Fashion largely determines what is studied in (aquatic) toxicology

I finished as editor-in-chief of Aquatic Toxicology at the end of July after 14 years. During that time, I have handled more than 6000 manuscripts, which makes it possible to evaluate, what the scientists are studying. It also gives an indication about what is funded, since adequate funding is a prerequisite of being able to carry out the research. Overall, I must say that I am disappointed, since it appears that the funders mainly support fashionable topics, and scientists are naturally willing to do what gives them funding. Associated with this is the positive correlation between fashionable topic and the impact factors of journals. If you have many articles in a journal on a fashionable topic, its impact factor increases, even if the real environmental relevance of the work were poor. I present some major problems, which are the result of trying to do fashionable things instead of thinking already at the outset, what the real environmental relevance of the studies is.

First, it seems that using the newest possible methodology enables you to do work, which has little importance, and still get funded. In the past 15 years the -omics methods have increasingly been utilized by environmental toxicologists. Although they give new possibilities, if properly interpreted and utilized, their improper use is common, and many conclusions are faulty. Most studies use a very small number of organisms, typically 3. This is far too small number for any conclusions with natural populations of animals, especially as their environmental responses may involve changes in variability. I suppose everybody accepts that the responses to toxicants depend on the functions of proteins and their disturbances. Yet, most studies forget this, and based on real-time PCR, microarray or RNA-seq data, which show an increase in steady-state mRNA level conclude that the function encoded by the gene studied has increased. However, this need not be so: if the protein activity decreases because of the action of a toxicant, transcription is increased as a compensatory response. Yet, even after the compensatory response, protein activity may be reduced. In fact, some studies have seen this happening, but have not indicated this obvious explanation just being surprised of the finding. While the above concerns the commonly used transcriptomics, one can find problems with proteomics and metabolomics also. Basically, since toxicants can only affect organisms, if they disturb some functions, functional measurements are required. The -omics data help in finding the genes and consecutively functions, which may be affected. The reason why this is seldom done is twofold: functional measurements are time-consuming, and it is hard to make them high-throughput; the methodology is usually classical and does not attract funders as the use of fancy methodology does.

Second, nanotoxicology was in fashion a couple of years back. Between 2010 and 2015 one could publish virtually anything showing that nanomaterials can be toxic. In most cases, the amount of nanomaterial used has little bearing to what the environmental levels are or may be in future. Yet, relating the toxic actions to nanoparticles to their environmental occurrence is virtually undone. I fear that the same is happening with the new fashionable topic: microplastics. Horror stories are told about the effects of microplastics. Yet, virtually nothing is known about the effects of environmentally occurring levels of microplastics on the function of organisms.

Third, climate change and interactions of toxicants with temperature or oxygen level, or other environmental variables has hitherto been understudied. This knowledge gap is presently being filled. However, a significant problem remains, and most studies do not even indicate its existence. The studies are typically short, often 1-20 days, and the temperature, carbon dioxide or oxygen level change are typically imposed with virtually no lag time using values expected to occur a hundred years from now. This means that the stress levels in the studies are completely different from naturally occurring ones.

Finally, we are suffering from the tyranny of the mean. Virtually always a toxicological response is considered to be a change in the mean of a parameter. Changes in variance are virtually never considered as a toxicological endpoint. One is considering the heterogeneity of data only as determining if data transformation is needed for statistical testing. Yet, when I went over many toxicological studies, I observed that in most of them variability changed without a change in mean. In those cases, variability is undoubtedly a more sensitive indicator of a toxicological response than the mean. We have pointed out the possible importance of variability as a toxicological end point (Nikinmaa, M., Anttila, K. Individual variation in aquatic toxicology: not only unwanted noise. Aquatic Toxicology 207, 29-33; open access)