Evolution of specialization in a plant-microbial mutualism is explained by the oscillation theory of speciation

This is a compilation of the different sets of data that were gathered and analyzed to understand the evolution of plant-microbial mutualisms in a genus of native legumes in California (Acmispon) described in the paper: "Torres-Martínez L, Porter SS, Wendlandt CE, Purcell J, Ortiz-Barbosa GS, Rothschild J, Lampe M, Warisha F, Le T, Weisberg AJ, Chang JH, Sachs JL. Evolution of specialization in a plant-microbial mutualism is explained by the oscillation theory of speciation. Evolution. 2021 Mar 30. doi: 10.1111/evo.14222". Here we use a multidisciplinary approach were niche modeling, phylogenetic approaches and a manipulative green house experiment were implemented to test two specific hypothesis on the evolution of specialization: whether generalist interactions evolve to become more specialized or if partner switches evolve without any change in niche breadth.

First, we assesed the rhizobia taxa that associate with Acmispon spp in nature by performing field collections of root nodules,bacteria isolation in cultures and subsequent sequencing to identify the genus of each isolate. Then, to evaluate how symbiont specificity has evolved within the genus we reconstructed the phylogeny of Acmispon using available sequence data and performed ancestral trait reconstruction ("SymbiontSpecificity").

Second, we collected and genotyped multiple maternal lines of two sister Acmispon spp ("RADseq"), A. strigosus and A. wrangelianus, within their California range and evaluated phylogenetic relationships among the different lines to test for phylogenetic signal on symbiosis specificity.

Third, to gather information on symbiosis specificity of the two sister species to their mutualist partners ("SymbiosisTraitData") we performed a reciprocal inoculation experiment where rhizobial strains of the Bradyrhizobium genus and that associate with A. strigosus in nature were inoculated to multiple maternal lines of both A. strigosus and A. wrangelianus under controlled green house conditions. Strains of the Mesorhizobium genus and that associate with A. wrangelianus in nature were also inoculated in the same maternal lines of both host species. Multiple symbiosis traits were measured to asses specificity, including the probability of nodulation, the number of nodules formed, the investment into nodulation and the relative host growth.

Lastly, we evaluated the levels of specialization of the two Acmispon host species along two abiotic niche axis ("Niche Modelling"), climatic and edaphic. We extracted climatic and soil data from occurrence records of the species and performed a principal component analysis to estimate the environmental space that the species occupy.

The main outcomes of these analyses were: 1) Most Acmispon spp exclusivery associate with one genus of rhizobia, but some exceptions exist and switches only occurred between Mesorhizobium and Bradyrhizobium and no other rhizobial genus. 2) A. strigosus and A. wrangelianus retain the ability to associate with ancestral rhizobial partner taxa, although this trend was not phylogenetically conserved across genotypes. Instead, genetic variation in specificity traits was observed within each host species. 3) A. strigosus genotypes tended to be more generalists than A. wrangelianus genotypes, suggesting asymetry along the mutualist niche. 4) Niche breadth of the climatic and edaphic axes was also broader in A. strigosus than A. wrangelianus. Together these findings, supported the oscillation mode of specialization.

Methods
The field collections of the rhizobial partners of Acmispon spp were performed across California at multiple UC Reserves over the spring of 2019 (February-March). The RADseq data was obtained from maternal lines grown under green house conditions. Library preparation was done in the laboratory of Dr. Jessica Purcel at the University of California, Riverside. Library sequencing was performed at UC Berkeley. Data analyses of RADseq reads were performed using Stacks. Trait data was analyzed through a series of generalized linear mixed models in R. For specific details about the methods implemented for each each data set please refer to the published manuscript in Evolution.