Spatiotemporal variation in guppy colour

One of the core processes underlying evolution is natural selection. Natural election is a process by which organisms adapt to their environments, and is why we have the phenotypic variation (a.k.a. biodiversity) we observe today. These days, we have literally thousands of estimates of selection for all sorts of organisms, and there are now several meta-analyses in the literature examining patterns of phenotypic selection. One of the patterns considered is temporal variation in selection; some analyses argue that there is considerable temporal variation in selection, whereas others argue there is very little. So which is it, and more importantly, how important is this temporal variation in our understanding of adaptation and evolution?One more thing to consider with temporal variation in selection is how important is this variation in comparison to other patterns of selection, with one obvious consideration being spatial variation.

Typically, when considering patterns of selection, we estimate selection by considering a trait under selection and a fitness measure. However, there are logistical problems with this. For example, what’s the best way to measure fitness? Survival? Sure, but if you don’t reproduce, is survival then a good estimate of fitness? Fecundity? Great, but what if none of your offspring make it to maturity – is fecundity then a good estimate of fitness? As you can see, quantifying fitness can be difficult, especially in nature.

An organism in nature. Is it fit or not?
Photo: Kiyoko Gotanda

So what’s an evolutionary biologist to do? How about considering the consequences of selection? If selection is ‘doing its job,’ then we should see changes in the trait under selection, right? If there is selection for larger body size within a population, then we would expect that population to evolve toward an increased mean body size. So, instead of trying to estimate selection, what if we looked at the phenotypic trait that is purportedly under selection? This might be handy because (1) quantifying a phenotypic trait through time is easier than trying to accurately estimate fitness in natural populations, (2) we can consider the importance of temporal variation in selection, because if selection is acting on a trait, and that selection is varying, then it’s reasonable to think that the trait would also vary through time, and (3) compare temporal variation to spatial variation to get a sense of the relative importance of variation in time and space. Thus, we decided to go out and measure spatiotemporal variation in an adaptive phenotypic trait.

First, we needed a system where we have some a priori knowledge about selection and adaptive traits. If we wandered into a random system and quantify a phenotypic trait, we would have no idea if that particular trait was an adaptive trait, nor if and where selection was acting, etc. We’re pretty lucky that there are now several systems we can choose from in which we know that adaptive traits are under selection; the system we chose is male colouration in the Trinidadian guppy system. For those unfamiliar with the guppy system, I refer you to our previous posts here, here, here, and here.

Lovely Trinidadian rainforests
Photo: Kiyoko Gotanda

So we trudged through the gorgeous rainforests of northern Trinidad, caught guppies, and photographed them. We sampled ~20 male guppies annually for six years at six different sites within two rivers. Within the Marianne River, we selected two high-predation (HP) sites and two low-predation (LP) sites, and within the Paria River, we selected two sites, both LP. Male guppies were photographed, and then I (with the help of several audiobooks) quantified their colour patterns from the digital photographs.

Colourful guppies

So what did we find? Not surprisingly, LP sites had more colour than HP sites. I know, you can stifle your yawns: no surprise there. Are you surprised? Here's some background information on why it's not a surprise. Two things about this were interesting, though. First, everyone assumed guppy colour was relatively temporally stable since the HP/LP colour divergence has been demonstrated many times. However, nobody has actually assessed colour annually for several years! If you take a quick peak below, there was some temporal variation, but the LP sites, with two exceptions, basically had more colour than HP sites for six years. The second thing to consider is that there are many potential sources of temporal variation in male guppy colour such as seasonal flooding, variation in parasite loads, and variation in predator composition and density. We don't see any evidence of that here, but this might be due to sampling at the annual level. If we sampled several times a year, we might find a different story. This is a perfect excuse for more trips to Trinidad!

Second, one of the LP sites achieved high colour differently than other sites, and that pattern is relatively consistent through time. Take a look above. The solid, black lines are LP sites, and the dashed lines are HP sites. So, above, you see most of the solid lines above the dashed lines, meaning all the LP sites generally have more colour (carotenoid, structural and melanins) than the HP sites. Now take a look below at graph only showing carotenoid (orange and yellows) colours. You can see one set of the solid, black lines hanging out with  two dashed lines.you've for an LP site hanging out with the HP sites in terms of carotenoid colours.  In other words, one of these LP sites (which happens to be called M16) and HP fish do not have a lot of orange colouration compared to the three other LP sites.

Now check out the structural (violet, silver, and blue) colour graph below. Again, one of the solid black lines seems to be hanging with the dashed line. That solid line happens to be M16 (legend in first graph above) again!  So one LP site (M16) and HP fish have more structural colouration than three other LP sites. So this one LP site (M16) has lots of colour, but unlike the other LP sites, achieved more total colour with more structural colouration and less carotenoid colouration. This pattern holds over six years, so it seems like whatever is happening at M16 isn’t going to be changing anytime soon…

So, what can the spatiotemporal patterns of male guppy colour tell us about selection? We can infer that spatial variation in selection is more important than temporal variation in selection on male guppy colour in our sites over our time frame. Now, before you bombard us with comments about how can we possibly make statements about selection when we didn’t measure selection, let us offer a few of our own comments. First, we recognize that the guppy system might be exceptional in terms of spatial variation in selection vs. temporal variation in selection. The guppy system is known for strong spatial variation in selection on adaptive traits, which is a reason why many researchers gravitate towards guppies for their research. Second, quantification of an adaptive trait is looking at the consequences of variation in selection. That being said, we are confident that we can interpret our results in the context of selection (without having directly estimated selection) because previous work has shown many known selective factors appear to vary more spatially than temporally (e.g. predation regime, canopy cover, and parasitism).  Third, we are not suggesting that our approach should replace the direct estimation of selection, but instead, that it should be used to provide important information that is complementary to our understanding of selection in natural populations.

Go forth, biologists! Estimate selection! Replicate it spatially and temporally! Quantify adaptive phenotypic traits! Replicate it spatially and temporally! Then, write a blog post here about your adventures and what you find. We’ll be waiting patiently!

Reference: Gotanda, K. M., and A. P. Hendry (2014). Using variation in an adaptive trait to consider the potential consequences of temporal variation in selection: spatiotemporal variation in guppy colour. Biological Journal of the Linnean Society 112:108-122.