations are feasible.

In our lab, we often compare populations of threespine stickleback to infer morphological function or evolutionary mechanisms. Similarities among populations of a species reflect both the effects of common ancestry and evolution since the most recent common ancestor. Distinguishing these influences on the phenotypic properties of related populations is a serious problem in comparative biology that can be minimized in the threespine stickleback. In recently deglaicated regions, freshwater threespine stickleback populations have been derived from marine and anadromous (sea-run) populations (collectively called "oceanic" within a relatively short period of time. Since marine and anadromous populations can disperse readily through the sea, but dispersal among freshwater drainages is slow and improbable, it is likely that postglacial populations from different drainages were derived independently from a common oceanic ancestor. Oceanic stickleback from different populations are not identical, but they exhibit limited phenotypic and genotypic variation compared to their freshwater descendants, from which they differ greatly. Thus, any oceanic population within a geographical area approximates ancestral phenotypes and genotypes for the local freshwater populations. Freshwater populations that are from separate drainages but share phenotypic traits that are absent in the common oceanic ancestor must have evolved those similarities independently. Since threespine stickleback occur in many lakes, large numbers of populations with similar phenotypes can be assembled to form samples of statistically independent observations in comparative studies. Repeated independent evolution of freshwater phenotypes from oceanic ancestors can be thought of as a highly replicated natural experiment. Repeated derivation of freshwater populations from oceanic ancestors and subsequent independent evolution in fresh water is an extremely valuable property of the threespine stickleback supermodel for biological research.

During the past decade, the extreme morphological variation exhibited by the threespine stickleback has attracted the interest of developmental geneticists. Their research on the same morphological traits that we study (see below) are rapidly yielding exciting insights into the genetic architecture of traits that vary within stickleback populations and evolve rapidly. More generally, a wide array of the specialized tools to investigate the genetic architecture and DNA sequence variation underlying any phenotypic difference among threespine stickleback is rapidly maturing. In the winter of 2005, the Broad Institute in Cambridge, Massachusetts began to sequence the genome of a specimen from one of the Alaskan populations (Bear Paw Lake) that we study. The combination of biological properties of the threespine stickleback, the wealth of knowledge of its behavior, ecology and evolution, and development stickleback genomics recently inspired Greg Gibson to dub the threespine stickleback a biological "supermodel" because it will be possible to integrate research across levels of causation, from genomic variation through gene expression, development, and function to population processes.


Our research concerns the description and explanation of morphological variation within and among freshwater populations of the threespine stickleback. The foundation of our research is to describe patterns of morphological variation in natural populations. Combined with environmental information, these descriptions can be used to frame hypotheses for the function of phenotypes, the identity of selection agents, and the roles of natural selection and geographical isolation in evolution. Of course, genotypic variation has a strong influence on the phenotype, and understanding the genetic variation and developmental processes underlying morphological variation in threespine stickleback are part of our research. Morphological variation among populations can also be used to identify candidate populations for the study of novel genetic variation.

Research in my lab exploits two primary field areas. We study postglacial populations in lakes and streams around Cook Inlet, Alaska. We have sampled over 200 populations of G. aculeatus from ecologically diverse lakes. Our research as well as that carried out by others has demonstrated that diet and predation regime are important factors in stickleback evolution. We exploit two environmental contrasts to investigate stickleback evolution: (1) lakes with abundant benthic prey versus those with abundant plankton, and (2) lakes with and without native predatory fishes.

The second primary field area is a diatomaceous earth quarry near Reno, Nevada. About 10,000,000 years ago, the rock in this quarry was laid down in recognizable annual layers (varves), which allows us to convert stratigraphic distance to years. The fossil stickleback in this quarry is known technically as Gasterosteus doryssus but belongs to the G. aculeatus species complex. Gasterosteus doryssus in this quarry is abundant, well preserved, and distributed almost continuously throughout the roughly 110,000 years represented by rocks exposed in the quarry. In addition, G. doryssus evolves within this deposit. The combination of fine temporal resolution, a reasonably long span of time, abundant, well preserved specimens, and evolutionary change led the late Stephen J. Gould to comment, “A remarkable series of studies by Michael A. Bell . . . surely a summum bonum for attainable temporal precision!

RESEARCH STRATEGY AND THE ROLE OF COLLABORATION: Because information ranging from the level of DNA sequence variation to the history of deglaciation in Alaska influence stickleback morphological evolution, I have broad interests that encompass genomics, development, functional morphology, ecology, natural selection, geographical variation, and paleontology. Admittedly, this breadth of interest creates the potential to become a jack of all trades and master of none. I try to minimize that problem by working almost exclusively with the threespine stickleback. This breadth of interest creates the potential to use the threespine stickleback supermodel to examine the interactions among levels of biological causation and produce a vertical integration among levels of causation from the level of the genome to population processes. This breadth of interest also necessitates collaboration with graduate students and colleagues who possess the background and technical skills that I lack. Thus, I depend on my graduate students, my closest collaborators, to bring crucial skills into the lab, and I have several senior collaborators who are all excellent in their own specialties.

CURRENT PROJECTS: Below I briefly describe current projects in our lab and note with whom each project is being done.

  • Covariation within populations and the direction of evolution. If one visualizes the set of morphological traits of a population as dimensions in a morphological hyperspace, it is possible to draw a line through that hyperspace that represents the direction of greatest variation. Extending Fisher’s Fundamental Theorem of Natural Selection from individual to multiple traits, this is the direction in which the response to natural selection should be greatest, assuming that these traits have similar heritability and the same effect on Darwinian fitness.

    We are studying a contemporary population in Loberg Lake, Alaska, where we have observed striking evolutionary change (Bell et al. 2004) of heritable traits (Aguirre et al. 2004) during the last 16 years (as of 2005). The Loberg Lake population was founded between 1983 and 1989 by oceanic ancestors. We can calculate the direction of evolution in the morphological hyperspace for this population and investigate the direction of greatest heritable variation in the ancestral population.

    We are making annual samples from this population and studying quantative genetics (genetic covariance or G matrix) of morphological traits in a local anadromous population from Rabbit Slough, which we are treating as the ancestor. We also plan to study the G matrix of the Loberg Lake population to compare genetic covariance structure in the ancestor and descendant lake populations. The genetic and evolutionary components of this project are being done in collaboration with F. James Rohlf and Windsor E. Aguirre. Our collaborator at the University of Alaska, Anchorage, Frank A. von Hippel, is studying assortative mating between the Loberg Lake and Rabbit Slough populations to see whether behavioral isolation has evolved that would make the young Loberg Lake isolate and ancestral anadromous populations separate biological species.

  • Stratigraphic variation - multivariate phenotypic variation and evolution. The diatomaceous earth quarry in Nevada contains a stratigraphic section comprising about 110,000 years of sediment. Stickleback at the base of the stratigraphic section in this quarry are highly divergent from typical threespine stickleback. They have a small vestige of the primitively robust pelvis and the dorsal spines are reduced from the ancestral number of three to one or zero. During the ensuing 93,000 years, this "low-armored" stickleback lineage (see figure) experienced moderate evolutionary change (Bell et al. 1985). Then it was joined by a second "spiny" stickleback lineage with the ancestral condition for dorsal spine number (i.e., 3) and pelvic structure (robust girdle with large pelvic spine; see figure). There was limited or no hybridization between these fossil stickleback species, as is observed in modern pairs of stickleback species. The low-spined form ceased to occur within about 100 years, and the spiny form persisted alone for another 21,500 years within the stratigraphic section. During this time, mean dorsal spine number declined from three to about one, and pelvic structure was reduced to a small vestige or may even be lost entirely. Matt Travis and I are studying rates and patterns of evolution of armor phenotyes within the spiny species. We showed that the evolutionary rates are not high enough and the direction of change is not consistent enough to exclude genetic drift as the cause of change. However, the loss of armor observed is consistent with absence of predatory fishes in the fossil deposit. This paradox indicates that rates and patterns of change in the fossil record be used alone to detect natural selection. With Mark Purnell and Paul Hart at the Univesity of Leicester, Matt Travis and I are investigating differences in tooth wear in modern Alaskan stickleback populations with known diets and fossil stickleback from this deposit. This work is in progress.With F. James Rohlf, Matt Travis and I are using the second spiny lineage to study the evolution of trait covariances. This study has produced some results that clearly are not artifacts but are difficult to interpret mechanistically. This work is in progress.

  • Genomics of evolutionary divergence. Our laboratory is not equipped to perform genetic analyses, and this work is done in the laboratories of collaborators. In the late 1990’s I worked with John Postlethwait, Charles Kimmel, and William Cresko at the University of Oregon on the genetic basis of armor reduction in lake populations from Cook Inlet (Cresko et al. 2004). This work also led to a study by Kimmel et al. (2005) on the genetic and developmental basis of shape change in the operculum of Alaskan freshwater stickleback. Since 2005, we have begun to work more closely with David Kingsley’s laboratory at Stanford University School of Medicine on stickleback genomics and development. Our current projects grew out of a common interest in the potential for variation in the genetic architecture for pelvic girdle reduction among threespine stickleback populations. Studies by Cheryl Tickle’s, Kingsley’s, Postlethwait’s groups had all implicated a transcription factor, Pitx1, as the major locus for evolution of pelvic reduction in G. aculeatus. Silencing of Pitx1 expression is associated with extreme pelvic reduction, but also has the interesting tendency to produce pelvic phenotypes in which the left vestige is larger than the right. Vickie Khalef, an undergraduate honors student working in our lab, showed that the left pelvic vestige is significantly more likely to be larger than the right in asymmetrical pelvic phenotypes from 20 of 27 Cook Inlet threespine stickleback populations, implicating silencing of Pitx1 in pelvic reduction. However, in three of the 27 populations, the right vestige is likely to be larger than the left, and a large sample from one population had equal frequencies of asymmetrical specimens with larger right or left pelvic vestiges. We reasoned that some other gene than Pitx1 is the primary cause for pelvic reduction in these populations. We have two studies under way to investigate this possibility. During the summer of 2005, Katie Ellis, another undergraduate honors student in our lab, and I performed a series of intrapopulation crosses using specimens from populations without pelvic reduction as controls. We also used specimens with extremely reduced pelvic phenotypes from populations with pelvic reduction and the usual left-biased pelvic asymmetry, populations with pelvic reduction and the unusual right-biased asymmetry or no asymmetry, and populations in which the pelvic spine is commonly missing but the girdle is intact. Fry from these crosses were preserved at developmental stages when Pitx1 is expressed in populations with normal pelvic phenotypes. A plasmid supplied by Kingsley’s lab (and produced with support of the Howard Hughes Medical Institute to Kingsley) was used to make a probe for pelvic girdle expression. Working in Howard Sirotkin’s lab at Stony Brook University, Katie Ellis is studying variation of Pitx1 expression in these populations. Some of the same populations with right-biased and non-directional pelvic asymmetry were used in F1 hybrid crosses with anadromous stickleback with normal pelvic expression in 2005 to produce F1 hybrids. The zygotes from these crosses were shipped to Kingsley’s laboratory, where they are being reared. When they become sexually mature, brother-sister crosses will be performed to produce F2 hybrids for use in linkage mapping.

  • Ontogeny of feeding performance in divergent threespine stickleback populations. This project is Matthew Travis' dissertation research. He is interested in differences in the morphology of the jaws, hyoid apparatus, and other structures involved in food capture. He is using three benthic-feeding, three planktivore, the young Loberg Lake (see above), and anadromous stickleback. Although static morphological differences are of interest, the focus of this research is the dynamics of food acquisition. Specimens ranging from 1 cm long fry to adults are used to study kinematics of feeding. Matt records high-speed digital motion images of landmarks on the jaws and head and the motion of food particles to infer the relationship of trophic form and mechanics to food acquisition.

  • Spatial cognition, neuroanatomy, and genetics. This research is our latest project, and Peter Park is doing the laboratory work for this study. This project will become his dissertation research project. It has long been known that birds that hide (cache) seeds are much better at learning and remembering (spatial cognition) where food has been hidden compared to close relatives that do not cache. Similar results have been obtained for mammals. The hippocampus, the region of the brain for spatial cognition in both groups, is larger in the species or sexes in which spatial cognition is more important. Interpretation of these associations has been controversial.

    Fig. 1. (left) Contrasting body and brain (inset) shape in limnetic (A) and benthic (B) threespine stickleback.

    I hypothesized that stickleback that eat benthic prey should have better spatial cognition than planktivores because the location of plankton is highly unpredictable and planktivores should have less use for spatial cognition. Peter Park’s preliminary data indicate that there are differences in the structure of the threespine stickleback brain in the area implicated in spatial cognition, and that individuals from benthic-feeding populations learn to solve mazes more efficiently than planktivores do.

    This research is in a preliminary state. Peter plans to determine the cellular differences responsible for variation in brain shape among populations. This will allow him to determine whether similiarities in spatial cognition have the same or a different cellular basis in different populations with similar spatial cognition and gross brain morphology. Once the cellular basis for differences in spatial cognition in stickleback is understood, it will be possible to use hybrids between planktivore and benthic-feeding stickleback to use linkage mapping to infer the genes responsible for differences in spatial learning ability and brain development between benthic-feeding and planktivore threespine stickleback populations.

    Fig. 2. (right) Principal component analysis of dorsolateral telencephalon (blue) shape in threespine stickleback. Plots on the inset grids represent telencephalon shapes at the adjacent PC maxima compared to its shape at the origin (0,0). PC1 mostly represents the longitudinal position of telencephalon lateral expansion; PC2 mostly represents the magnitude of lateral expansion. Anadromous and benthic populations are restricted to the upper right quadrant, corresponding to an anterior position (PC1) of moderate telencephalon expansion (PC2).

  • Other projects. A number of related projects that support our interests in the evolution of phenotypic diversity are in progress. (1) I am studying a fossil threespine stickleback from a marine deposit. It appears to represent a marine G. aculeatus with highly armored morphology, which is comparable to that of modern marine threespine stickleback. However, this specimen predates the oldest described fossil threespine stickleback. This specimen will be important to calibrate molecular clocks for the threespine stickleback, (2) Matt Travis is studying diet in several Cook Inlet populations of G. aculeatus. Up until now, we have inferred diet from the trophic morphology and body form of the stickleback populations and lake bathymetry. However, we had never examined their stomach content to verify these inferences. Stomach content analysis largely confirms our inferences, but there have been some surprises. (3) Windsor Aguirre is using microsatellites to study the genetic structure of a system of populations in one Cook Inlet drainage to infer the level of genetic similarity among populations and its potential to explain phenotypic similarities. He is also using microsatellites to infer whether morphologically constrastsing individuals from the same lake represent one or two demes. (4) Windsor Aguirre is also studying shape variation among years in anadromous populations of G. aculeatus from Rabbit Slough, a population we have used for most of our genetic crosses.

DISCLAIMER: The research described in this website was performed with grants from National Science Foundation, which requires the following statement: "Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the National Science Foundation."