How does a plant become toxic, and why? It might seem obvious a plant becomes toxic to defend itself, but investing in chemical defences isn’t cheap. So it is thought that different organs will have varying toxicity, depending on their value to the plant, and the likelihood of attack.

Karen Martinez-Swatson and colleagues investigated the defences of Thapsia garganica, the deadly carrot. The common name might sound like a joke, but the plant’s poisons have a serious effect on anything foolish enough to eat it.

“The anti-herbivory compounds from Thapsia garganica have been used for millennia,” co-author on  the paper, Christopher Barnes, said. “There are records of ancient Romans using the plant for dieting, and with the sickness and diarrhea that thapsigargin induces, it was probably pretty effective (but definitely not recommended). More recently, mipsagargin (a prodrug of thapsigargin) was being used in clinical trials to treat skin cancer. Using laboratory experiments, there is also evidence of thapsigargin is also extremely toxic to many different Eukaryotes.”

The project is a collaboration between many authors, instigated by Martiz-Swatson, Barnes explained. “The project was derived from Karen Martinez-Swatson’s ERC funded PhD work into the whole Thapsia genus, as they all produce thapsigargins. She was interested in the variation in defence compounds between species, but was also finding huge differences in thapsigargins within single species.”

“Its use in both nature and society is what drew my attention to the plant,” Barnes added. “With my background in ecology, I was interested in why the same species in relatively close proximity could vary so much in their defence compounds. After reading the literature into this within species variation in plant chemical defences, I found that there were separate models for different components of plant variation (for example between different tissue types and temporal variation), and even multiple models predicting the same thing that would give different predictions. After some confusion, we decided it would make more sense to test multiple models simultaneously and see which gave the best results.