Novel Cytonuclear Combinations Modify Arabidopsis thaliana Seed Physiology and Vigor

Dormancy and germination vigor are complex traits of primary importance for adaptation and agriculture. Intraspecific variation in cytoplasmic genomes and cytonuclear interactions were previously reported to affect germination in Arabidopsis using novel cytonuclear combinations that disrupt co-adaptation between natural variants of nuclear and cytoplasmic genomes. However, specific aspects of dormancy and germination vigor were not thoroughly explored, nor the parental contributions to the genetic effects. Here, we specifically assessed dormancy, germination performance and longevity of seeds from Arabidopsis plants with natural and new genomic compositions. All three traits were modified by cytonuclear reshuffling. Both depth and release rate of dormancy could be modified by a changing of cytoplasm. Significant changes on dormancy and germination performance due to specific cytonuclear interacting combinations mainly occurred in opposite directions, consistent with the idea that a single physiological consequence of the new genetic combination affected both traits oppositely. However, this was not always the case. Interestingly, the ability of parental accessions to contribute to significant cytonuclear interactions modifying the germination phenotype was different depending on whether they provided the nuclear or cytoplasmic genetic compartment. The observed deleterious effects of novel cytonuclear combinations (in comparison with the nuclear parent) were consistent with a contribution of cytonuclear interactions to germination adaptive phenotypes. More surprisingly, we also observed favorable effects of novel cytonuclear combinations, suggesting suboptimal genetic combinations exist in natural populations for these traits. Reduced sensitivity to exogenous ABA and faster endogenous ABA decay during germination were observed in a novel cytonuclear combination that also exhibited enhanced longevity and better germination performance, compared to its natural nuclear parent. Taken together, our results strongly support that cytoplasmic genomes represent an additional resource of natural variation for breeding seed vigor traits.

Hence the cytoplasm effects in each nuclear background should be interpreted as averaged on the harvests and the shelves.
An effect of the cytonuclear interacting combination was tested for all pairs of cytoplasms (c, c') and for all pairs of nuclei (n, n'). As no third order interaction involving both nucleus and cytoplasm was included in the selected model, cytonuclear interacting combinations could be tested independently from other model terms. Each was tested with the hypothesis where CN states for the model term corresponding to the second-order interaction between cytoplasm and nucleus.
-For 3PH, the selected model was the one with all the first and second order terms plus 4 third order interactions shelf x cytoplasm x nucleus (SCN), harvest x cytoplasm x nucleus (HCN), shelf x harvest x nucleus and harvest x temperature x nucleus.
Effect of each foreign cytoplasm (c) in each natural accession nuclear background (c', n') was Hence the cytonuclear interacting combination effects should be interpreted as averaged on the harvests.
Data from germination performance after storage.
As the salt concentration was adjusted to the tolerance of the nuclear background of the lines, and in order to allow testing of cytonuclear interacting combinations, the data were analyzed considering the salt term was encoded as a binary factor (presence or absence).
The selected model was the one with the first order terms salt, cytoplasm and nucleus plus all second-order interactions, namely cytoplasm x nucleus, cytoplasm x salt and nucleus x salt. As the third-order interaction cytoplasm x nucleus x salt was not included in the selected model, The results were analyzed separately for the two nuclear backgrounds.
For some levels of the aging treatment time length (aatime = 0 and 40) for Ct-1nuclear background, the percentage of germination was exactly equal to 0 or 100%. This led to an unstable estimation of the model parameters due to the data being linearly separable.
Consequently, the data corresponding to these aatime levels were discarded in the analyses for the two nuclear backgrounds. Hence the cytoplasm effect should be interpreted as averaged on the three aatimes.
Germination with exogenous ABA Seed germination was analyzed as previously described (see main text material and methods).
The selected model was the complete model: cytoplasm (C) + treatment (T) + cytoplasm x treatment (CT). Because we were interested in sensitivity to ABA, we tested the interaction term CT for the three cytoplasm pairs (c,c'), and the treatment pair (a,w, for ABA and water), with the hypothesis H0 {(CT ca -CT c'a ) -(CT cw -CT c'w ' ) = 0}

Dosages of ABA content
The data from dry seeds, seeds germinated on water and seeds germinated on salt were analyzed independently. In each case the data from the two independent experiments were jointly analyzed.
For dry seeds and seeds germinated on water, the selected models did not contain the cytoplasm x experiment interaction term, so in these two cases, cytoplasm effects for the three cytoplasm pairs (c, c') could be tested independently from the experiment with the hypothesis

Dosages of ion content
The ratios Na + /K + were analyzed only for NaCl treated seeds, independently at T0 (just after stratification) and at T20 (after 20 hours in germination conditions). The linear models thus only contained the cytoplasm term and the cytoplasm effects were tested for the three pairs of cytoplasms (c, c') with the hypothesis H0 {(C c -C c' ) = 0}