Home People Faculty Jeffrey K. Conner Research Description
Jeffrey K. Conner's Research PDF Print E-mail

1. Evolution of floral morphology

A primary research interest is in the evolution of functional integration among traits, that is, how do groups of traits such as an organ or organ system evolve to work together to perform a single function? The flower is an excellent model system to address this question. Flowers are made up of a number of different parts (petals, pistils, stamens, etc.) that must work together to perform one crucial function: sexual reproduction.

Since each of the floral parts is itself a complex trait, affected by many gene loci and the environment, the tools of quantitative genetics are appropriate for this research. Genetic integration among any two complex traits can be quantified as the genetic correlation, which measures the degree to which the two traits are affected by the same genes (pleiotropy) or the same groups of genes (linkage disequilibrium). If there is selection for increased functional integration between two traits, then an increased genetic correlation can evolve, so that the two traits are then inherited together.

This evolution of increased correlation for functional integration may have occurred between the lengths of the filaments and corolla tube in several species of Brassicaceae, including my main study species, wild radish (Raphanus raphanistrum).

The correlation between filament and corolla tube determines the position of the anthers relative to the opening of the corolla tube, often referred to as anther exsertion (Fig. 1). Since the insect pollinators remain in the open part of the corolla, flowers with anthers that are placed too high or two low may not be as successful at placing their pollen on insects as flowers with intermediate anther placement. If so, this would create selection for an increased correlation between filament and corolla tube lengths to maintain this intermediate anther position regardless of flower size.

conner1

Figure 1. Schematic cross-section of a wild radish flower, showing the corolla tube (a false tube formed by the four petal claws) and one of the four long and one of the two short filaments. Anther exsertion is defined as filament length minus corolla tube length.

We have found that this correlation is significantly greater than the average correlation among floral parts in wild radish and five other species in the mustard family (Conner and Sterling 1995; Conner et al. unpub.). Comparative phylogenetic analysis (in collaboration with Anna Wiese and Alan Prather at MSU) indicates that the high correlation evolved from an ancestral lower correlation, but likely only twice in the large family Brassicaceae (Fig. 2). In the studies outlined below (funded by two grants from NSF and one from USDA), we have used the correlation between filament and corolla tube in wild radish as a model system for studying the evolution of genetic correlations and constraints on adaptive evolution.

conner2

Figure 2. Preliminary phylogenetic tree of the family Brassicaceae, with Cleome and Polanisia representing the Capparaceae (outgroup). Patterns of correlation between filament and corolla tube are indicated with open and filled boxes, and the phylogeny of this trait was mapped onto the tree using MacClade. A filled box means that the filament -corolla tube correlations are significantly greater than the rest of the floral correlations. Note that six of the species have no boxes; we have yet to measure correlations in these species. Most of the phylogeny is based on a single minimum-length tree from parsimony analysis of a 780 bp sequence from the chloroplast gene ndhF (Price, unpub.). Measurements of selection on correlations: To see whether plants with intermediate anther position do have higher fitness as hypothesized, we measured selection on anther position and several other floral traits in three field seasons in Illinois. Since anther position should mainly affect male fitness, that is, the ability to sire seeds on other plants, measurements of male success were necessary. Indeed, theory predicts that most selection on floral traits should be through differences in male fitness rather than female fitness (seed production), but since male fitness is so difficult to measure this prediction has gone largely untested. Lifetime male fitness was determined by molecular genetic paternity analysis of over 6500 offspring. Plants with intermediate anther position did have the highest fitness in at least one of the three years (Morgan and Conner 2001), suggesting that selection could have caused the increased filament-corolla tube correlation in wild radish. More selection on floral morphology occurred through differences in male fitness than female fitness, in agreement with the theory mentioned above (Conner et al. 1996a; Morgan and Conner 2001).

Artificial selection on anther position: The selection on anther position in only one of three field seasons provides only modest support for my adaptive hypothesis. However, the high correlation between filament and corolla tube means that there is little variability in anther position, so the fitness consequences of having extremely high or low anthers are difficult to observe. This is an example of a central problem in evolutionary biology: attempting to reconstruct what may have happened in the past by studying present-day populations.

In collaboration with Keith Karoly at Reed College, we performed six generations of artificial selection designed to create more variability in anther position. This is an attempt to "turn back the evolutionary clock", by recreating the ancestral condition, which is a lower filament-corolla tube correlation. We selected for increased anther exsertion in two lines, decreased anther exsertion in another two lines, and had two randomly-mated lines as controls. Results indicate responses to selection in the predicted directions: anther exsertion increased in the two lines selected to increase, and decreased in the two lines selected for a decrease. When these lines are combined together with the two random-mated control lines, the result is a composite population with the ancestral lower correlation (Fig. 3). In the summer of 2001, the plants resulting from these selection experiments were used in studies of selection through male fitness similar to those described above, to determine the fitness consequences of more extreme anther positions than exist in present-day populations. We are currently performing paternity analysis with microsatellite markers on the offspring to determine selection through male fitness.

conner6

Figure 3. Results for the artificial selection on anther exsertion experiment. The left panel shows the half-sib family means in the base population, with the arrows depicting the direction of selection applied, and the right shows full-sib family means in the lines after five or six generations of selection. Note that the correlation in the composite population dropped from 0.83 to 0.40, and that the high and low lines are displaced at right angles from the major axis of the correlation (where the controls lie). Interestingly, the correlations within the high, low, and control lines were unchanged relative to the base population, ranging from 0.76 to 0.85 (see Conner 2003 Ecology for further discussion). The different ranges in the two plots are due to different estimation methods.

Evolution of dimorphic anther height: The studies outlined above focus on the position of the long filament anthers relative to the opening of the corolla tube, but like most of the over 3,000 species in the family Brassicaceae, wild radish has four long and two short stamens (Fig. 1). Therefore, each flower is dimorphic for anther height. In a parallel series of studies, also in collaboration with Keith Karoly, we are exploring the reasons for the evolutionary stasis of dimorphic anther heights across this large plant family. One possible explanation is that there is no additive genetic variation for this dimorphism, so it cannot evolve. We have shown that this is not the case in wild radish, both using sibling analysis (Karoly and Conner 2000) and artificial selection for decreased dimorphism. Another possibility is that there is selection for dimorphism, i.e., it is an adaptation. In work with an REU in my lab, Amber Rice, and Martin Morgan of Washington State, we showed that dimorphism in anther height led to higher male fitness (seed siring success) in one of three years (Conner et al. 2003a). As we did for anther exsertion in 2001, the composite population formed from the selected and control lines was placed in the field in the summer of 2002 to measure selection with this increased variation in anther exsertion; results await completion of microsatellite paternity analysis. A final possible explanation for maintenance of anther dimorphism is that it is constrained by genetic correlations with other traits under selection. To test this, we measured many traits on a sample of about 630 plants from the selected and control lines to look for evidence of correlated responses to our artificial selection. This large dataset is still being analyzed.

Mechanisms of genetic correlations: A critical piece of information for understanding the evolution of correlations is the genetic mechanism causing the correlation. However, little is known empirically about the mechanisms that generate genetic correlations. To determine the mechanisms of correlations, we subjected two replicates of 300 plants each to nine generations of random mating using hand-pollination in the greenhouse. Means, correlations, variances, and covariances were all remarkably stable over the nine generations, despite considerable statistical power to detect changes (Conner 2002). This strongly suggests that the correlations are caused by pleiotropy rather than linkage disequilibrium, despite evidence for correlational selection acting on the filaments and corolla tube.

QTL mapping: Our random-mating experiment shows that the strong correlation between filament and corolla tube is due to pleiotropy. Quantitative genetic theory suggests that a strong pleiotropic correlation like this should constrain the independent evolution of the two correlated traits. However, our artificial selection experiment clearly demonstrates that independent evolution of filament and corolla tube can occur, because this is the only way to get increases and decreases in anther exsertion. Thus our two sets of traditional, statistical quantitative genetic results disagree. Despite the great power and utility of statistical quantitative genetics, it treats the genome as a black box; it does not provide information on how many gene loci are affecting the traits, where those loci are located in the genome, what the allele frequencies are at those loci, and what proteins the loci code for. The newer technique of QTL mapping is a first step at breaking open the black box of the genetics of quantitative traits, and should provide the means of resolving the discrepancy between our two sets of results in wild radish.

My long-term goal is to use QTL mapping and other molecular techniques to provide a comprehensive view of the evolution of genetic correlations at three hierarchical levels. At the microevolutionary level, the genetic mechanism underlying floral correlations within a natural population will be determined. QTL mapping has rarely been used within a natural population, which is the unit within which evolution by natural selection occurs. At the interface between micro- and macroevolution, genes causing the difference in anther exsertion between our divergent selected lines are currently being mapped. These lines are experimental analogues of natural populations undergoing speciation by disruptive selection. This work is in collaboration with my MSU colleagues Alan Prather and Jim Hancock, and is funded by USDA, MSU, and the Michigan State Agricultural Experiment Station We have measured floral traits on over 500 F2 progeny resulting from a cross between our selection lines. We are currently genotyping these F2 plants at 15 microsatellite markers that differ between our parental plants and will use approximately 60 additional AFLP markers. Finally, at the macroevolutionary level, the genes that cause differences in correlation patterns between closely related species will be mapped. Based on current information, I plan to cross Barbarea vulgaris and B. verna (we have successfully crossed these species). The former shows the high filament - corolla tube correlation, while the latter does not. This will allow me to identify gene regions that have caused the evolution of high correlation from the lower ancestral correlation.

Field measurements of genetic variances, covariances, and selection on breeding values: A quantitative-genetic half-sibling analysis of seven floral traits and lifetime female fitness was conducted in the field at Kellogg Biological Station. Four major conclusions can be reached from the results: (1) There was significant additive genetic variance for lifetime female fitness (seed production); this is an important evolutionary parameter that has rarely been estimated in the field. (2) Phenotypic selection analysis corroborated female fitness results from Illinois; that is, selection for increased flower size and number through female fitness differences. Selection analysis based on breeding values showed that this selection was real, not just an artifact of environmental correlations. (3) Flower number should respond to this strong selection, since there was significant genetic variance for flower number in the field and no strong genetic correlations with other measured traits. (4) Heritabilities of the floral traits were much lower in the field compared to earlier greenhouse estimates (Conner and Via 1993); this was due to both increased environmental variance and decreased additive genetic variance expressed in the field (Conner et al 2003b).

Summary: The studies outlined above integrate genetics and ecology into studies of the evolution of complex traits. They also span the range from microevolutionary studies within populations to macroevolutionary studies of genetic differences between related species. The results will be critical to our understanding of floral evolution in the model plant family Brassicaceae. More importantly, these results should improve our understanding of how genetic correlations affect the integration and evolution of complex traits in any species, using floral morphology as a model system. Finally, because wild radish is an economically important weed worldwide, our work may have agricultural applications as it addresses mechanisms of rapid adaptation of a weed to novel environments.

2. Evolutionary ecology of plant-animal interactions

With several different collaborators, I am working on a variety of projects addressing the evolution of both plant-pollinator and plant-herbivore interactions. Some of this work addresses how these interactions are being altered by anthropogenic environmental change.

Genetic variation in induced resistance to herbivory: Using quantitative genetic techniques, Anurag Agrawal (University of Toronto) and I measured genetic variation in induced resistance to herbivory. We found evidence for genetic variation in induction of gluconsinolates, a defensive chemical, and corresponding evidence for genetic variance in induced resistance to cabbage butterfly herbivory (Agrawal et al. 2002). This plasticity (induced defense) was costly to the plants, however, as families with higher plasticity had lower fitness in the absence of herbivory.

Evolution of tolerance to herbivory: Plant responses to herbivory can be divided into two components, plant defenses and tolerance. Tolerance refers to the ability of plants to grow and reproduce following herbivore damage; i.e., the ability to compensate. Most studies to date addressing the evolution of plant resistance have focused on defense, while much less is known about how tolerance evolves. My colleague, Ken Paige (University of Illinois), has studied a special case of tolerance in scarlet gilia (Ipomopsis aggregata). He has found several populations of scarlet gilia that overcompensate, that is, have higher lifetime male and female fitness when eaten by deer and elk. However, other populations of scarlet gilia studied by Paige and others have lower fitness after ungulate herbivory. In our collaborative project, funded by NSF, we are using reciprocal transplants among two pairs of populations (one that does and one that does not overcompensate in each pair) to determine the relative importance of genetics versus the environment in determining differences in tolerance. In addition, we will test for differences among the four populations in natural selection on traits associated with tolerance.

Effects of insect herbivory on plant-pollinator interactions: Sharon Strauss (UC Davis) and I studied three-way herbivore-plant-pollinator interactions in wild radish (funded by NSF). In early work, we have found that the consumption of leaves by herbivores reduces flower size, attractiveness to insect pollinators, and pollen production (Strauss et al. 1996). We subsequently estimated the effect of herbivory on seed-siring ability and total male fitness in two large field experiments, using the same molecular genetic paternity analyses that we have used in our studies of selection on floral traits. We found no effect of herbivory on male fitness when plants were in potted arrays in the field, and an increase in male fitness when plants were grown from seed in the field (Strauss et al. 2001). A greenhouse experiment showed no effect of herbivory on pollen competitive ability. These results are surprising given the reduction in floral traits due to herbivory cited above, and may reflect differences in allocation to different components of male and female fitness under different conditions. The results of this work should be of interest to a broad spectrum of ecologists and evolutionary biologists, because the vast majority of herbivory studies to date have measured only effects on female fitness.

Effects of enhanced ultraviolet radiation on plant-pollinator interactions: My lab group, in collaboration with Gene Robinson of the University of Illinois and Jim Cane of Auburn University, studied the effects of increased UV-B radiation on plant-pollinator interactions and total population fitness (funded by USDA). Our goal was to understand how destruction of the ozone layer will affect the crucial mutualistic relationship between plants and their insect pollinators. We exposed populations of field mustard (Brassica rapa) and black mustard (B. nigra) to control and enhanced doses of ultraviolet radiation, at levels designed to mimic current and possible future conditions on the earth. In studies of plants that were grown under greenhouse conditions, lifetime seed production actually increased under increased UV-B in three out of four cases, with little effect on seed quality (Feldheim and Conner 1996). In a parallel field study, however, increased UV-B caused significant declines in seed production in one of the two species, and declines in seed quality in both (Conner and Zangori 1997). Therefore, under more stressful field conditions, Brassica is apparently less able to handle UV-B than in a more benign greenhouse environment. However, in subsequent greenhouse studies we found no evidence for interactions between UV-B and either water and nutrient stress in Brassica (Conner and Zangori 1998) or intraspecific competition in Phacelia (Conner and Neumeier 2001). These studies also found that the other stresses had a much greater effect on plant fitness than did UV-B. Future work could include applying the techniques that we use in our floral evolution work to determine the potential for evolutionary adaptation to UV-B.Fig

Last Updated on Friday, 01 October 2010 19:21