Today we introduce a new guest blogger. Sean Hoban “uses population and genetic models as well as optimization techniques to develop practical guidelines for ex situ conservation of plants for botanic gardens, seed banks, and breeding programs.” Below, in the first of two posts, he sets the scene for us. Tomorrow, he will tell us about recent advances in how to collect maximum seed diversity as efficiently as possible. Welcome to the blog, Sean.
Climate change, species’ extinctions, loss of natural habitats, crop failures… In the new millennium we face many challenges for society and the environment. Large-scale initiatives for collecting and saving seed can help by enabling crop improvement, promotion and sharing of biodiversity, protecting rare species, and restoring ecosystems. If the right genes are in seed collections, plant breeders can make crosses to transfer traits like tolerance to disease and environmental stresses, ultimately boosting crop resilience. Or, if seeds are used for reintroduction, a diversity of genes can help a population cope with environmental change. Maximizing genetic and trait diversity is thus a key task on a seed collector’s mind. At the same time, we don’t have enough resources to visit every location in a species’ geographic range, nor to store immense quantities of seed. Thus the challenge is to get enough of the right diversity in a small-ish batch of seed. Here I’ll discuss progress on this challenge, including recent advances.
Where to sample from, and how much? We need to be efficient, because we have limited resources, but also be effective, because any variation we miss may be lost in threatened natural populations. In the 1970s, a key innovation was made by academic researchers AHD Brown and DR Marshall. They used simple probability models to determine that sampling 30-60 individuals in a population would ensure capturing alleles that were ‘not rare’ (not below 0.05 frequency in the population). The recommendation of ‘approximately 50’ samples quickly caught on, and has been a common basis of protocols ever since, including recent protocols for many botanic gardens, restoration companies, the United States Forest Service and Bureau of Land Management, and others.
Recent work has begun to revisit this decades-old recommendation, and might lead to more refined, and more effective and efficient, collections. It comes down to recognizing that ’50 samples’ may in fact be too few or too many for some species and population.
The Brown & Marshall guideline, as it is often called, has key limitations. Specifically it assumes that every sample will contain a random selection of genes from the population. However, this assumption rarely holds true! For example, when we sample adjacent plants that are relatives we are quite likely to capture some of the same genetic variants rather than new, random ones — a phenomenon called redundancy. Redundancy in a collection means fewer unique genetic variants than expected, meaning that the 50 samples guideline may capture less (or much less) variation than we expect it too. Another limitation: Brown & Marshall recommend sampling every population that can be reached. However, populations differ in size and connectivity — some populations may share a good deal of genetic variation and thus sampling nearby populations may also lead to redundancy.
These limitations have been pointed out repeatedly over the decades by authors (CPC 1991 ((Center for Plant Conservation, CPC. 1991. Genetics and conservation of rare plants, edited by D. A. Falk and K. E. Holsinger. New York: Oxford University Press.)) ; Guerrant 2004, 2014) who say we should sample differently for abundant or common species, or for species whose life history and reproductive strategies are different. After all, these characteristics affect how genetic variation is distributed in landscapes, so they will likely affect how well sampling strategies perform. We may even need to sample differently in different populations — large vs. small populations, for example. Mitchell McGlaughlin and I have recently provided evidence for such assertions, using different approaches. Our works shows precisely how limiting the generic Brown & Marshall strategy can be — and how we can improve our sampling guidelines. I’ll talk about that tomorrow.