Proc. B Special Issue: Can Random Processes Drive Parallel Evolutionary Responses to Cities?

Continuing our coverage of the recent Proc. B Special Issue on urban evolution, James Santangelo (PhD candidate at University of Toronto Mississauga) tells us about his recent manuscript:

One of the outstanding questions in evolutionary biology concerns the extent to which different species — or different populations of a single species — evolve the same genes or traits when exposed to the same environment. This “parallel” (or “convergent”) evolution has important consequences for our understanding of how adaptation proceeds in nature: high levels of parallelism suggest the paths that evolution can take are limited, thereby increasing our ability to predict how species (or populations) will respond to environmental change. Among the many systems where parallel evolution has been studied, urban environments are emerging as models for assessing the ubiquity (or lack thereof) of parallel evolutionary responses among natural populations, due in part to their global replication and relative consistency in the environmental changes they elicit.

When biologists observe the same traits repeatedly evolving across populations, it has traditionally been assumed that evolutionary forces like natural selection are at work, molding populations to match their environments. After all, how could random processes (e.g. genetic drift) generate the same pattern over and over again in independent populations? In our recent paper from the special feature on urban environments, we challenge this assumption and show that, under certain conditions, parallel evolutionary responses can arise out of random evolutionary processes.

Figure_2
Figure 2: HCN results from the interaction of two genes: Ac produces the cyanogenic precursors & Li produces the enzyme that converts the precursors into HCN. Plants need at least one dominant allele (i.e. Ac and Li) at each locus for HCN.

Our work builds on previous work by Thompson and colleagues identifying the repeated loss of hydrogen cyanide (HCN, a potent defense against chewing insects) from urban white clover (Trifolium repens) populations (Figure 1) across multiple cities in eastern North America. After seeing this work presented at a lab meeting, prof. Rob Ness was unconvinced by the “selectionist” (i.e. natural selection) explanation offered by the authors for the repeated evolution of HCN clines, suggesting instead that they had not ruled out a neutral (i.e. genetic drift) explanation.

Rob recognized that the two-locus genetic architecture of HCN made populations particularly susceptible to decreases in HCN frequency due only to random processes (Figure 2). Rob went back to his office and generated a simple simulation: start a single population with equal frequencies of alleles at both genes underlying HCN, continuously sample alleles every generation for multiple generations, and see what happens to the frequency of HCN over time. So, what happens? The frequency of HCN goes down with every passing generation (Figure 3)! Great! Done deal, right? It’s all drift!

Figure_3
Figure 3: The frequency of HCN goes down over time in a population where each generation is produced by sampling alleles at each gene from the previous generation. Consistent with theory, this effect is stronger for small populations due to greater sampling error resulting in more drastic changes in the frequency of alleles. This occurs because random fluctuation at either gene are more likely to result in the loss of HCN from populations (Figure 1).

Not quite…Rob’s simulation lacked the realism of nature. A single population of constant size evolving for a few generations is too simple to be convincing. We needed to add some of the complexity we’ve come to expect of natural systems. So, I expanded Rob’s initial simulation first to include multiple populations along a linear landscape, then to include a gradient in population sizes (i.e. drift) along this landscape, gene flow among populations, and finally varying strengths of selection acting in populations (Figure 4, selection not included in this post). We even simulated yet more complex scenarios involving population colonization and bottlenecks followed by population growth. The main question of the paper became: How do genetic drift, gene flow, and natural selection interact in the formation of spatial clines in HCN? What did we find?

Figure_4
Figure 4: Our simulations involved generating a spatial gradient in population sizes as a way of controlling the amount of genetic drift in populations (smaller population = stronger drift). Gene flow could happen among all populations, every generation.

Surprisingly (or maybe not), our results are largely consistent with Rob’s initial simulation. Under a gradient in population sizes, the strong drift in small populations results in a decrease in HCN frequencies, whereas HCN frequencies remain unchanged in large, weakly drifting populations. The effect is a spatial cline in HCN frequency, with stronger clines resulting from steeper gradients in populations sizes (Figure 5). As expected, gene flow constrains the formation and reduces the strength of clines by homogenizing alleles among neighboring populations (Figure 6). Nonetheless, clines do form and the picture remains the same: gradients in the strength of genetic drift drive the formation of spatial clines in HCN.

Figure_5
Figure 5: Steeper gradients in the strength of genetic drift (i.e. greater difference between the largest and smallest population) resulted in stronger phenotypic clines in HCN (measured as the mean slope across 1000 simulations).
Figure_6
Figure 6: Increasing gene flow weakened the strength of phenotypic clines in HCN, consistent with gene flow homogenizing alleles across populations. However, clines formed even under unrealistically high levels of gene flow (e.g. 0.1 = 10% exchange of alleles among populations).

 

 

 

 

 

 

 

 

 

So, what does this mean for the urban-rural HCN clines observed across cities? We’ve shown that genetic drift can cause clines, but does it? After all, cities are frequently associated with reductions in population sizes and increased strength of genetic drift, exactly the conditions we would expect to drive decreases in HCN frequencies based on our simulations. Despite this, I don’t think drift is the main player here. A separate paper by Johnson and colleagues published in the same special feature found rampant gene flow among urban-rural clover populations and no decreased genetic diversity in urban populations, suggesting no increased drift in the city. Additionally, our simulations showed that while drift can cause clines, these clines tend to be very weak relative to those caused by selection, the latter of which are more consistent in strength with the clines observed by Thompson and colleagues.

Future work sampling more cities and generating genome-wide SNP data for these samples will help better assess the contribution of genetic drift to the evolution of HCN clines and nail down some of the environmental factors responsible for driving reduced HCN frequencies in urban populations. One such effort is the Global Urban Evolution Project, which has solicited help from ~550 scientists and collaborators from around the world to sample clover in their local cities. This large-scale parallel evolution project promises to provide insight not only into the evolution of HCN clines in cities, but more broadly into the ubiquity of phenotypic and genetic evolution in nature.

I’ll end on a cautionary note. Genetic drift is unlikely the main force driving reduced HCN frequencies in urban populations. Nonetheless, urban populations often experience increased drift and the type of non-additive genetics underlying HCN is a common genetic architecture in nature. Whenever traits arise via interactions among multiple genes, chance fluctuations in the frequency of any one gene may result in consistent changes in the corresponding trait. Therefore, in cases of repeated phenotypic evolution in cities (and otherwise), it’s important that drift be ruled out as a viable mechanism driving these parallel responses; parallel phenotypic evolution need not be the result of natural selection.

 


About the author:

James is a PhD candidate in the Department of Ecology and Evolutionary Biology at the University of Toronto with prof. Marc Johnson and prof. Rob Ness. He is broadly interested in parallel evolution — the recurrence of similar phenotypes and genotypes due to adaptation to similar environments or via other evolutionary processes (e.g. genetic drift). He uses white clover evolving in cities as a model for testing fundamental questions about the ubiquity of parallelism in nature. You can read more about his work here and check out some of his code here.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Blog at WordPress.com.

Up ↑

Create your website at WordPress.com
Get started
%d bloggers like this: