Other Research I've done, and sometimes still do...

The wild oat project is an experimental approach to understanding adaptation across heterogeneous environments. I've worked with a number of other approaches over the years.

Surveys of molecular variation among populations of pine

Identifying targets of diversifying selection from comparisons and contrasts among genes and traits

Theoretical models of adaptation to heterogeneous environments.

 

Surveys of molecular variation among populations of pine

Gene flow is the exchange of genes between populations by migration. It is an important evolutionary force in that it tends to homogenize populations and prevent their divergence into separate lineages. It is common to infer patterns of restricted gene flow from the patterns of genetic difference between populations, but this process assumes that populations are at equilibrium between gene flow and drift. Recently, there has been interest in the degree to which past events affect the present day pattern of genetic variation.

Some of my earlier work focused on pines, which provide a unique opportunity to examine gene flow through seeds and pollen separately. The Pinaceae show paternal inheritance of chloroplasts (cpDNA) and maternal inheritance of mitochondria (mtDNA). At several spatial scales, and in two different species, Jeff Mitton and I showed that cpDNA exhibits much less spatial structure both within and among populations than does mtDNA, supporting the hypothesis that wind dispersed pollen is the main agent of gene flow among extant populations. In an area in Montana, near the continental divide where two historically separate populations of ponderosa pine (Pinus ponderosa) have come back into contact since the Wisconsin Ice age (Map, Left) both maternally and paternally inherited markers show steep clines through this zone, indicating little gene flow since secondary contact. The mtDNA cline is particularly steep indicating that once the trees colonized an empty site, little seed exchange with neighboring sites took place.

But allozyme (nuclear) markers show very little differentiation - a pattern that would normally be associated with widespread gene exchange. Along an elevational gradient in Limber pine (Pinus flexilis), we found a similar conundrum - but in that case, the gene flow by pollen appeared to be quite high. The allozymes still showed very little differentiation, it was RAPD markers that showed sharp divergence and that were inconsistent with the patterns of pollen and seed flow inferred from cpDNA and mtDNA. This implies that something interesting is afoot. Either the populations are out of equilibrium between migration and drift, or some of the markers are under natural selection.

Identifying targets of diversifying selection from comparisons and contrasts among genes and traits

It is common to infer the action of selection, by noting that migration and drift will be the same for all the nuclear loci - this means that they are all expected to have the same pattern, and any locus that stands out as different from the rest might be the target of selection. So for example, in ponderosa pine, where most allozyme loci show very little difference between western (var ponderosa) and eastern (var scopulorum) varieties, the few loci that are strongly diverged might be candidates for diversifying natural selection

The problem, is that migration and drift are random processes, so there's a lot of variation in the patterns seen at different loci. This variation can be explicitly modeled using an approach known as the coalescent, which simulates the time to the most recent common ancestor of all the alleles in your sample. Two loci in the same populations can have vastly different evolutionary histories (measured as the time to their most recent common ancestors), and therefore can have very different patterns of genetic variation (Figure, Left). This is particularly pronounced when the populations have not yet come into equilibrium with the level of gene flow.

Against this heterogeneous background, it is very difficult to pick out the true targets of selection. Because we know something of the pattern of history (from pollen cores and packrat midden remains) and gene flow (see below) of ponderosa pine, I had a pretty good picture of the evolutionary scenario to simulate that was independent of the allozyme genetic data I wanted to analyze. When I simulated the distribution of Fst for two isolated populations, it turns out to be highly skewed with a long tail. The observations (top) fit this expected distribution pretty well. Because there is no difference between the observed distribution and the distribution expected under neutral processes, I conclude that we have no evidence for diversifying selection on these loci. Note that this is not the same thing as saying that selection doesn't occur, it just means that we haven't got any information about what it's acting on.

This approach has been extended to quantitative traits. Sewall Wright (1952) worked out the relationship between patterns of divergence at selectively neutral genes, and that at the quantitative trait they influence. So it is (conceptually) fairly simple to make an estimate of divergence from quantitative traits, and compare that with the expectation for neutral genes. (Logistically it is a good bit harder than it sounds, because one has to raise a large [1000's] number of individuals in family structured groups all properly randomized in a common garden experiment). The general pattern is that quantitative traits tend to be more diverged than single electrophoretic markers. This implies that quantitative traits tend more often to diverge in response to natural selection across heterogeneous habitats than do single genes.

But Qst as it's called, is subject to the same caveat as single loci. The distribution of Qst values under neutrality is just as highly variable as single loci. So picking out a particular trait as being the target of diversifying selection is just as fraught with error as for single genes (above). The general pattern is robust, because the overall distribution of Fst and Qst are different. The difficulty lies in identifying a particular trait as the target of selection, just because it lies in the tail of the distribution.

Theoretical models of adaptation to heterogeneous environments.

Since most of the traits that are likely to adapt populations to their local environments are quantitative traits affected by multiple genes, I began to wonder what pattern of genetic variation we might expect at quantitative trait loci - the individual genes that affect quantitative traits. I conducted a simulation of selection acting on a simple additive polygenic trait, with the optimum trait value varying among populations (ie, high trait values were selected for in one environment, medium in others, and low values favoured in still other populations). Surprisingly, the QTL's showed exactly the same pattern as neutral markers, even though the average trait value diverged in response to selection. (Or to put it technically, Fst of QTL;s matched that of neutral markers even though Qst responded to selection). This paradox was resolved by noticing that the variance of the trait had a component of linkage disequilibrium associated with it. With negative disequilibrium (a negative association between the allelic effects at different loci) variation in the two loci tend to cancel out - With positive disequilibrium, they re-inforce one another (Figure right). This may go some way to explaining how populations might adapt to local environments in the face of high gene flow.

It will take a lot of experimental work to figure out whether this is how populations actually adapt in the real world. But a recently published paper by Par Ingvarsson's group in Sweden seems to suggest they might. They study the adaptation of phenology (the timing of annual events like leaf emergence) in European poplars (Populus trichocarpus) to the local climate in which the population lives. They showed that one of the candidate genes underlying phenology was much less diverged along a North-South gradient than is phenology itself.

other research

Limber Pine at High Elevation (Sheep Mtn, Colorado)

Ponderosa Pine in Boulder Mtn Park, Boulder, Colorado

Juniper in Arches NP, Utah

Alpine Fir at treeline, Chicago Basin, Colorado

Limber Pine at Low Elevation, (Vedawoo, Wyoming)