Interview with the authors: can we identify the acting selective regime in evolution experiments?

Rapid adaptation to novel conditions is an exciting and growing area in evolutionary research due, at least in part, to our desire to understand the effects of climate change, introduced species, and other conservation-related concerns. However, our ability to detect this evolution is fraught with both biological realities and technical difficulties. A recent paper by Drs. Pfenninger and Foucault, published in Molecular Ecology, illustrate how deep resequencing of replicated experimental populations can fail to provide evolutionary insights, even with extreme selective pressures, due to adaptation to unintentional environmental conditions that overwhelm the genomic signals of the intended selection. This rapid adaptation, in this case to captivity, is an interesting phenomenon that is almost certain to alter other experimental systems, including those that take place in the field. In addition to more details on this fascinating study, the interview with Dr. Pfenninger below also provides an interesting view into technical issues the research team faced: a short time after the initial publication of their manuscript, they discovered a bug in allele frequency calling software that they used! 

Swarming flight of Chironomus midges over a small puddle. Photo credit: Markus Pfenninger.

What led to your interest in this topic / what was the motivation for this study? 

I wanted to know whether rapid adaptation of a natural population to an environmental stressor, in this case temperature, is possible and, if so, by which processes in detail. Apart from this being a fundamental question in population genetics, it is a crucial issue for biodiversity in the ongoing global change.

This is the official and completely true answer – but, to be completely honest, not the entire story: I wanted to see, analyse and prove evolution by natural selection hands-on. Because it’s one thing to teach something gleaned from literature and another to have seen it with your own eyes.

What difficulties did you run into along the way? 

There were actually quite a few: finding a suitable PhD student, technical difficulties with the experimental facilities that almost killed the long term experiment after some months, to mention only the most important ones.

And finally, of course, the almost detrimental issue with a bugged software tool: A few days after the official publication in January, a student reanalysing the data from a different angle, stumbled over unexplainable inconsistencies between the raw data and the allele-frequencies inferred from them. You can imagine the shock it gave me!

When I looked into the problem, I quickly found out that the allele frequencies had little to do with the raw data for most, but perfidiously not all positions in the genome, in particular not the first few on the first scaffold – that’s why the error escaped my attention during a cursory check. The allele frequencies were extracted by a software tool which indeed produced consistently wrong results – a task in principle so simple that systematic checking would have required to write a second tool that exactly did what the first should have done in the first place.

I immediately contacted the authors of the tool and they promptly confirmed that the version we used contained this bug. They did nothing wrong, though. Once they discovered the bug a few months ago, they had promptly updated the tool and documented the error in the release notes. In fact, it appears that the wrong version was on the server for a few days only. Unfortunately it was exactly during the time when we downloaded it – and who looks into the release notes after a tool seemingly did without a hitch what it was supposed to?

I had no choice but to contact the editorial office of Molecular Ecology, informing Genevieve Horn that parts of the publication were flawed and should probably be retracted. At the same time, I started reanalysing the complete data set with a correct version of the tool. Fortunately, after a hard week of number crunching, it turned out that the wrong values were highly correlated in terms of location and allele frequencies to the true values so that some numerical values, but none of the study’s conclusions, needed to be revised. The journal agreed that in this case, a correction article would be sufficient and here it is.

I have to say that everyone, from the software authors to the editor in chief, I have dealt with in this affair has responded greatly and I want to express my deep gratitude here. Given this experience with Molecular Ecology, I can only encourage everyone to address such unfortunate as perhaps unavoidable mistakes immediately and openly.

What is the biggest or most surprising innovation highlighted in this study? 

The rather unsettling major result of the study was the realisation that it is nearly impossible to experimentally manipulate the selection regime of a natural population in a targeted, predictable manner. I think, however, that such “failures” finally advance science by showing which approaches are worth pursuing and which not. Besides this more philosophical aspect showed the study the impressive power of rapid polygenic adaptation.

Moving forward, what are the next steps in this area of research?

I am currently moving into analysing population genomic time series from the field to get an idea on the selective forces acting on natural populations.

What would your message be for students about to start developing or using novel techniques in Molecular Ecology?

Have a good plan, be ready to revise it once the plan meets reality and be prepared for setbacks, remain critical about your results and incorporate appropriate controls. But perhaps most importantly, always take your time to think what you are currently doing and what should be the next steps.

What have you learned about methods and resources development over the course of this project? 

Obviously to even more thoroughly back-check every single analysis. Beyond this, I realised the value and potential of population genomic time series analysis.

Describe the significance of this research for the general scientific community in one sentence.

An evolutionary experiment tells you something about the experiment – not necessarily about nature.

Describe the significance of this research for your scientific community in one sentence.

The acting selective regime in evolutionary experiments is difficult to predict and to manipulate – but perhaps it may be inferred from the results.

Pfenninger M and Foucault Q. 2020. Genomic processes underlying rapid adaptation of a natural Chironomus ripariuspopulation to unintendedly applied experimental selection pressures Molecular Ecology 59:536-548.

Summary from the authors: Latent Dirichlet Allocation reveals spatial and taxonomic structure in a DNA-based census of soil biodiversity from a tropical forest

Biodiversity inventories can now be built by collecting and sequencing DNA from the environment, which is not only easier, faster and cheaper than direct observation, but also much more comprehensive and systematic. This gives in particular unprecedented access to little-known microbial diversity. Tapping these data to answer community ecology questions, however, can prove a daunting task, as classical statistical approaches often fall short of the size and complexity of molecular datasets. To uncover the spatial structure of soil biodiversity over 12 ha of primary tropical forest in French Guiana, we borrowed a probabilistic model from text analysis. After demonstrating the performance of the method on simulated data, we used it to capture the co-occurrence and covariance patterns of more than 25,000 taxa of bacteria, protists, fungi and metazoans across 1,131 soil samples, collected every 10 m – a dataset that led to a previous publication in Mol. Ecol. (Zinger et al. 2019). We find that, even though the forest plot is at first sight rather uniform, bacteria, protists and fungi are all clearly structured into three assemblages matching the environmental heterogeneity of the plot, whereas metazoans are unstructured at that scale. We then work though the practical problems ecologists may encounter using this approach, such as whether to use presence-absence or read-count data, how to choose the number of assemblages and how to assess the robustness of the results. Finally, we discuss the potential use of related methods in community ecology and biogeography, and argue that probabilistic models are a way forward for analyzing the ever-expanding amount of data generated by the field.

Left: Primary tropical forest understory on the plot where the data were collected, Nouragues ecological research station, French Guiana. Right: Spatial distribution of assemblages of co-occurring soil taxa (OTUs), obtained by Latent Dirichlet Allocation from OTU presence-absence information only, over a 12-ha plot of plateau forest sampled every 10 m (top); and two main axes of environmental variation over the same forest plot, derived from Airborne Laser Scanning (bottom). Bacteria, protists and fungi exhibit a spatial pattern matching the environmental heterogeneity of the plot: the blue, green and red assemblages match the terra firme, hydromorphic and exposed rock parts of the plot, respectively. In contrast, metazoans such as annelids can be shown to be spatially unstructured at that scale. Sampled locations are indicated by dark dots, and values have been interpolated between samples using kriging.

Full article:
Sommeria-Klein G, Zinger L, Coissac E, et al. (2020). Latent Dirichlet Allocation reveals spatial and taxonomic structure in a DNA-based census of soil biodiversity from a tropical forest. Molecular Ecology Resour. 20:371–386.

Zinger, L., Taberlet, P., et al. (2019). Body size determines soil community assembly
in a tropical forest. Molecular Ecology, 28(3), 528–543.