Reproductive Isolation in Polyploid Complexes

Change in ploidy has long been considered tantamount to instant reproductive isolation – even instant speciation! This idea is well-supported in diploid-tetraploid systems, but what about higher ploidy levels (hexaploid, octoploid, etc)?

New Phyt Fig 3c
Germination success of 2X-4X and 4X-6X interploid crosses.

My work demonstrates that, in a polyploid complex containing diploids, tetraploids, and hexaploids, reproductive isolation is much lower between the two polyploid cytotypes (tetraploid and hexaploid) than between diploids and tetraploids.

This decreased reproductive isolation is found in interploid F1 crosses (Sutherland and Galloway, 2017), in interploid backcross (Sutherland et al., in prep), and in natural mixed-ploidy populations (Sutherland et al., in prep).

Polyploid Mating System Evolution
Although work in other systems has shown that polyploidy can result in increased self-compatibility, Campanula rotundifolia has been considered universally self-incompatible. However, anecdotal accounts describe selfed seed found in cultivated and natural populations.

So what is correct?

Both! Self-incompatibility is affected by both polyploidy and demographic history in C. rotundifolia (Sutherland, Quarles, and Galloway, 2017): Central European diploids and tetraploids are self-incompatible, Western European tetraploids and hexaploids have some self-compatibility, and North American plants can be completely self-compatible!

AJB Fig 2a
Index of self-incompatibility (ISI) of C. rotundifolia populations. ISI of 1 = completely self-incompatible.

Supervised Learning Approaches to Polyploid Estimation
As genome and transcriptome sequencing becomes cheaper and bioinformatic techniques improve, it is becoming possible to not only identify extant polyploids, but also previous rounds of whole-genome duplication in organisms that are currently diploid. However, current methods of identifying such ancient polyploidy are slow, poorly standardized, or overestimate the number of ancient genome duplications. As a postdoc in the Barker Lab at the University of Arizona, I am working on supervised-learning techniques to quickly and reproducibly estimate ancient genome duplication events from a wide range of eukaryotic transcriptomes.

AJB Fig 2a
Early testing of training set size. Pairs of simulated genomes, one positive and one negative for an ancient genome duplication event. Simulated datasets containing 20,000 to 100,000 pairs were used to train models, which were then tested for accuracy against real plant and insect transcriptomes.